# coding=utf-8 import glob, hashlib, os, re, shutil, subprocess, sys import tools.shared from tools.shared import * from runner import RunnerCore, path_from_root, checked_sanity, test_modes, get_bullet_library class T(RunnerCore): # Short name, to make it more fun to use manually on the commandline def is_emscripten_abi(self): return not ('i386-pc-linux-gnu' in COMPILER_OPTS or self.env.get('EMCC_LLVM_TARGET') == 'i386-pc-linux-gnu') def test_hello_world(self): test_path = path_from_root('tests', 'core', 'test_hello_world') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) assert 'EMSCRIPTEN_GENERATED_FUNCTIONS' not in open(self.in_dir('src.cpp.o.js')).read(), 'must not emit this unneeded internal thing' def test_intvars(self): if self.emcc_args == None: return self.skip('needs ta2') test_path = path_from_root('tests', 'core', 'test_intvars') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sintvars(self): Settings.CORRECT_SIGNS = 1 # Relevant to this test Settings.CORRECT_OVERFLOWS = 0 # We should not need overflow correction to get this right test_path = path_from_root('tests', 'core', 'test_sintvars') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, force_c=True) def test_i64(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('i64 mode 1 requires ta2') src = ''' #include int main() { long long a = 0x2b00505c10; long long b = a >> 29; long long c = a >> 32; long long d = a >> 34; printf("*%Ld,%Ld,%Ld,%Ld*\\n", a, b, c, d); unsigned long long ua = 0x2b00505c10; unsigned long long ub = ua >> 29; unsigned long long uc = ua >> 32; unsigned long long ud = ua >> 34; printf("*%Ld,%Ld,%Ld,%Ld*\\n", ua, ub, uc, ud); long long x = 0x0000def123450789ULL; // any bigger than this, and we long long y = 0x00020ef123456089ULL; // start to run into the double precision limit! printf("*%Ld,%Ld,%Ld,%Ld,%Ld*\\n", x, y, x | y, x & y, x ^ y, x >> 2, y << 2); printf("*"); long long z = 13; int n = 0; while (z > 1) { printf("%.2f,", (float)z); // these must be integers! z = z >> 1; n++; } printf("*%d*\\n", n); return 0; } ''' self.do_run(src, '*184688860176,344,43,10*\n*184688860176,344,43,10*\n*245127260211081,579378795077769,808077213656969,16428841631881,791648372025088*\n*13.00,6.00,3.00,*3*') src = r''' #include #include #include int64_t returner1() { return 0x0000def123450789ULL; } int64_t returner2(int test) { while (test > 10) test /= 2; // confuse the compiler so it doesn't eliminate this function return test > 5 ? 0x0000def123450123ULL : 0ULL; } void modifier1(int64_t t) { t |= 12; printf("m1: %Ld\n", t); } void modifier2(int64_t &t) { t |= 12; } int truthy() { int x = time(0); while (x > 10) { x |= 7; x /= 2; } return x < 3; } struct IUB { int c; long long d; }; IUB iub[] = { { 55, 17179869201 }, { 122, 25769803837 }, }; int main(int argc, char **argv) { int64_t x1 = 0x1234def123450789ULL; int64_t x2 = 0x1234def123450788ULL; int64_t x3 = 0x1234def123450789ULL; printf("*%Ld\n%d,%d,%d,%d,%d\n%d,%d,%d,%d,%d*\n", x1, x1==x2, x1x2, x1>=x2, // note: some rounding in the printing! x1==x3, x1x3, x1>=x3); printf("*%Ld*\n", returner1()); printf("*%Ld*\n", returner2(30)); uint64_t maxx = -1ULL; printf("*%Lu*\n*%Lu*\n", maxx, maxx >> 5); // Make sure params are not modified if they shouldn't be int64_t t = 123; modifier1(t); printf("*%Ld*\n", t); modifier2(t); printf("*%Ld*\n", t); // global structs with i64s printf("*%d,%Ld*\n*%d,%Ld*\n", iub[0].c, iub[0].d, iub[1].c, iub[1].d); // Bitshifts { int64_t a = -1; int64_t b = a >> 29; int64_t c = a >> 32; int64_t d = a >> 34; printf("*%Ld,%Ld,%Ld,%Ld*\n", a, b, c, d); uint64_t ua = -1; int64_t ub = ua >> 29; int64_t uc = ua >> 32; int64_t ud = ua >> 34; printf("*%Ld,%Ld,%Ld,%Ld*\n", ua, ub, uc, ud); } // Nonconstant bitshifts { int64_t a = -1; int64_t b = a >> (29 - argc + 1); int64_t c = a >> (32 - argc + 1); int64_t d = a >> (34 - argc + 1); printf("*%Ld,%Ld,%Ld,%Ld*\n", a, b, c, d); uint64_t ua = -1; int64_t ub = ua >> (29 - argc + 1); int64_t uc = ua >> (32 - argc + 1); int64_t ud = ua >> (34 - argc + 1); printf("*%Ld,%Ld,%Ld,%Ld*\n", ua, ub, uc, ud); } // Math mixtures with doubles { uint64_t a = 5; double b = 6.8; uint64_t c = a * b; if (truthy()) printf("*%d,%d,%d*\n", (int)&a, (int)&b, (int)&c); // printing addresses prevents optimizations printf("*prod:%llu*\n", c); } // Basic (rounded, for now) math. Just check compilation. int64_t a = 0x1234def123450789ULL; a--; if (truthy()) a--; // confuse optimizer int64_t b = 0x1234000000450789ULL; b++; if (truthy()) b--; // confuse optimizer printf("*%Ld,%Ld,%Ld,%Ld*\n", (a+b)/5000, (a-b)/5000, (a*3)/5000, (a/5)/5000); a -= 17; if (truthy()) a += 5; // confuse optimizer b -= 17; if (truthy()) b += 121; // confuse optimizer printf("*%Lx,%Lx,%Lx,%Lx*\n", b - a, b - a/2, b/2 - a, b - 20); if (truthy()) a += 5/b; // confuse optimizer if (truthy()) b += 121*(3+a/b); // confuse optimizer printf("*%Lx,%Lx,%Lx,%Lx*\n", a - b, a - b/2, a/2 - b, a - 20); return 0; } ''' self.do_run(src, '*1311918518731868041\n' + '0,0,0,1,1\n' + '1,0,1,0,1*\n' + '*245127260211081*\n' + '*245127260209443*\n' + '*18446744073709551615*\n' + '*576460752303423487*\n' + 'm1: 127\n' + '*123*\n' + '*127*\n' + '*55,17179869201*\n' + '*122,25769803837*\n' + '*-1,-1,-1,-1*\n' + '*-1,34359738367,4294967295,1073741823*\n' + '*-1,-1,-1,-1*\n' + '*-1,34359738367,4294967295,1073741823*\n' + '*prod:34*\n' + '*524718382041609,49025451137,787151111239120,52476740749274*\n' + '*ffff210edd000002,91990876ea283be,f6e5210edcdd7c45,1234000000450765*\n' + '*def122fffffe,91adef1232283bb,f6e66f78915d7c42,1234def123450763*\n') src = r''' #include #include int main() { long long i,j,k; i = 0; j = -1, k = 1; printf( "*\n" ); printf( "%s\n", i > j ? "Ok": "Fail" ); printf( "%s\n", k > i ? "Ok": "Fail" ); printf( "%s\n", k > j ? "Ok": "Fail" ); printf( "%s\n", i < j ? "Fail": "Ok" ); printf( "%s\n", k < i ? "Fail": "Ok" ); printf( "%s\n", k < j ? "Fail": "Ok" ); printf( "%s\n", (i-j) >= k ? "Ok": "Fail" ); printf( "%s\n", (i-j) <= k ? "Ok": "Fail" ); printf( "%s\n", i > std::numeric_limits::min() ? "Ok": "Fail" ); printf( "%s\n", i < std::numeric_limits::max() ? "Ok": "Fail" ); printf( "*\n" ); } ''' self.do_run(src, '*\nOk\nOk\nOk\nOk\nOk\nOk\nOk\nOk\nOk\nOk\n*') # stuff that also needs sign corrections Settings.CORRECT_SIGNS = 1 src = r''' #include #include int main() { // i32 vs i64 int32_t small = -1; int64_t large = -1; printf("*%d*\n", small == large); small++; printf("*%d*\n", small == large); uint32_t usmall = -1; uint64_t ularge = -1; printf("*%d*\n", usmall == ularge); return 0; } ''' self.do_run(src, '*1*\n*0*\n*0*\n') def test_i64_b(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_b') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i64_cmp(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_cmp') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i64_cmp2(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_cmp2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i64_double(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_double') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i64_umul(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_umul') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i64_precise(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') src = r''' #include #include int main() { uint64_t x = 0, y = 0; for (int i = 0; i < 64; i++) { x += 1ULL << i; y += x; x /= 3; y *= 5; printf("unsigned %d: %llu,%llu,%llu,%llu,%llu,%llu,%llu,%llu,%llu\n", i, x, y, x+y, x-y, x*y, y ? x/y : 0, x ? y/x : 0, y ? x%y : 0, x ? y%x : 0); } int64_t x2 = 0, y2 = 0; for (int i = 0; i < 64; i++) { x2 += 1LL << i; y2 += x2; x2 /= 3 * (i % 7 ? -1 : 1); y2 *= 5 * (i % 2 ? -1 : 1); printf("signed %d: %lld,%lld,%lld,%lld,%lld,%lld,%lld,%lld,%lld\n", i, x2, y2, x2+y2, x2-y2, x2*y2, y2 ? x2/y2 : 0, x2 ? y2/x2 : 0, y2 ? x2%y2 : 0, x2 ? y2%x2 : 0); } return 0; } ''' self.do_run(src, open(path_from_root('tests', 'i64_precise.txt')).read()) # Verify that even if we ask for precision, if it is not needed it is not included Settings.PRECISE_I64_MATH = 1 src = ''' #include #include int main(int argc, char **argv) { uint64_t x = 2125299906845564, y = 1225891506842664; if (argc == 12) { x = x >> 1; y = y >> 1; } x = x & 12ULL; y = y | 12ULL; x = x ^ y; x <<= 2; y >>= 3; printf("*%llu, %llu*\\n", x, y); } ''' self.do_run(src, '*4903566027370624, 153236438355333*') code = open(os.path.join(self.get_dir(), 'src.cpp.o.js')).read() assert 'goog.math.Long' not in code, 'i64 precise math should not have been included if not actually used' # But if we force it to be included, it is. First, a case where we don't need it Settings.PRECISE_I64_MATH = 2 self.do_run(open(path_from_root('tests', 'hello_world.c')).read(), 'hello') code = open(os.path.join(self.get_dir(), 'src.cpp.o.js')).read() assert 'goog.math.Long' in code, 'i64 precise math should be included if forced' # and now one where we do self.do_run(r''' #include int main( int argc, char ** argv ) { unsigned long a = 0x60DD1695U; unsigned long b = 0xCA8C4E7BU; unsigned long long c = (unsigned long long)a * b; printf( "c = %016llx\n", c ); return 0; } ''', 'c = 4ca38a6bd2973f97') def test_i64_llabs(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') Settings.PRECISE_I64_MATH = 2 test_path = path_from_root('tests', 'core', 'test_i64_llabs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i64_zextneg(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_zextneg') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i64_7z(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_7z') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, ['hallo']) def test_i64_i16(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_i16') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i64_qdouble(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_qdouble') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i64_varargs(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('full i64 stuff only in ta2') test_path = path_from_root('tests', 'core', 'test_i64_varargs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, 'waka fleefl asdfasdfasdfasdf'.split(' ')) def test_double_varargs(self): test_path = path_from_root('tests', 'core', 'test_double_varargs') src, output = (test_path + s for s in ('.c', '.out')) self.do_run_from_file(src, output) def test_i32_mul_precise(self): if self.emcc_args == None: return self.skip('needs ta2') test_path = path_from_root('tests', 'core', 'test_i32_mul_precise') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i32_mul_semiprecise(self): if Settings.ASM_JS: return self.skip('asm is always fully precise') Settings.PRECISE_I32_MUL = 0 # we want semiprecise here test_path = path_from_root('tests', 'core', 'test_i32_mul_semiprecise') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_i16_emcc_intrinsic(self): Settings.CORRECT_SIGNS = 1 # Relevant to this test test_path = path_from_root('tests', 'core', 'test_i16_emcc_intrinsic') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_double_i64_conversion(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('needs ta2') test_path = path_from_root('tests', 'core', 'test_double_i64_conversion') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_float32_precise(self): Settings.PRECISE_F32 = 1 test_path = path_from_root('tests', 'core', 'test_float32_precise') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_negative_zero(self): test_path = path_from_root('tests', 'core', 'test_negative_zero') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_literal_negative_zero(self): if self.emcc_args == None: return self.skip('needs emcc') test_path = path_from_root('tests', 'core', 'test_literal_negative_zero') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_llvm_intrinsics(self): if self.emcc_args == None: return self.skip('needs ta2') Settings.PRECISE_I64_MATH = 2 # for bswap64 test_path = path_from_root('tests', 'core', 'test_llvm_intrinsics') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_bswap64(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('needs ta2') test_path = path_from_root('tests', 'core', 'test_bswap64') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sha1(self): if self.emcc_args == None: return self.skip('needs ta2') self.do_run(open(path_from_root('tests', 'sha1.c')).read(), 'SHA1=15dd99a1991e0b3826fede3deffc1feba42278e6') def test_asmjs_unknown_emscripten(self): if self.emcc_args == None: return self.skip('needs emcc') if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for asmjs-unknown-emscripten target test') if os.environ.get('EMCC_FAST_COMPILER') == '0': return self.skip('fastcomp needed for asmjs-unknonw-emscripten target') self.do_run(open(path_from_root('tests', 'asmjs-unknown-emscripten.c')).read(), '') def test_cube2md5(self): if self.emcc_args == None: return self.skip('needs emcc') if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for accurate math') self.emcc_args += ['--embed-file', 'cube2md5.txt'] shutil.copyfile(path_from_root('tests', 'cube2md5.txt'), os.path.join(self.get_dir(), 'cube2md5.txt')) self.do_run(open(path_from_root('tests', 'cube2md5.cpp')).read(), open(path_from_root('tests', 'cube2md5.ok')).read()) def test_cube2hash(self): # extra testing for various codegen modes for x86 in [0, 1] if self.run_name == 'asm2' else [0]: print 'x86', x86 try: old_x86 = os.environ.get('EMCC_LLVM_TARGET') or '' if x86: os.environ['EMCC_LLVM_TARGET'] = "i386-pc-linux-gnu" try: old_fastcomp = os.environ.get('EMCC_FAST_COMPILER') or '' if x86: os.environ['EMCC_FAST_COMPILER'] = "0" try: old_chunk_size = os.environ.get('EMSCRIPT_MAX_CHUNK_SIZE') or '' for chunk_size in ['1', old_chunk_size]: # test splitting out each function to a chunk in emscripten.py (21 functions here) print ' chunks', chunk_size os.environ['EMSCRIPT_MAX_CHUNK_SIZE'] = chunk_size # A good test of i64 math if Settings.USE_TYPED_ARRAYS != 2: return self.skip('requires ta2 C-style memory aliasing') self.do_run('', 'Usage: hashstring ', libraries=self.get_library('cube2hash', ['cube2hash.bc'], configure=None, cache_name_extra=str(x86)), includes=[path_from_root('tests', 'cube2hash')]) for text, output in [('fleefl', '892BDB6FD3F62E863D63DA55851700FDE3ACF30204798CE9'), ('fleefl2', 'AA2CC5F96FC9D540CA24FDAF1F71E2942753DB83E8A81B61'), ('64bitisslow', '64D8470573635EC354FEE7B7F87C566FCAF1EFB491041670')]: self.do_run('', 'hash value: ' + output, [text], no_build=True) finally: os.environ['EMSCRIPT_MAX_CHUNK_SIZE'] = old_chunk_size finally: if x86: if old_fastcomp: os.environ['EMCC_FAST_COMPILER'] = old_fastcomp else: del os.environ['EMCC_FAST_COMPILER'] finally: if x86: if old_x86: os.environ['EMCC_LLVM_TARGET'] = old_x86 else: del os.environ['EMCC_LLVM_TARGET'] def test_unaligned(self): if Settings.QUANTUM_SIZE == 1: return self.skip('No meaning to unaligned addresses in q1') src = r''' #include struct S { double x; int y; }; int main() { // the 64-bit value here will not be 8-byte aligned S s0[3] = { {0x12a751f430142, 22}, {0x17a5c85bad144, 98}, {1, 1}}; char buffer[10*sizeof(S)]; int b = int(buffer); S *s = (S*)(b + 4-b%8); s[0] = s0[0]; s[1] = s0[1]; s[2] = s0[2]; printf("*%d : %d : %d\n", sizeof(S), ((unsigned int)&s[0]) % 8 != ((unsigned int)&s[1]) % 8, ((unsigned int)&s[1]) - ((unsigned int)&s[0])); s[0].x++; s[0].y++; s[1].x++; s[1].y++; printf("%.1f,%d,%.1f,%d\n", s[0].x, s[0].y, s[1].x, s[1].y); return 0; } ''' # TODO: A version of this with int64s as well if self.is_emscripten_abi(): return self.skip('LLVM marks the reads of s as fully aligned, making this test invalid') else: self.do_run(src, '*12 : 1 : 12\n328157500735811.0,23,416012775903557.0,99\n') return # TODO: continue to the next part here # Test for undefined behavior in C. This is not legitimate code, but does exist if Settings.USE_TYPED_ARRAYS != 2: return self.skip('No meaning to unaligned addresses without t2') src = r''' #include int main() { int x[10]; char *p = (char*)&x[0]; p++; short *q = (short*)p; *q = 300; printf("*%d:%d*\n", *q, ((int)q)%2); int *r = (int*)p; *r = 515559; printf("*%d*\n", *r); long long *t = (long long*)p; *t = 42949672960; printf("*%Ld*\n", *t); return 0; } ''' try: self.do_run(src, '*300:1*\n*515559*\n*42949672960*\n') except Exception, e: assert 'must be aligned' in str(e), e # expected to fail without emulation def test_align64(self): src = r''' #include // inspired by poppler enum Type { A = 10, B = 20 }; struct Object { Type type; union { int intg; double real; char *name; }; }; struct Principal { double x; Object a; double y; }; int main(int argc, char **argv) { int base = argc-1; Object *o = NULL; printf("%d,%d\n", sizeof(Object), sizeof(Principal)); printf("%d,%d,%d,%d\n", (int)&o[base].type, (int)&o[base].intg, (int)&o[base].real, (int)&o[base].name); printf("%d,%d,%d,%d\n", (int)&o[base+1].type, (int)&o[base+1].intg, (int)&o[base+1].real, (int)&o[base+1].name); Principal p, q; p.x = p.y = q.x = q.y = 0; p.a.type = A; p.a.real = 123.456; *(&q.a) = p.a; printf("%.2f,%d,%.2f,%.2f : %.2f,%d,%.2f,%.2f\n", p.x, p.a.type, p.a.real, p.y, q.x, q.a.type, q.a.real, q.y); return 0; } ''' if self.is_emscripten_abi(): self.do_run(src, '''16,32 0,8,8,8 16,24,24,24 0.00,10,123.46,0.00 : 0.00,10,123.46,0.00 ''') else: self.do_run(src, '''12,28 0,4,4,4 12,16,16,16 0.00,10,123.46,0.00 : 0.00,10,123.46,0.00 ''') def test_unsigned(self): Settings.CORRECT_SIGNS = 1 # We test for exactly this sort of thing here Settings.CHECK_SIGNS = 0 src = ''' #include const signed char cvals[2] = { -1, -2 }; // compiler can store this is a string, so -1 becomes \FF, and needs re-signing int main() { { unsigned char x = 200; printf("*%d*\\n", x); unsigned char y = -22; printf("*%d*\\n", y); } int varey = 100; unsigned int MAXEY = -1, MAXEY2 = -77; printf("*%u,%d,%u*\\n", MAXEY, varey >= MAXEY, MAXEY2); // 100 >= -1? not in unsigned! int y = cvals[0]; printf("*%d,%d,%d,%d*\\n", cvals[0], cvals[0] < 0, y, y < 0); y = cvals[1]; printf("*%d,%d,%d,%d*\\n", cvals[1], cvals[1] < 0, y, y < 0); // zext issue - see mathop in jsifier unsigned char x8 = -10; unsigned long hold = 0; hold += x8; int y32 = hold+50; printf("*%u,%u*\\n", hold, y32); // Comparisons x8 = 0; for (int i = 0; i < 254; i++) x8++; // make it an actual 254 in JS - not a -2 printf("*%d,%d*\\n", x8+1 == 0xff, x8+1 != 0xff); // 0xff may be '-1' in the bitcode return 0; } ''' self.do_run(src, '*4294967295,0,4294967219*\n*-1,1,-1,1*\n*-2,1,-2,1*\n*246,296*\n*1,0*') # Now let's see some code that should just work in USE_TYPED_ARRAYS == 2, but requires # corrections otherwise if Settings.USE_TYPED_ARRAYS == 2: Settings.CORRECT_SIGNS = 0 Settings.CHECK_SIGNS = 1 if not Settings.ASM_JS else 0 else: Settings.CORRECT_SIGNS = 1 Settings.CHECK_SIGNS = 0 src = ''' #include int main() { { unsigned char x; unsigned char *y = &x; *y = -1; printf("*%d*\\n", x); } { unsigned short x; unsigned short *y = &x; *y = -1; printf("*%d*\\n", x); } /*{ // This case is not checked. The hint for unsignedness is just the %u in printf, and we do not analyze that unsigned int x; unsigned int *y = &x; *y = -1; printf("*%u*\\n", x); }*/ { char x; char *y = &x; *y = 255; printf("*%d*\\n", x); } { char x; char *y = &x; *y = 65535; printf("*%d*\\n", x); } { char x; char *y = &x; *y = 0xffffffff; printf("*%d*\\n", x); } return 0; } ''' self.do_run(src, '*255*\n*65535*\n*-1*\n*-1*\n*-1*') def test_bitfields(self): if self.emcc_args is None: Settings.SAFE_HEAP = 0 # bitfields do loads on invalid areas, by design test_path = path_from_root('tests', 'core', 'test_bitfields') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_floatvars(self): test_path = path_from_root('tests', 'core', 'test_floatvars') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_closebitcasts(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('requires ta2') test_path = path_from_root('tests', 'core', 'closebitcasts') src, output = (test_path + s for s in ('.c', '.txt')) self.do_run_from_file(src, output) def test_fast_math(self): if self.emcc_args is None: return self.skip('requires emcc') Building.COMPILER_TEST_OPTS += ['-ffast-math'] test_path = path_from_root('tests', 'core', 'test_fast_math') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, ['5', '6', '8']) def test_zerodiv(self): test_path = path_from_root('tests', 'core', 'test_zerodiv') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_zero_multiplication(self): test_path = path_from_root('tests', 'core', 'test_zero_multiplication') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_isnan(self): test_path = path_from_root('tests', 'core', 'test_isnan') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_globaldoubles(self): test_path = path_from_root('tests', 'core', 'test_globaldoubles') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_math(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('requires ta2') test_path = path_from_root('tests', 'core', 'test_math') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_erf(self): test_path = path_from_root('tests', 'core', 'test_erf') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_math_hyperbolic(self): src = open(path_from_root('tests', 'hyperbolic', 'src.c'), 'r').read() expected = open(path_from_root('tests', 'hyperbolic', 'output.txt'), 'r').read() self.do_run(src, expected) def test_math_lgamma(self): if self.emcc_args is None: return self.skip('requires emcc') if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for accurate math') test_path = path_from_root('tests', 'math', 'lgamma') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_frexp(self): test_path = path_from_root('tests', 'core', 'test_frexp') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_rounding(self): test_path = path_from_root('tests', 'core', 'test_rounding') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_fcvt(self): if self.emcc_args is None: return self.skip('requires emcc') test_path = path_from_root('tests', 'core', 'test_fcvt') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_llrint(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('requires ta2') test_path = path_from_root('tests', 'core', 'test_llrint') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_getgep(self): # Generated code includes getelementptr (getelementptr, 0, 1), i.e., GEP as the first param to GEP test_path = path_from_root('tests', 'core', 'test_getgep') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_multiply_defined_symbols(self): a1 = "int f() { return 1; }" a1_name = os.path.join(self.get_dir(), 'a1.c') open(a1_name, 'w').write(a1) a2 = "void x() {}" a2_name = os.path.join(self.get_dir(), 'a2.c') open(a2_name, 'w').write(a2) b1 = "int f() { return 2; }" b1_name = os.path.join(self.get_dir(), 'b1.c') open(b1_name, 'w').write(b1) b2 = "void y() {}" b2_name = os.path.join(self.get_dir(), 'b2.c') open(b2_name, 'w').write(b2) main = r''' #include int f(); int main() { printf("result: %d\n", f()); return 0; } ''' main_name = os.path.join(self.get_dir(), 'main.c') open(main_name, 'w').write(main) Building.emcc(a1_name) Building.emcc(a2_name) Building.emcc(b1_name) Building.emcc(b2_name) Building.emcc(main_name) liba_name = os.path.join(self.get_dir(), 'liba.a') Building.emar('cr', liba_name, [a1_name + '.o', a2_name + '.o']) libb_name = os.path.join(self.get_dir(), 'libb.a') Building.emar('cr', libb_name, [b1_name + '.o', b2_name + '.o']) all_name = os.path.join(self.get_dir(), 'all.bc') Building.link([main_name + '.o', liba_name, libb_name], all_name) self.do_ll_run(all_name, 'result: 1') def test_if(self): test_path = path_from_root('tests', 'core', 'test_if') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_if_else(self): test_path = path_from_root('tests', 'core', 'test_if_else') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_loop(self): test_path = path_from_root('tests', 'core', 'test_loop') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_stack(self): Settings.INLINING_LIMIT = 50 test_path = path_from_root('tests', 'core', 'test_stack') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strings(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('musl libc needs ta2') test_path = path_from_root('tests', 'core', 'test_strings') src, output = (test_path + s for s in ('.in', '.out')) for named in (0, 1): print named if os.environ.get('EMCC_FAST_COMPILER') != '0' and named: continue # no named globals in fastcomp Settings.NAMED_GLOBALS = named self.do_run_from_file(src, output, ['wowie', 'too', '74']) if self.emcc_args == []: gen = open(self.in_dir('src.cpp.o.js')).read() assert ('var __str1;' in gen) == named def test_strcmp_uni(self): test_path = path_from_root('tests', 'core', 'test_strcmp_uni') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strndup(self): if self.run_name.startswith('s_'): return self.skip('musl libc strndup() assumes that C strings can be loaded via i16 and i32 loads.') test_path = path_from_root('tests', 'core', 'test_strndup') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_errar(self): test_path = path_from_root('tests', 'core', 'test_errar') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_mainenv(self): test_path = path_from_root('tests', 'core', 'test_mainenv') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_funcs(self): test_path = path_from_root('tests', 'core', 'test_funcs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_structs(self): test_path = path_from_root('tests', 'core', 'test_structs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) gen_struct_src = ''' #include #include #include "emscripten.h" struct S { int x, y; }; int main() { S* a = {{gen_struct}}; a->x = 51; a->y = 62; printf("*%d,%d*\\n", a->x, a->y); {{del_struct}}(a); return 0; } ''' def test_mallocstruct(self): self.do_run(self.gen_struct_src.replace('{{gen_struct}}', '(S*)malloc(sizeof(S))').replace('{{del_struct}}', 'free'), '*51,62*') def test_newstruct(self): if self.emcc_args is None: return self.skip('requires emcc') self.do_run(self.gen_struct_src.replace('{{gen_struct}}', 'new S').replace('{{del_struct}}', 'delete'), '*51,62*') def test_addr_of_stacked(self): test_path = path_from_root('tests', 'core', 'test_addr_of_stacked') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_globals(self): test_path = path_from_root('tests', 'core', 'test_globals') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_linked_list(self): test_path = path_from_root('tests', 'core', 'test_linked_list') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sup(self): src = ''' #include struct S4 { int x; }; // size: 4 struct S4_2 { short x, y; }; // size: 4, but for alignment purposes, 2 struct S6 { short x, y, z; }; // size: 6 struct S6w { char x[6]; }; // size: 6 also struct S6z { int x; short y; }; // size: 8, since we align to a multiple of the biggest - 4 struct C___ { S6 a, b, c; int later; }; struct Carr { S6 a[3]; int later; }; // essentially the same, but differently defined struct C__w { S6 a; S6w b; S6 c; int later; }; // same size, different struct struct Cp1_ { int pre; short a; S6 b, c; int later; }; // fillers for a struct Cp2_ { int a; short pre; S6 b, c; int later; }; // fillers for a (get addr of the other filler) struct Cint { S6 a; int b; S6 c; int later; }; // An int (different size) for b struct C4__ { S6 a; S4 b; S6 c; int later; }; // Same size as int from before, but a struct struct C4_2 { S6 a; S4_2 b; S6 c; int later; }; // Same size as int from before, but a struct with max element size 2 struct C__z { S6 a; S6z b; S6 c; int later; }; // different size, 8 instead of 6 int main() { #define TEST(struc) \\ { \\ struc *s = 0; \\ printf("*%s: %d,%d,%d,%d<%d*\\n", #struc, (int)&(s->a), (int)&(s->b), (int)&(s->c), (int)&(s->later), sizeof(struc)); \\ } #define TEST_ARR(struc) \\ { \\ struc *s = 0; \\ printf("*%s: %d,%d,%d,%d<%d*\\n", #struc, (int)&(s->a[0]), (int)&(s->a[1]), (int)&(s->a[2]), (int)&(s->later), sizeof(struc)); \\ } printf("sizeofs:%d,%d\\n", sizeof(S6), sizeof(S6z)); TEST(C___); TEST_ARR(Carr); TEST(C__w); TEST(Cp1_); TEST(Cp2_); TEST(Cint); TEST(C4__); TEST(C4_2); TEST(C__z); return 0; } ''' if Settings.QUANTUM_SIZE == 1: self.do_run(src, 'sizeofs:6,8\n*C___: 0,3,6,9<24*\n*Carr: 0,3,6,9<24*\n*C__w: 0,3,9,12<24*\n*Cp1_: 1,2,5,8<24*\n*Cp2_: 0,2,5,8<24*\n*Cint: 0,3,4,7<24*\n*C4__: 0,3,4,7<24*\n*C4_2: 0,3,5,8<20*\n*C__z: 0,3,5,8<28*') else: self.do_run(src, 'sizeofs:6,8\n*C___: 0,6,12,20<24*\n*Carr: 0,6,12,20<24*\n*C__w: 0,6,12,20<24*\n*Cp1_: 4,6,12,20<24*\n*Cp2_: 0,6,12,20<24*\n*Cint: 0,8,12,20<24*\n*C4__: 0,8,12,20<24*\n*C4_2: 0,6,10,16<20*\n*C__z: 0,8,16,24<28*') def test_assert(self): test_path = path_from_root('tests', 'core', 'test_assert') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_libcextra(self): if self.emcc_args is None: return self.skip('needs emcc for libcextra') test_path = path_from_root('tests', 'core', 'test_libcextra') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_regex(self): if self.emcc_args is None: return self.skip('needs emcc for libcextra') test_path = path_from_root('tests', 'core', 'test_regex') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_longjmp(self): test_path = path_from_root('tests', 'core', 'test_longjmp') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_longjmp2(self): test_path = path_from_root('tests', 'core', 'test_longjmp2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_longjmp3(self): test_path = path_from_root('tests', 'core', 'test_longjmp3') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_longjmp4(self): test_path = path_from_root('tests', 'core', 'test_longjmp4') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_longjmp_funcptr(self): test_path = path_from_root('tests', 'core', 'test_longjmp_funcptr') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_longjmp_repeat(self): test_path = path_from_root('tests', 'core', 'test_longjmp_repeat') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_longjmp_stacked(self): test_path = path_from_root('tests', 'core', 'test_longjmp_stacked') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_longjmp_exc(self): test_path = path_from_root('tests', 'core', 'test_longjmp_exc') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_longjmp_throw(self): for disable_throw in [0, 1]: print disable_throw Settings.DISABLE_EXCEPTION_CATCHING = disable_throw test_path = path_from_root('tests', 'core', 'test_longjmp_throw') src, output = (test_path + s for s in ('.cpp', '.out')) self.do_run_from_file(src, output) def test_setjmp_many(self): src = r''' #include #include int main(int argc) { jmp_buf buf; for (int i = 0; i < NUM; i++) printf("%d\n", setjmp(buf)); if (argc-- == 1131) longjmp(buf, 11); return 0; } ''' for maxx in [Settings.MAX_SETJMPS/2, Settings.MAX_SETJMPS, 2*Settings.MAX_SETJMPS]: Settings.MAX_SETJMPS = maxx for num in [maxx, maxx+1]: print maxx, num self.do_run(src.replace('NUM', str(num)), '0\n' * num if num <= Settings.MAX_SETJMPS or not Settings.ASM_JS else 'build with a higher value for MAX_SETJMPS') def test_setjmp_many_2(self): if os.environ.get('EMCC_FAST_COMPILER') == '0': return self.skip('non-fastcomp do not hit the limit.') src = r''' #include #include jmp_buf env; void luaWork(int d){ int x; printf("d is at %d\n", d); longjmp(env, 1); } int main() { const int ITERATIONS=25; for(int i = 0; i < ITERATIONS; i++){ if(!setjmp(env)){ luaWork(i); } } return 0; } ''' self.do_run(src, r'''d is at 19 too many setjmps in a function call, build with a higher value for MAX_SETJMPS''') def test_exceptions(self): if Settings.QUANTUM_SIZE == 1: return self.skip("we don't support libcxx in q1") if self.emcc_args is None: return self.skip('need emcc to add in libcxx properly') Settings.EXCEPTION_DEBUG = 1 Settings.DISABLE_EXCEPTION_CATCHING = 0 if '-O2' in self.emcc_args: self.emcc_args += ['--closure', '1'] # Use closure here for some additional coverage src = ''' #include void thrower() { printf("infunc..."); throw(99); printf("FAIL"); } int main() { try { printf("*throw..."); throw(1); printf("FAIL"); } catch(...) { printf("caught!"); } try { thrower(); } catch(...) { printf("done!*\\n"); } return 0; } ''' self.do_run(src, '*throw...caught!infunc...done!*') Settings.DISABLE_EXCEPTION_CATCHING = 1 self.do_run(src, 'Exception catching is disabled, this exception cannot be caught. Compile with -s DISABLE_EXCEPTION_CATCHING=0') src = ''' #include class MyException { public: MyException(){ std::cout << "Construct..."; } MyException( const MyException & ) { std::cout << "Copy..."; } ~MyException(){ std::cout << "Destruct..."; } }; int function() { std::cout << "Throw..."; throw MyException(); } int function2() { return function(); } int main() { try { function2(); } catch (MyException & e) { std::cout << "Caught..."; } try { function2(); } catch (MyException e) { std::cout << "Caught..."; } return 0; } ''' Settings.DISABLE_EXCEPTION_CATCHING = 0 if '-O2' in self.emcc_args: self.emcc_args.pop() ; self.emcc_args.pop() # disable closure to work around a closure bug self.do_run(src, 'Throw...Construct...Caught...Destruct...Throw...Construct...Copy...Caught...Destruct...Destruct...') def test_exceptions_2(self): if self.emcc_args is None: return self.skip('need emcc to add in libcxx properly') if self.run_name == 'asm2x86': return self.skip('TODO') Settings.DISABLE_EXCEPTION_CATCHING = 0 for safe in [0,1]: print safe Settings.SAFE_HEAP = safe test_path = path_from_root('tests', 'core', 'test_exceptions_2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_exceptions_white_list(self): Settings.DISABLE_EXCEPTION_CATCHING = 2 Settings.EXCEPTION_CATCHING_WHITELIST = ["__Z12somefunctionv"] Settings.INLINING_LIMIT = 50 # otherwise it is inlined and not identified test_path = path_from_root('tests', 'core', 'test_exceptions_white_list') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_exceptions_white_list_2(self): Settings.DISABLE_EXCEPTION_CATCHING = 2 Settings.EXCEPTION_CATCHING_WHITELIST = ["_main"] Settings.INLINING_LIMIT = 50 # otherwise it is inlined and not identified test_path = path_from_root('tests', 'core', 'test_exceptions_white_list_2') src, output = (test_path + s for s in ('.c', '.out')) self.do_run_from_file(src, output) def test_exceptions_uncaught(self): if self.emcc_args is None: return self.skip('no libcxx inclusion without emcc') Settings.DISABLE_EXCEPTION_CATCHING = 0 src = r''' #include #include struct X { ~X() { printf("exception? %s\n", std::uncaught_exception() ? "yes" : "no"); } }; int main() { printf("exception? %s\n", std::uncaught_exception() ? "yes" : "no"); try { X x; throw 1; } catch(...) { printf("exception? %s\n", std::uncaught_exception() ? "yes" : "no"); } printf("exception? %s\n", std::uncaught_exception() ? "yes" : "no"); return 0; } ''' self.do_run(src, 'exception? no\nexception? yes\nexception? no\nexception? no\n') src = r''' #include #include int main() { std::ofstream os("test"); os << std::unitbuf << "foo"; // trigger a call to std::uncaught_exception from // std::basic_ostream::sentry::~sentry std::cout << "success"; } ''' self.do_run(src, 'success') def test_exceptions_typed(self): Settings.DISABLE_EXCEPTION_CATCHING = 0 Settings.SAFE_HEAP = 0 # Throwing null will cause an ignorable null pointer access. test_path = path_from_root('tests', 'core', 'test_exceptions_typed') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_exceptions_multi(self): Settings.DISABLE_EXCEPTION_CATCHING = 0 test_path = path_from_root('tests', 'core', 'test_exceptions_multi') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_exceptions_std(self): if self.emcc_args is None: return self.skip('requires emcc') Settings.DISABLE_EXCEPTION_CATCHING = 0 self.emcc_args += ['-s', 'SAFE_HEAP=0'] test_path = path_from_root('tests', 'core', 'test_exceptions_std') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_exceptions_alias(self): Settings.DISABLE_EXCEPTION_CATCHING = 0 test_path = path_from_root('tests', 'core', 'test_exceptions_alias') src, output = (test_path + s for s in ('.c', '.out')) self.do_run_from_file(src, output) def test_async_exit(self): open('main.c', 'w').write(r''' #include #include #include "emscripten.h" void main_loop() { exit(EXIT_SUCCESS); } int main() { emscripten_set_main_loop(main_loop, 60, 0); return 0; } ''') Popen([PYTHON, EMCC, 'main.c']).communicate() self.assertNotContained('Reached an unreachable!', run_js(self.in_dir('a.out.js'), stderr=STDOUT)) def test_exit_stack(self): if self.emcc_args is None: return self.skip('requires emcc') if Settings.ASM_JS: return self.skip('uses report_stack without exporting') Settings.INLINING_LIMIT = 50 src = r''' #include #include extern "C" { extern void report_stack(int x); } char moar() { char temp[125]; for (int i = 0; i < 125; i++) temp[i] = i*i; for (int i = 1; i < 125; i++) temp[i] += temp[i-1]/2; if (temp[100] != 99) exit(1); return temp[120]; } int main(int argc, char *argv[]) { report_stack((int)alloca(4)); printf("*%d*\n", moar()); return 0; } ''' open(os.path.join(self.get_dir(), 'pre.js'), 'w').write(''' var initialStack = -1; var _report_stack = function(x) { Module.print('reported'); initialStack = x; } var Module = { postRun: function() { Module.print('Exit Status: ' + EXITSTATUS); Module.print('postRun'); assert(initialStack == STACKTOP, [initialStack, STACKTOP]); Module.print('ok.'); } }; ''') self.emcc_args += ['--pre-js', 'pre.js'] self.do_run(src, '''reported\nExit Status: 1\npostRun\nok.\n''') def test_class(self): test_path = path_from_root('tests', 'core', 'test_class') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_inherit(self): test_path = path_from_root('tests', 'core', 'test_inherit') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_isdigit_l(self): if self.emcc_args is None: return self.skip('no libcxx inclusion without emcc') test_path = path_from_root('tests', 'core', 'test_isdigit_l') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_iswdigit(self): if self.emcc_args is None: return self.skip('no libcxx inclusion without emcc') test_path = path_from_root('tests', 'core', 'test_iswdigit') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_polymorph(self): if self.emcc_args is None: return self.skip('requires emcc') test_path = path_from_root('tests', 'core', 'test_polymorph') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_segfault(self): if self.emcc_args is None: return self.skip('SAFE_HEAP without ta2 means we check types too, which hide segfaults') Settings.SAFE_HEAP = 1 for addr in ['0', 'new D2()']: print addr src = r''' #include struct Classey { virtual void doIt() = 0; }; struct D1 : Classey { virtual void doIt() { printf("fleefl\n"); } }; struct D2 : Classey { virtual void doIt() { printf("marfoosh\n"); } }; int main(int argc, char **argv) { Classey *p = argc == 100 ? new D1() : (Classey*)%s; p->doIt(); return 0; } ''' % addr self.do_run(src, 'segmentation fault' if addr.isdigit() else 'marfoosh') def test_safe_dyncalls(self): if Settings.ASM_JS: return self.skip('asm does not support missing function stack traces') if Settings.SAFE_HEAP: return self.skip('safe heap warning will appear instead') if self.emcc_args is None: return self.skip('need libc') Settings.SAFE_DYNCALLS = 1 for cond, body, work in [(True, True, False), (True, False, False), (False, True, True), (False, False, False)]: print cond, body, work src = r''' #include struct Classey { virtual void doIt() = 0; }; struct D1 : Classey { virtual void doIt() BODY; }; int main(int argc, char **argv) { Classey *p = argc COND 100 ? new D1() : NULL; printf("%p\n", p); p->doIt(); return 0; } '''.replace('COND', '==' if cond else '!=').replace('BODY', r'{ printf("all good\n"); }' if body else '') # typically we get the dyncall error message from SAFE_DYNCALLS, however llvm opts can devirtualize the # call in -O2, leading to the much nicer error message specifically about the missing function self.do_run(src, ('dyncall error: vi', 'missing function: _ZN2D14doItEv') if not work else 'all good') def test_dynamic_cast(self): if self.emcc_args is None: return self.skip('need libcxxabi') test_path = path_from_root('tests', 'core', 'test_dynamic_cast') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_dynamic_cast_b(self): if self.emcc_args is None: return self.skip('need libcxxabi') test_path = path_from_root('tests', 'core', 'test_dynamic_cast_b') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_dynamic_cast_2(self): if self.emcc_args is None: return self.skip('need libcxxabi') test_path = path_from_root('tests', 'core', 'test_dynamic_cast_2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_funcptr(self): test_path = path_from_root('tests', 'core', 'test_funcptr') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_mathfuncptr(self): test_path = path_from_root('tests', 'core', 'test_mathfuncptr') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_funcptrfunc(self): test_path = path_from_root('tests', 'core', 'test_funcptrfunc') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_funcptr_namecollide(self): test_path = path_from_root('tests', 'core', 'test_funcptr_namecollide') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, force_c=True) def test_emptyclass(self): if self.emcc_args is None: return self.skip('requires emcc') test_path = path_from_root('tests', 'core', 'test_emptyclass') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_alloca(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('non-ta2 may have unaligned allocas') test_path = path_from_root('tests', 'core', 'test_alloca') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, force_c=True) def test_rename(self): src = open(path_from_root('tests', 'stdio', 'test_rename.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_alloca_stack(self): if self.emcc_args is None: return # too slow in other modes test_path = path_from_root('tests', 'core', 'test_alloca_stack') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, force_c=True) def test_stack_byval(self): if self.emcc_args is None: return # too slow in other modes test_path = path_from_root('tests', 'core', 'test_stack_byval') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_stack_varargs(self): if self.emcc_args is None: return # too slow in other modes Settings.INLINING_LIMIT = 50 Settings.TOTAL_STACK = 1024 test_path = path_from_root('tests', 'core', 'test_stack_varargs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_stack_varargs2(self): if self.emcc_args is None: return # too slow in other modes Settings.TOTAL_STACK = 1024 src = r''' #include #include void func(int i) { } int main() { for (int i = 0; i < 1024; i++) { printf("%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d\n", i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i); } printf("ok!\n"); return 0; } ''' self.do_run(src, 'ok!') print 'with return' src = r''' #include #include int main() { for (int i = 0; i < 1024; i++) { int j = printf("%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d", i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i); printf(" (%d)\n", j); } printf("ok!\n"); return 0; } ''' self.do_run(src, 'ok!') print 'with definitely no return' src = r''' #include #include #include void vary(const char *s, ...) { va_list v; va_start(v, s); char d[20]; vsnprintf(d, 20, s, v); puts(d); // Try it with copying va_list tempva; va_copy(tempva, v); vsnprintf(d, 20, s, tempva); puts(d); va_end(v); } int main() { for (int i = 0; i < 1024; i++) { int j = printf("%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d,%d", i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i, i); printf(" (%d)\n", j); vary("*cheez: %d+%d*", 99, 24); vary("*albeit*"); } printf("ok!\n"); return 0; } ''' self.do_run(src, 'ok!') def test_stack_void(self): Settings.INLINING_LIMIT = 50 test_path = path_from_root('tests', 'core', 'test_stack_void') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_life(self): if self.emcc_args is None: return self.skip('need c99') self.emcc_args += ['-std=c99'] src = open(path_from_root('tests', 'life.c'), 'r').read() self.do_run(src, '''-------------------------------- [] [] [][][] [] [] [] [][] [] [] [] [] [][] [][] [][][] [] [] [] [] [] [][] [] [] [] [][] [] [] [] [] [][][][] [][] [][] [] [][][] [] [] [] [][] [][] [][] [][][] [][] [][][] [] [] [][] [][] [] [][][] [] [][][] [] [][] [][] [][] [] [][] [][] [][] [][] [] [][] [][] [] [] [][] [] [][][] [] [] [][] [] [] [] [] [] [] [] [][][] [] [][][] [] -------------------------------- ''', ['2'], force_c=True) def test_array2(self): test_path = path_from_root('tests', 'core', 'test_array2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_array2b(self): test_path = path_from_root('tests', 'core', 'test_array2b') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_constglobalstructs(self): test_path = path_from_root('tests', 'core', 'test_constglobalstructs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_conststructs(self): test_path = path_from_root('tests', 'core', 'test_conststructs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_bigarray(self): if self.emcc_args is None: return self.skip('need ta2 to compress type data on zeroinitializers') test_path = path_from_root('tests', 'core', 'test_bigarray') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_mod_globalstruct(self): test_path = path_from_root('tests', 'core', 'test_mod_globalstruct') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_pystruct(self): src = ''' #include // Based on CPython code union PyGC_Head { struct { union PyGC_Head *gc_next; union PyGC_Head *gc_prev; size_t gc_refs; } gc; long double dummy; /* force worst-case alignment */ } ; struct gc_generation { PyGC_Head head; int threshold; /* collection threshold */ int count; /* count of allocations or collections of younger generations */ }; #define NUM_GENERATIONS 3 #define GEN_HEAD(n) (&generations[n].head) /* linked lists of container objects */ static struct gc_generation generations[NUM_GENERATIONS] = { /* PyGC_Head, threshold, count */ {{{GEN_HEAD(0), GEN_HEAD(0), 0}}, 700, 0}, {{{GEN_HEAD(1), GEN_HEAD(1), 0}}, 10, 0}, {{{GEN_HEAD(2), GEN_HEAD(2), 0}}, 10, 0}, }; int main() { gc_generation *n = NULL; printf("*%d,%d,%d,%d,%d,%d,%d,%d*\\n", (int)(&n[0]), (int)(&n[0].head), (int)(&n[0].head.gc.gc_next), (int)(&n[0].head.gc.gc_prev), (int)(&n[0].head.gc.gc_refs), (int)(&n[0].threshold), (int)(&n[0].count), (int)(&n[1]) ); printf("*%d,%d,%d*\\n", (int)(&generations[0]) == (int)(&generations[0].head.gc.gc_next), (int)(&generations[0]) == (int)(&generations[0].head.gc.gc_prev), (int)(&generations[0]) == (int)(&generations[1]) ); int x1 = (int)(&generations[0]); int x2 = (int)(&generations[1]); printf("*%d*\\n", x1 == x2); for (int i = 0; i < NUM_GENERATIONS; i++) { PyGC_Head *list = GEN_HEAD(i); printf("%d:%d,%d\\n", i, (int)list == (int)(list->gc.gc_prev), (int)list ==(int)(list->gc.gc_next)); } printf("*%d,%d,%d*\\n", sizeof(PyGC_Head), sizeof(gc_generation), int(GEN_HEAD(2)) - int(GEN_HEAD(1))); } ''' if Settings.QUANTUM_SIZE == 1: # Compressed memory. Note that sizeof() does give the fat sizes, however! self.do_run(src, '*0,0,0,1,2,3,4,5*\n*1,0,0*\n*0*\n0:1,1\n1:1,1\n2:1,1\n*12,20,5*') else: if self.is_emscripten_abi(): self.do_run(src, '*0,0,0,4,8,16,20,24*\n*1,0,0*\n*0*\n0:1,1\n1:1,1\n2:1,1\n*16,24,24*') else: self.do_run(src, '*0,0,0,4,8,12,16,20*\n*1,0,0*\n*0*\n0:1,1\n1:1,1\n2:1,1\n*12,20,20*') def test_ptrtoint(self): if self.emcc_args is None: return self.skip('requires emcc') runner = self def check_warnings(output): runner.assertEquals(filter(lambda line: 'Warning' in line, output.split('\n')).__len__(), 4) test_path = path_from_root('tests', 'core', 'test_ptrtoint') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, output_processor=check_warnings) def test_sizeof(self): if self.emcc_args is None: return self.skip('requires emcc') # Has invalid writes between printouts Settings.SAFE_HEAP = 0 test_path = path_from_root('tests', 'core', 'test_sizeof') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, [], lambda x, err: x.replace('\n', '*')) def test_llvm_used(self): Building.LLVM_OPTS = 3 test_path = path_from_root('tests', 'core', 'test_llvm_used') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_emscripten_api(self): #if Settings.MICRO_OPTS or Settings.RELOOP or Building.LLVM_OPTS: return self.skip('FIXME') test_path = path_from_root('tests', 'core', 'test_emscripten_api') src, output = (test_path + s for s in ('.in', '.out')) check = ''' def process(filename): src = open(filename, 'r').read() # TODO: restore this (see comment in emscripten.h) assert '// hello from the source' in src ''' Settings.EXPORTED_FUNCTIONS = ['_main', '_save_me_aimee'] self.do_run_from_file(src, output, post_build=check) # test EXPORT_ALL Settings.EXPORTED_FUNCTIONS = [] Settings.EXPORT_ALL = 1 self.do_run_from_file(src, output, post_build=check) def test_emscripten_get_now(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('requires ta2') if self.run_name == 'slow2asm': self.emcc_args += ['--closure', '1'] # Use closure here for some additional coverage self.do_run(open(path_from_root('tests', 'emscripten_get_now.cpp')).read(), 'Timer resolution is good.') def test_emscripten_get_compiler_setting(self): test_path = path_from_root('tests', 'core', 'emscripten_get_compiler_setting') src, output = (test_path + s for s in ('.c', '.out')) self.do_run(open(src).read(), 'You must build with -s RETAIN_COMPILER_SETTINGS=1') Settings.RETAIN_COMPILER_SETTINGS = 1 self.do_run(open(src).read(), open(output).read().replace('waka', EMSCRIPTEN_VERSION)) def test_inlinejs(self): if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') if os.environ.get('EMCC_FAST_COMPILER') != '0': return self.skip('fastcomp only supports EM_ASM') test_path = path_from_root('tests', 'core', 'test_inlinejs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) if self.emcc_args == []: # opts will eliminate the comments out = open('src.cpp.o.js').read() for i in range(1, 5): assert ('comment%d' % i) in out def test_inlinejs2(self): if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') if os.environ.get('EMCC_FAST_COMPILER') != '0': return self.skip('fastcomp only supports EM_ASM') test_path = path_from_root('tests', 'core', 'test_inlinejs2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_inlinejs3(self): test_path = path_from_root('tests', 'core', 'test_inlinejs3') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_memorygrowth(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('memory growth is only supported with typed arrays mode 2') # With typed arrays in particular, it is dangerous to use more memory than TOTAL_MEMORY, # since we then need to enlarge the heap(s). src = r''' #include #include #include #include #include "emscripten.h" int main(int argc, char **argv) { char *buf1 = (char*)malloc(100); char *data1 = "hello"; memcpy(buf1, data1, strlen(data1)+1); float *buf2 = (float*)malloc(100); float pie = 4.955; memcpy(buf2, &pie, sizeof(float)); printf("*pre: %s,%.3f*\n", buf1, buf2[0]); int totalMemory = emscripten_run_script_int("TOTAL_MEMORY"); char *buf3 = (char*)malloc(totalMemory+1); buf3[argc] = (int)buf2; if (argc % 7 == 6) printf("%d\n", memcpy(buf3, buf1, argc)); char *buf4 = (char*)malloc(100); float *buf5 = (float*)malloc(100); //printf("totalMemory: %d bufs: %d,%d,%d,%d,%d\n", totalMemory, buf1, buf2, buf3, buf4, buf5); assert((int)buf4 > (int)totalMemory && (int)buf5 > (int)totalMemory); printf("*%s,%.3f*\n", buf1, buf2[0]); // the old heap data should still be there memcpy(buf4, buf1, strlen(data1)+1); memcpy(buf5, buf2, sizeof(float)); printf("*%s,%.3f*\n", buf4, buf5[0]); // and the new heap space should work too return 0; } ''' # Fail without memory growth self.do_run(src, 'Cannot enlarge memory arrays.') fail = open('src.cpp.o.js').read() # Win with it Settings.ALLOW_MEMORY_GROWTH = 1 self.do_run(src, '*pre: hello,4.955*\n*hello,4.955*\n*hello,4.955*') win = open('src.cpp.o.js').read() if self.emcc_args and '-O2' in self.emcc_args: # Make sure ALLOW_MEMORY_GROWTH generates different code (should be less optimized) code_start = 'var TOTAL_MEMORY = ' fail = fail[fail.find(code_start):] win = win[win.find(code_start):] assert len(fail) < len(win), 'failing code - without memory growth on - is more optimized, and smaller' def test_ssr(self): # struct self-ref src = ''' #include // see related things in openjpeg typedef struct opj_mqc_state { unsigned int qeval; int mps; struct opj_mqc_state *nmps; struct opj_mqc_state *nlps; } opj_mqc_state_t; static opj_mqc_state_t mqc_states[2] = { {0x5600, 0, &mqc_states[2], &mqc_states[3]}, {0x5602, 1, &mqc_states[3], &mqc_states[2]}, }; int main() { printf("*%d*\\n", (int)(mqc_states+1)-(int)mqc_states); for (int i = 0; i < 2; i++) printf("%d:%d,%d,%d,%d\\n", i, mqc_states[i].qeval, mqc_states[i].mps, (int)mqc_states[i].nmps-(int)mqc_states, (int)mqc_states[i].nlps-(int)mqc_states); return 0; } ''' if Settings.QUANTUM_SIZE == 1: self.do_run(src, '''*4*\n0:22016,0,8,12\n1:22018,1,12,8\n''') else: self.do_run(src, '''*16*\n0:22016,0,32,48\n1:22018,1,48,32\n''') def test_tinyfuncstr(self): if self.emcc_args is None: return self.skip('requires emcc') test_path = path_from_root('tests', 'core', 'test_tinyfuncstr') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_llvmswitch(self): Settings.CORRECT_SIGNS = 1 test_path = path_from_root('tests', 'core', 'test_llvmswitch') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) # By default, when user has not specified a -std flag, Emscripten should always build .cpp files using the C++03 standard, # i.e. as if "-std=c++03" had been passed on the command line. On Linux with Clang 3.2 this is the case, but on Windows # with Clang 3.2 -std=c++11 has been chosen as default, because of # < jrose> clb: it's deliberate, with the idea that for people who don't care about the standard, they should be using the "best" thing we can offer on that platform def test_cxx03_do_run(self): test_path = path_from_root('tests', 'core', 'test_cxx03_do_run') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_bigswitch(self): if self.run_name != 'default': return self.skip('TODO: issue #781') src = open(path_from_root('tests', 'bigswitch.cpp')).read() self.do_run(src, '''34962: GL_ARRAY_BUFFER (0x8892) 26214: what? 35040: GL_STREAM_DRAW (0x88E0) ''', args=['34962', '26214', '35040']) def test_indirectbr(self): Building.COMPILER_TEST_OPTS = filter(lambda x: x != '-g', Building.COMPILER_TEST_OPTS) test_path = path_from_root('tests', 'core', 'test_indirectbr') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_indirectbr_many(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('blockaddr > 255 requires ta2') test_path = path_from_root('tests', 'core', 'test_indirectbr_many') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_pack(self): src = ''' #include #include #pragma pack(push,1) typedef struct header { unsigned char id; unsigned short colour; unsigned char desc; } header; #pragma pack(pop) typedef struct fatheader { unsigned char id; unsigned short colour; unsigned char desc; } fatheader; int main( int argc, const char *argv[] ) { header h, *ph = 0; fatheader fh, *pfh = 0; printf("*%d,%d,%d*\\n", sizeof(header), (int)((int)&h.desc - (int)&h.id), (int)(&ph[1])-(int)(&ph[0])); printf("*%d,%d,%d*\\n", sizeof(fatheader), (int)((int)&fh.desc - (int)&fh.id), (int)(&pfh[1])-(int)(&pfh[0])); return 0; } ''' if Settings.QUANTUM_SIZE == 1: self.do_run(src, '*4,2,3*\n*6,2,3*') else: self.do_run(src, '*4,3,4*\n*6,4,6*') def test_varargs(self): if Settings.QUANTUM_SIZE == 1: return self.skip('FIXME: Add support for this') if not self.is_emscripten_abi(): return self.skip('we do not support all varargs stuff without asmjs-unknown-emscripten') test_path = path_from_root('tests', 'core', 'test_varargs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_varargs_byval(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('FIXME: Add support for this') if self.is_emscripten_abi(): return self.skip('clang cannot compile this code with that target yet') src = r''' #include #include typedef struct type_a { union { double f; void *p; int i; short sym; } value; } type_a; enum mrb_vtype { MRB_TT_FALSE = 0, /* 0 */ MRB_TT_CLASS = 9 /* 9 */ }; typedef struct type_b { enum mrb_vtype tt:8; } type_b; void print_type_a(int argc, ...); void print_type_b(int argc, ...); int main(int argc, char *argv[]) { type_a a; type_b b; a.value.p = (void*) 0x12345678; b.tt = MRB_TT_CLASS; printf("The original address of a is: %p\n", a.value.p); printf("The original type of b is: %d\n", b.tt); print_type_a(1, a); print_type_b(1, b); return 0; } void print_type_a(int argc, ...) { va_list ap; type_a a; va_start(ap, argc); a = va_arg(ap, type_a); va_end(ap); printf("The current address of a is: %p\n", a.value.p); } void print_type_b(int argc, ...) { va_list ap; type_b b; va_start(ap, argc); b = va_arg(ap, type_b); va_end(ap); printf("The current type of b is: %d\n", b.tt); } ''' self.do_run(src, '''The original address of a is: 0x12345678 The original type of b is: 9 The current address of a is: 0x12345678 The current type of b is: 9 ''') def test_functionpointer_libfunc_varargs(self): test_path = path_from_root('tests', 'core', 'test_functionpointer_libfunc_varargs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_structbyval(self): Settings.INLINING_LIMIT = 50 # part 1: make sure that normally, passing structs by value works src = r''' #include struct point { int x, y; }; void dump(struct point p) { p.x++; // should not modify p.y++; // anything in the caller! printf("dump: %d,%d\n", p.x, p.y); } void dumpmod(struct point *p) { p->x++; // should not modify p->y++; // anything in the caller! printf("dump: %d,%d\n", p->x, p->y); } int main( int argc, const char *argv[] ) { point p = { 54, 2 }; printf("pre: %d,%d\n", p.x, p.y); dump(p); void (*dp)(point p) = dump; // And, as a function pointer dp(p); printf("post: %d,%d\n", p.x, p.y); dumpmod(&p); dumpmod(&p); printf("last: %d,%d\n", p.x, p.y); return 0; } ''' self.do_run(src, 'pre: 54,2\ndump: 55,3\ndump: 55,3\npost: 54,2\ndump: 55,3\ndump: 56,4\nlast: 56,4') # Check for lack of warning in the generated code (they should appear in part 2) generated = open(os.path.join(self.get_dir(), 'src.cpp.o.js')).read() assert 'Casting a function pointer type to another with a different number of arguments.' not in generated, 'Unexpected warning' # part 2: make sure we warn about mixing c and c++ calling conventions here if not (self.emcc_args is None or self.emcc_args == []): return # Optimized code is missing the warning comments header = r''' struct point { int x, y; }; ''' open(os.path.join(self.get_dir(), 'header.h'), 'w').write(header) supp = r''' #include #include "header.h" void dump(struct point p) { p.x++; // should not modify p.y++; // anything in the caller! printf("dump: %d,%d\n", p.x, p.y); } ''' supp_name = os.path.join(self.get_dir(), 'supp.c') open(supp_name, 'w').write(supp) main = r''' #include #include "header.h" #ifdef __cplusplus extern "C" { #endif void dump(struct point p); #ifdef __cplusplus } #endif int main( int argc, const char *argv[] ) { struct point p = { 54, 2 }; printf("pre: %d,%d\n", p.x, p.y); dump(p); void (*dp)(struct point p) = dump; // And, as a function pointer dp(p); printf("post: %d,%d\n", p.x, p.y); return 0; } ''' main_name = os.path.join(self.get_dir(), 'main.cpp') open(main_name, 'w').write(main) Building.emcc(supp_name) Building.emcc(main_name) all_name = os.path.join(self.get_dir(), 'all.bc') Building.link([supp_name + '.o', main_name + '.o'], all_name) # This will fail! See explanation near the warning we check for, in the compiler source code output = Popen([PYTHON, EMCC, all_name], stderr=PIPE).communicate() # Check for warning in the generated code generated = open(os.path.join(self.get_dir(), 'src.cpp.o.js')).read() if 'i386-pc-linux-gnu' in COMPILER_OPTS: assert 'Casting a function pointer type to a potentially incompatible one' in output[1], 'Missing expected warning' else: print >> sys.stderr, 'skipping C/C++ conventions warning check, since not i386-pc-linux-gnu' def test_stdlibs(self): if self.emcc_args is None: return self.skip('requires emcc') if Settings.USE_TYPED_ARRAYS == 2: # Typed arrays = 2 + safe heap prints a warning that messes up our output. Settings.SAFE_HEAP = 0 src = ''' #include #include #include #include void clean() { printf("*cleaned*\\n"); } int comparer(const void *a, const void *b) { int aa = *((int*)a); int bb = *((int*)b); return aa - bb; } int main() { // timeofday timeval t; gettimeofday(&t, NULL); printf("*%d,%d\\n", int(t.tv_sec), int(t.tv_usec)); // should not crash // atexit atexit(clean); // qsort int values[6] = { 3, 2, 5, 1, 5, 6 }; qsort(values, 5, sizeof(int), comparer); printf("*%d,%d,%d,%d,%d,%d*\\n", values[0], values[1], values[2], values[3], values[4], values[5]); printf("*stdin==0:%d*\\n", stdin == 0); // check that external values are at least not NULL printf("*%%*\\n"); printf("*%.1ld*\\n", 5); printf("*%.1f*\\n", strtod("66", NULL)); // checks dependency system, as our strtod needs _isspace etc. printf("*%ld*\\n", strtol("10", NULL, 0)); printf("*%ld*\\n", strtol("0", NULL, 0)); printf("*%ld*\\n", strtol("-10", NULL, 0)); printf("*%ld*\\n", strtol("12", NULL, 16)); printf("*%lu*\\n", strtoul("10", NULL, 0)); printf("*%lu*\\n", strtoul("0", NULL, 0)); printf("*%lu*\\n", strtoul("-10", NULL, 0)); printf("*malloc(0)!=0:%d*\\n", malloc(0) != 0); // We should not fail horribly printf("tolower_l: %c\\n", tolower_l('A', 0)); return 0; } ''' self.do_run(src, '*1,2,3,5,5,6*\n*stdin==0:0*\n*%*\n*5*\n*66.0*\n*10*\n*0*\n*-10*\n*18*\n*10*\n*0*\n*4294967286*\n*malloc(0)!=0:1*\ntolower_l: a\n*cleaned*') src = r''' #include #include int main() { bool x = true; bool y = false; printf("*%d*\n", x != y); return 0; } ''' self.do_run(src, '*1*', force_c=True) def test_strtoll_hex(self): if self.emcc_args is None: return self.skip('requires emcc') # tests strtoll for hex strings (0x...) test_path = path_from_root('tests', 'core', 'test_strtoll_hex') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strtoll_dec(self): if self.emcc_args is None: return self.skip('requires emcc') # tests strtoll for decimal strings (0x...) test_path = path_from_root('tests', 'core', 'test_strtoll_dec') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strtoll_bin(self): if self.emcc_args is None: return self.skip('requires emcc') # tests strtoll for binary strings (0x...) test_path = path_from_root('tests', 'core', 'test_strtoll_bin') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strtoll_oct(self): if self.emcc_args is None: return self.skip('requires emcc') # tests strtoll for decimal strings (0x...) test_path = path_from_root('tests', 'core', 'test_strtoll_oct') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strtol_hex(self): if self.run_name.startswith('s_'): return self.skip('Needs musl libc.') # tests strtoll for hex strings (0x...) test_path = path_from_root('tests', 'core', 'test_strtol_hex') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strtol_dec(self): if self.run_name.startswith('s_'): return self.skip('Needs musl libc.') # tests strtoll for decimal strings (0x...) test_path = path_from_root('tests', 'core', 'test_strtol_dec') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strtol_bin(self): if self.run_name.startswith('s_'): return self.skip('Needs musl libc.') # tests strtoll for binary strings (0x...) test_path = path_from_root('tests', 'core', 'test_strtol_bin') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strtol_oct(self): if self.run_name.startswith('s_'): return self.skip('Needs musl libc.') # tests strtoll for decimal strings (0x...) test_path = path_from_root('tests', 'core', 'test_strtol_oct') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_atexit(self): # Confirms they are called in reverse order test_path = path_from_root('tests', 'core', 'test_atexit') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_pthread_specific(self): if self.emcc_args is None: return self.skip('requires emcc') src = open(path_from_root('tests', 'pthread', 'specific.c'), 'r').read() expected = open(path_from_root('tests', 'pthread', 'specific.c.txt'), 'r').read() self.do_run(src, expected, force_c=True) def test_tcgetattr(self): src = open(path_from_root('tests', 'termios', 'test_tcgetattr.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_time(self): # XXX Not sure what the right output is here. Looks like the test started failing with daylight savings changes. Modified it to pass again. src = open(path_from_root('tests', 'time', 'src.c'), 'r').read() expected = open(path_from_root('tests', 'time', 'output.txt'), 'r').read() expected2 = open(path_from_root('tests', 'time', 'output2.txt'), 'r').read() self.do_run(src, [expected, expected2], extra_emscripten_args=['-H', 'libc/time.h']) #extra_emscripten_args=['-H', 'libc/fcntl.h,libc/sys/unistd.h,poll.h,libc/math.h,libc/langinfo.h,libc/time.h']) def test_timeb(self): # Confirms they are called in reverse order test_path = path_from_root('tests', 'core', 'test_timeb') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_time_c(self): test_path = path_from_root('tests', 'core', 'test_time_c') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_gmtime(self): test_path = path_from_root('tests', 'core', 'test_gmtime') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strptime_tm(self): test_path = path_from_root('tests', 'core', 'test_strptime_tm') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strptime_days(self): test_path = path_from_root('tests', 'core', 'test_strptime_days') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strptime_reentrant(self): test_path = path_from_root('tests', 'core', 'test_strptime_reentrant') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strftime(self): test_path = path_from_root('tests', 'core', 'test_strftime') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_intentional_fault(self): # Some programs intentionally segfault themselves, we should compile that into a throw src = r''' int main () { *(volatile char *)0 = 0; return *(volatile char *)0; } ''' self.do_run(src, 'fault on write to 0' if not Settings.ASM_JS else 'abort()') def test_trickystring(self): test_path = path_from_root('tests', 'core', 'test_trickystring') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_statics(self): # static initializers save i16 but load i8 for some reason (or i64 and load i8) if Settings.SAFE_HEAP: Settings.SAFE_HEAP = 3 Settings.SAFE_HEAP_LINES = ['src.cpp:19', 'src.cpp:26', 'src.cpp:28'] test_path = path_from_root('tests', 'core', 'test_statics') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_copyop(self): if self.emcc_args is None: return self.skip('requires emcc') # clang generated code is vulnerable to this, as it uses # memcpy for assignments, with hardcoded numbers of bytes # (llvm-gcc copies items one by one). See QUANTUM_SIZE in # settings.js. test_path = path_from_root('tests', 'core', 'test_copyop') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_memcpy_memcmp(self): test_path = path_from_root('tests', 'core', 'test_memcpy_memcmp') src, output = (test_path + s for s in ('.in', '.out')) def check(result, err): return hashlib.sha1(result).hexdigest() self.do_run_from_file(src, output, output_nicerizer = check) def test_memcpy2(self): test_path = path_from_root('tests', 'core', 'test_memcpy2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_memcpy3(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('need ta2') test_path = path_from_root('tests', 'core', 'test_memcpy3') src, output = (test_path + s for s in ('.c', '.out')) self.do_run_from_file(src, output) def test_memset(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('need ta2') test_path = path_from_root('tests', 'core', 'test_memset') src, output = (test_path + s for s in ('.c', '.out')) self.do_run_from_file(src, output) def test_getopt(self): if self.emcc_args is None: return self.skip('needs emcc for libc') test_path = path_from_root('tests', 'core', 'test_getopt') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, args=['-t', '12', '-n', 'foobar']) def test_getopt_long(self): if self.emcc_args is None: return self.skip('needs emcc for libc') test_path = path_from_root('tests', 'core', 'test_getopt_long') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, args=['--file', 'foobar', '-b']) def test_memmove(self): test_path = path_from_root('tests', 'core', 'test_memmove') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_memmove2(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('need ta2') test_path = path_from_root('tests', 'core', 'test_memmove2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_memmove3(self): test_path = path_from_root('tests', 'core', 'test_memmove3') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_flexarray_struct(self): test_path = path_from_root('tests', 'core', 'test_flexarray_struct') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_bsearch(self): if Settings.QUANTUM_SIZE == 1: return self.skip('Test cannot work with q1') test_path = path_from_root('tests', 'core', 'test_bsearch') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_nestedstructs(self): src = ''' #include #include "emscripten.h" struct base { int x; float y; union { int a; float b; }; char c; }; struct hashtableentry { int key; base data; }; struct hashset { typedef hashtableentry entry; struct chain { entry elem; chain *next; }; // struct chainchunk { chain chains[100]; chainchunk *next; }; }; struct hashtable : hashset { hashtable() { base *b = NULL; entry *e = NULL; chain *c = NULL; printf("*%d,%d,%d,%d,%d,%d|%d,%d,%d,%d,%d,%d,%d,%d|%d,%d,%d,%d,%d,%d,%d,%d,%d,%d*\\n", sizeof(base), int(&(b->x)), int(&(b->y)), int(&(b->a)), int(&(b->b)), int(&(b->c)), sizeof(hashtableentry), int(&(e->key)), int(&(e->data)), int(&(e->data.x)), int(&(e->data.y)), int(&(e->data.a)), int(&(e->data.b)), int(&(e->data.c)), sizeof(hashset::chain), int(&(c->elem)), int(&(c->next)), int(&(c->elem.key)), int(&(c->elem.data)), int(&(c->elem.data.x)), int(&(c->elem.data.y)), int(&(c->elem.data.a)), int(&(c->elem.data.b)), int(&(c->elem.data.c)) ); } }; struct B { char buffer[62]; int last; char laster; char laster2; }; struct Bits { unsigned short A : 1; unsigned short B : 1; unsigned short C : 1; unsigned short D : 1; unsigned short x1 : 1; unsigned short x2 : 1; unsigned short x3 : 1; unsigned short x4 : 1; }; int main() { hashtable t; // Part 2 - the char[] should be compressed, BUT have a padding space at the end so the next // one is aligned properly. Also handle char; char; etc. properly. B *b = NULL; printf("*%d,%d,%d,%d,%d,%d,%d,%d,%d*\\n", int(b), int(&(b->buffer)), int(&(b->buffer[0])), int(&(b->buffer[1])), int(&(b->buffer[2])), int(&(b->last)), int(&(b->laster)), int(&(b->laster2)), sizeof(B)); // Part 3 - bitfields, and small structures Bits *b2 = NULL; printf("*%d*\\n", sizeof(Bits)); return 0; } ''' if Settings.QUANTUM_SIZE == 1: # Compressed memory. Note that sizeof() does give the fat sizes, however! self.do_run(src, '*16,0,1,2,2,3|20,0,1,1,2,3,3,4|24,0,5,0,1,1,2,3,3,4*\n*0,0,0,1,2,62,63,64,72*\n*2*') else: # Bloated memory; same layout as C/C++ self.do_run(src, '*16,0,4,8,8,12|20,0,4,4,8,12,12,16|24,0,20,0,4,4,8,12,12,16*\n*0,0,0,1,2,64,68,69,72*\n*2*') def test_runtimelink(self): return self.skip('BUILD_AS_SHARED_LIB=2 is deprecated') if Building.LLVM_OPTS: return self.skip('LLVM opts will optimize printf into puts in the parent, and the child will still look for puts') if Settings.ASM_JS: return self.skip('asm does not support runtime linking') main, supp = self.setup_runtimelink_test() self.banned_js_engines = [NODE_JS] # node's global scope behaves differently than everything else, needs investigation FIXME Settings.LINKABLE = 1 Settings.BUILD_AS_SHARED_LIB = 2 Settings.NAMED_GLOBALS = 1 self.build(supp, self.get_dir(), self.in_dir('supp.cpp')) shutil.move(self.in_dir('supp.cpp.o.js'), self.in_dir('liblib.so')) Settings.BUILD_AS_SHARED_LIB = 0 Settings.RUNTIME_LINKED_LIBS = ['liblib.so']; self.do_run(main, 'supp: 54,2\nmain: 56\nsupp see: 543\nmain see: 76\nok.') def can_dlfcn(self): if os.environ.get('EMCC_FAST_COMPILER') != '0': self.skip('todo in fastcomp') return False if self.emcc_args and '--memory-init-file' in self.emcc_args: for i in range(len(self.emcc_args)): if self.emcc_args[i] == '--memory-init-file': self.emcc_args = self.emcc_args[:i] + self.emcc_args[i+2:] break if Settings.ASM_JS: Settings.DLOPEN_SUPPORT = 1 else: Settings.NAMED_GLOBALS = 1 if not self.is_emscripten_abi(): self.skip('need asmjs-unknown-emscripten for dlfcn support') return False else: return True def prep_dlfcn_lib(self): if Settings.ASM_JS: Settings.MAIN_MODULE = 0 Settings.SIDE_MODULE = 1 else: Settings.BUILD_AS_SHARED_LIB = 1 Settings.INCLUDE_FULL_LIBRARY = 0 def prep_dlfcn_main(self): if Settings.ASM_JS: Settings.MAIN_MODULE = 1 Settings.SIDE_MODULE = 0 else: Settings.BUILD_AS_SHARED_LIB = 0 Settings.INCLUDE_FULL_LIBRARY = 1 dlfcn_post_build = ''' def process(filename): src = open(filename, 'r').read().replace( '// {{PRE_RUN_ADDITIONS}}', "FS.createLazyFile('/', 'liblib.so', 'liblib.so', true, false);" ) open(filename, 'w').write(src) ''' def test_dlfcn_basic(self): if not self.can_dlfcn(): return self.prep_dlfcn_lib() lib_src = ''' #include class Foo { public: Foo() { printf("Constructing lib object.\\n"); } }; Foo global; ''' dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.cpp') self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() src = ''' #include #include class Bar { public: Bar() { printf("Constructing main object.\\n"); } }; Bar global; int main() { dlopen("liblib.so", RTLD_NOW); return 0; } ''' self.do_run(src, 'Constructing main object.\nConstructing lib object.\n', post_build=self.dlfcn_post_build) def test_dlfcn_i64(self): if not self.can_dlfcn(): return if not Settings.ASM_JS: return self.skip('TODO') self.prep_dlfcn_lib() Settings.EXPORTED_FUNCTIONS = ['_foo'] lib_src = ''' int foo(int x) { return (long long)x / (long long)1234; } ''' dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.c') self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() Settings.EXPORTED_FUNCTIONS = ['_main'] src = r''' #include #include #include typedef int (*intfunc)(int); void *p; int main() { p = malloc(1024); void *lib_handle = dlopen("liblib.so", 0); printf("load %p\n", lib_handle); intfunc x = (intfunc)dlsym(lib_handle, "foo"); printf("foo func %p\n", x); if (p == 0) return 1; printf("|%d|\n", x(81234567)); return 0; } ''' self.do_run(src, '|65830|', post_build=self.dlfcn_post_build) def test_dlfcn_qsort(self): if not self.can_dlfcn(): return if Settings.USE_TYPED_ARRAYS == 2: Settings.CORRECT_SIGNS = 1 # Needed for unsafe optimizations self.prep_dlfcn_lib() Settings.EXPORTED_FUNCTIONS = ['_get_cmp'] lib_src = ''' int lib_cmp(const void* left, const void* right) { const int* a = (const int*) left; const int* b = (const int*) right; if(*a > *b) return 1; else if(*a == *b) return 0; else return -1; } typedef int (*CMP_TYPE)(const void*, const void*); extern "C" CMP_TYPE get_cmp() { return lib_cmp; } ''' dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.cpp') self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() Settings.EXPORTED_FUNCTIONS = ['_main', '_malloc'] src = ''' #include #include #include typedef int (*CMP_TYPE)(const void*, const void*); int main_cmp(const void* left, const void* right) { const int* a = (const int*) left; const int* b = (const int*) right; if(*a < *b) return 1; else if(*a == *b) return 0; else return -1; } int main() { void* lib_handle; CMP_TYPE (*getter_ptr)(); CMP_TYPE lib_cmp_ptr; int arr[5] = {4, 2, 5, 1, 3}; qsort((void*)arr, 5, sizeof(int), main_cmp); printf("Sort with main comparison: "); for (int i = 0; i < 5; i++) { printf("%d ", arr[i]); } printf("\\n"); lib_handle = dlopen("liblib.so", RTLD_NOW); if (lib_handle == NULL) { printf("Could not load lib.\\n"); return 1; } getter_ptr = (CMP_TYPE (*)()) dlsym(lib_handle, "get_cmp"); if (getter_ptr == NULL) { printf("Could not find func.\\n"); return 1; } lib_cmp_ptr = getter_ptr(); qsort((void*)arr, 5, sizeof(int), lib_cmp_ptr); printf("Sort with lib comparison: "); for (int i = 0; i < 5; i++) { printf("%d ", arr[i]); } printf("\\n"); return 0; } ''' self.do_run(src, 'Sort with main comparison: 5 4 3 2 1 *Sort with lib comparison: 1 2 3 4 5 *', output_nicerizer=lambda x, err: x.replace('\n', '*'), post_build=self.dlfcn_post_build) if Settings.ASM_JS and SPIDERMONKEY_ENGINE and os.path.exists(SPIDERMONKEY_ENGINE[0]): out = run_js('liblib.so', engine=SPIDERMONKEY_ENGINE, full_output=True, stderr=STDOUT) if 'asm' in out: self.validate_asmjs(out) def test_dlfcn_data_and_fptr(self): if Settings.ASM_JS: return self.skip('this is not a valid case - libraries should not be able to access their parents globals willy nilly') if not self.can_dlfcn(): return if Building.LLVM_OPTS: return self.skip('LLVM opts will optimize out parent_func') self.prep_dlfcn_lib() lib_src = ''' #include int global = 42; extern void parent_func(); // a function that is defined in the parent void lib_fptr() { printf("Second calling lib_fptr from main.\\n"); parent_func(); // call it also through a pointer, to check indexizing void (*p_f)(); p_f = parent_func; p_f(); } extern "C" void (*func(int x, void(*fptr)()))() { printf("In func: %d\\n", x); fptr(); return lib_fptr; } ''' dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.cpp') Settings.EXPORTED_FUNCTIONS = ['_func'] Settings.EXPORTED_GLOBALS = ['_global'] self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() Settings.LINKABLE = 1 src = ''' #include #include #include typedef void (*FUNCTYPE(int, void(*)()))(); FUNCTYPE func; void EMSCRIPTEN_KEEPALIVE parent_func() { printf("parent_func called from child\\n"); } void main_fptr() { printf("First calling main_fptr from lib.\\n"); } int main() { void* lib_handle; FUNCTYPE* func_fptr; // Test basic lib loading. lib_handle = dlopen("liblib.so", RTLD_NOW); if (lib_handle == NULL) { printf("Could not load lib.\\n"); return 1; } // Test looked up function. func_fptr = (FUNCTYPE*) dlsym(lib_handle, "func"); // Load twice to test cache. func_fptr = (FUNCTYPE*) dlsym(lib_handle, "func"); if (func_fptr == NULL) { printf("Could not find func.\\n"); return 1; } // Test passing function pointers across module bounds. void (*fptr)() = func_fptr(13, main_fptr); fptr(); // Test global data. int* global = (int*) dlsym(lib_handle, "global"); if (global == NULL) { printf("Could not find global.\\n"); return 1; } printf("Var: %d\\n", *global); return 0; } ''' Settings.EXPORTED_FUNCTIONS = ['_main'] Settings.EXPORTED_GLOBALS = [] self.do_run(src, 'In func: 13*First calling main_fptr from lib.*Second calling lib_fptr from main.*parent_func called from child*parent_func called from child*Var: 42*', output_nicerizer=lambda x, err: x.replace('\n', '*'), post_build=self.dlfcn_post_build) def test_dlfcn_alias(self): if Settings.ASM_JS: return self.skip('this is not a valid case - libraries should not be able to access their parents globals willy nilly') Settings.LINKABLE = 1 Settings.NAMED_GLOBALS = 1 if Building.LLVM_OPTS == 2: return self.skip('LLVM LTO will optimize away stuff we expect from the shared library') lib_src = r''' #include extern int parent_global; extern "C" void func() { printf("Parent global: %d.\n", parent_global); } ''' dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.cpp') Settings.BUILD_AS_SHARED_LIB = 1 Settings.EXPORTED_FUNCTIONS = ['_func'] self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) src = r''' #include int parent_global = 123; int main() { void* lib_handle; void (*fptr)(); lib_handle = dlopen("liblib.so", RTLD_NOW); fptr = (void (*)())dlsym(lib_handle, "func"); fptr(); parent_global = 456; fptr(); return 0; } ''' Settings.BUILD_AS_SHARED_LIB = 0 Settings.INCLUDE_FULL_LIBRARY = 1 Settings.EXPORTED_FUNCTIONS = ['_main'] self.do_run(src, 'Parent global: 123.*Parent global: 456.*', output_nicerizer=lambda x, err: x.replace('\n', '*'), post_build=self.dlfcn_post_build, extra_emscripten_args=['-H', 'libc/fcntl.h,libc/sys/unistd.h,poll.h,libc/math.h,libc/time.h,libc/langinfo.h']) Settings.INCLUDE_FULL_LIBRARY = 0 def test_dlfcn_varargs(self): if Settings.ASM_JS: return self.skip('this is not a valid case - libraries should not be able to access their parents globals willy nilly') if not self.can_dlfcn(): return Settings.LINKABLE = 1 if Building.LLVM_OPTS == 2: return self.skip('LLVM LTO will optimize things that prevent shared objects from working') if Settings.QUANTUM_SIZE == 1: return self.skip('FIXME: Add support for this') self.prep_dlfcn_lib() lib_src = r''' void print_ints(int n, ...); extern "C" void func() { print_ints(2, 13, 42); } ''' dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.cpp') Settings.EXPORTED_FUNCTIONS = ['_func'] self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() src = r''' #include #include #include #include void print_ints(int n, ...) { va_list args; va_start(args, n); for (int i = 0; i < n; i++) { printf("%d\n", va_arg(args, int)); } va_end(args); } int main() { void* lib_handle; void (*fptr)(); print_ints(2, 100, 200); lib_handle = dlopen("liblib.so", RTLD_NOW); assert(lib_handle); fptr = (void (*)())dlsym(lib_handle, "func"); fptr(); return 0; } ''' Settings.EXPORTED_FUNCTIONS = ['_main'] self.do_run(src, '100\n200\n13\n42\n', post_build=self.dlfcn_post_build) def test_dlfcn_self(self): if Settings.USE_TYPED_ARRAYS == 1: return self.skip('Does not work with USE_TYPED_ARRAYS=1') if os.environ.get('EMCC_FAST_COMPILER') != '0': return self.skip('todo in fastcomp') Settings.DLOPEN_SUPPORT = 1 def post(filename): with open(filename) as f: for line in f: if 'var SYMBOL_TABLE' in line: table = line break else: raise Exception('Could not find symbol table!') table = table[table.find('{'):table.rfind('}')+1] # ensure there aren't too many globals; we don't want unnamed_addr assert table.count(',') <= 4 test_path = path_from_root('tests', 'core', 'test_dlfcn_self') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, post_build=(None, post)) def test_dlfcn_unique_sig(self): if not self.can_dlfcn(): return self.prep_dlfcn_lib() lib_src = ''' #include int myfunc(int a, int b, int c, int d, int e, int f, int g, int h, int i, int j, int k, int l, int m) { return 13; } ''' Settings.EXPORTED_FUNCTIONS = ['_myfunc'] dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.c') self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() src = ''' #include #include #include typedef int (*FUNCTYPE)(int, int, int, int, int, int, int, int, int, int, int, int, int); int main() { void *lib_handle; FUNCTYPE func_ptr; lib_handle = dlopen("liblib.so", RTLD_NOW); assert(lib_handle != NULL); func_ptr = (FUNCTYPE)dlsym(lib_handle, "myfunc"); assert(func_ptr != NULL); assert(func_ptr(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) == 13); puts("success"); return 0; } ''' Settings.EXPORTED_FUNCTIONS = ['_main', '_malloc'] self.do_run(src, 'success', force_c=True, post_build=self.dlfcn_post_build) def test_dlfcn_stacks(self): if not self.can_dlfcn(): return self.prep_dlfcn_lib() lib_src = ''' #include #include #include int myfunc(const char *input) { char bigstack[1024] = { 0 }; // make sure we didn't just trample the stack! assert(!strcmp(input, "foobar")); snprintf(bigstack, sizeof(bigstack), input); return strlen(bigstack); } ''' Settings.EXPORTED_FUNCTIONS = ['_myfunc'] dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.c') self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() src = ''' #include #include #include #include typedef int (*FUNCTYPE)(const char *); int main() { void *lib_handle; FUNCTYPE func_ptr; char str[128]; snprintf(str, sizeof(str), "foobar"); // HACK: Use strcmp in the main executable so that it doesn't get optimized out and the dynamic library // is able to use it. assert(!strcmp(str, "foobar")); lib_handle = dlopen("liblib.so", RTLD_NOW); assert(lib_handle != NULL); func_ptr = (FUNCTYPE)dlsym(lib_handle, "myfunc"); assert(func_ptr != NULL); assert(func_ptr(str) == 6); puts("success"); return 0; } ''' Settings.EXPORTED_FUNCTIONS = ['_main', '_malloc', '_strcmp'] self.do_run(src, 'success', force_c=True, post_build=self.dlfcn_post_build) def test_dlfcn_funcs(self): if not self.can_dlfcn(): return self.prep_dlfcn_lib() lib_src = r''' #include #include #include typedef void (*voidfunc)(); typedef void (*intfunc)(int); void callvoid(voidfunc f) { f(); } void callint(voidfunc f, int x) { f(x); } void void_0() { printf("void 0\n"); } void void_1() { printf("void 1\n"); } voidfunc getvoid(int i) { switch(i) { case 0: return void_0; case 1: return void_1; default: return NULL; } } void int_0(int x) { printf("int 0 %d\n", x); } void int_1(int x) { printf("int 1 %d\n", x); } intfunc getint(int i) { switch(i) { case 0: return int_0; case 1: return int_1; default: return NULL; } } ''' Settings.EXPORTED_FUNCTIONS = ['_callvoid', '_callint', '_getvoid', '_getint'] dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.c') self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() src = r''' #include #include #include typedef void (*voidfunc)(); typedef void (*intfunc)(int); typedef void (*voidcaller)(voidfunc); typedef void (*intcaller)(intfunc, int); typedef voidfunc (*voidgetter)(int); typedef intfunc (*intgetter)(int); void void_main() { printf("main.\n"); } void int_main(int x) { printf("main %d\n", x); } int main() { printf("go\n"); void *lib_handle; lib_handle = dlopen("liblib.so", RTLD_NOW); assert(lib_handle != NULL); voidcaller callvoid = (voidcaller)dlsym(lib_handle, "callvoid"); assert(callvoid != NULL); callvoid(void_main); intcaller callint = (intcaller)dlsym(lib_handle, "callint"); assert(callint != NULL); callint(int_main, 201); voidgetter getvoid = (voidgetter)dlsym(lib_handle, "getvoid"); assert(getvoid != NULL); callvoid(getvoid(0)); callvoid(getvoid(1)); intgetter getint = (intgetter)dlsym(lib_handle, "getint"); assert(getint != NULL); callint(getint(0), 54); callint(getint(1), 9000); assert(getint(1000) == NULL); puts("ok"); return 0; } ''' Settings.EXPORTED_FUNCTIONS = ['_main', '_malloc'] self.do_run(src, '''go main. main 201 void 0 void 1 int 0 54 int 1 9000 ok ''', force_c=True, post_build=self.dlfcn_post_build) def test_dlfcn_mallocs(self): if not Settings.ASM_JS: return self.skip('needs asm') if not self.can_dlfcn(): return Settings.TOTAL_MEMORY = 64*1024*1024 # will be exhausted without functional malloc/free self.prep_dlfcn_lib() lib_src = r''' #include #include #include #include void *mallocproxy(int n) { return malloc(n); } void freeproxy(void *p) { free(p); } ''' Settings.EXPORTED_FUNCTIONS = ['_mallocproxy', '_freeproxy'] dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.c') self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() src = open(path_from_root('tests', 'dlmalloc_proxy.c')).read() Settings.EXPORTED_FUNCTIONS = ['_main', '_malloc', '_free'] self.do_run(src, '''*294,153*''', force_c=True, post_build=self.dlfcn_post_build) def test_dlfcn_longjmp(self): if not self.can_dlfcn(): return self.prep_dlfcn_lib() lib_src = r''' #include void jumpy(jmp_buf buf) { static int i = 0; i++; if (i == 10) longjmp(buf, i); printf("pre %d\n", i); } ''' Settings.EXPORTED_FUNCTIONS = ['_jumpy'] dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.c') self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() src = r''' #include #include #include #include typedef void (*jumpfunc)(jmp_buf); int main() { printf("go!\n"); void *lib_handle; lib_handle = dlopen("liblib.so", RTLD_NOW); assert(lib_handle != NULL); jumpfunc jumpy = (jumpfunc)dlsym(lib_handle, "jumpy"); assert(jumpy); jmp_buf buf; int jmpval = setjmp(buf); if (jmpval == 0) { while (1) jumpy(buf); } else { printf("out!\n"); } return 0; } ''' Settings.EXPORTED_FUNCTIONS = ['_main', '_malloc', '_free'] self.do_run(src, '''go! pre 1 pre 2 pre 3 pre 4 pre 5 pre 6 pre 7 pre 8 pre 9 out! ''', post_build=self.dlfcn_post_build, force_c=True) def zzztest_dlfcn_exceptions(self): # TODO: make this work. need to forward tempRet0 across modules if not self.can_dlfcn(): return Settings.DISABLE_EXCEPTION_CATCHING = 0 self.prep_dlfcn_lib() lib_src = r''' extern "C" { int ok() { return 65; } int fail() { throw 123; } } ''' Settings.EXPORTED_FUNCTIONS = ['_ok', '_fail'] dirname = self.get_dir() filename = os.path.join(dirname, 'liblib.cpp') self.build(lib_src, dirname, filename) shutil.move(filename + '.o.js', os.path.join(dirname, 'liblib.so')) self.prep_dlfcn_main() src = r''' #include #include #include typedef int (*intfunc)(); int main() { printf("go!\n"); void *lib_handle; lib_handle = dlopen("liblib.so", RTLD_NOW); assert(lib_handle != NULL); intfunc okk = (intfunc)dlsym(lib_handle, "ok"); intfunc faill = (intfunc)dlsym(lib_handle, "fail"); assert(okk && faill); try { printf("ok: %d\n", okk()); } catch(...) { printf("wha\n"); } try { printf("fail: %d\n", faill()); } catch(int x) { printf("int %d\n", x); } try { printf("fail: %d\n", faill()); } catch(double x) { printf("caught %f\n", x); } return 0; } ''' Settings.EXPORTED_FUNCTIONS = ['_main', '_malloc', '_free'] self.do_run(src, '''go! ok: 65 int 123 ok ''', post_build=self.dlfcn_post_build) def test_rand(self): src = r'''#include #include #include int main() { // we need RAND_MAX to be a bitmask (power of 2 minus 1). this assertions guarantees // if RAND_MAX changes the test failure will focus attention on that issue here. assert(RAND_MAX == 0x7fffffff); srand(0xdeadbeef); for(int i = 0; i < 10; ++i) printf("%d\n", rand()); unsigned int seed = 0xdeadbeef; for(int i = 0; i < 10; ++i) printf("%d\n", rand_r(&seed)); return 0; } ''' expected = '''2073540312 730128159 1365227432 1337224527 792390264 1952655743 983994184 1982845871 1210574360 1479617503 2073540312 730128159 1365227432 1337224527 792390264 1952655743 983994184 1982845871 1210574360 1479617503 ''' self.do_run(src, expected) def test_strtod(self): if self.emcc_args is None: return self.skip('needs emcc for libc') if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for accurate math') src = r''' #include #include int main() { char* endptr; printf("\n"); printf("%g\n", strtod("0", &endptr)); printf("%g\n", strtod("0.", &endptr)); printf("%g\n", strtod("0.0", &endptr)); printf("%g\n", strtod("-0.0", &endptr)); printf("%g\n", strtod("1", &endptr)); printf("%g\n", strtod("1.", &endptr)); printf("%g\n", strtod("1.0", &endptr)); printf("%g\n", strtod("z1.0", &endptr)); printf("%g\n", strtod("0.5", &endptr)); printf("%g\n", strtod(".5", &endptr)); printf("%g\n", strtod(".a5", &endptr)); printf("%g\n", strtod("123", &endptr)); printf("%g\n", strtod("123.456", &endptr)); printf("%g\n", strtod("-123.456", &endptr)); printf("%g\n", strtod("1234567891234567890", &endptr)); printf("%g\n", strtod("1234567891234567890e+50", &endptr)); printf("%g\n", strtod("84e+220", &endptr)); printf("%g\n", strtod("123e-50", &endptr)); printf("%g\n", strtod("123e-250", &endptr)); printf("%g\n", strtod("123e-450", &endptr)); printf("%g\n", strtod("0x6", &endptr)); printf("%g\n", strtod("-0x0p+0", &endptr)); char str[] = " 12.34e56end"; printf("%g\n", strtod(str, &endptr)); printf("%d\n", endptr - str); printf("%g\n", strtod("84e+420", &endptr)); printf("%.12f\n", strtod("1.2345678900000000e+08", NULL)); return 0; } ''' expected = ''' 0 0 0 -0 1 1 1 0 0.5 0.5 0 123 123.456 -123.456 1.23457e+18 1.23457e+68 8.4e+221 1.23e-48 1.23e-248 0 6 -0 1.234e+57 10 inf 123456789.000000000000 ''' self.do_run(src, re.sub(r'\n\s+', '\n', expected)) self.do_run(src.replace('strtod', 'strtold'), re.sub(r'\n\s+', '\n', expected)) # XXX add real support for long double def test_strtok(self): test_path = path_from_root('tests', 'core', 'test_strtok') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_parseInt(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('i64 mode 1 requires ta2') if Settings.QUANTUM_SIZE == 1: return self.skip('Q1 and I64_1 do not mix well yet') src = open(path_from_root('tests', 'parseInt', 'src.c'), 'r').read() expected = open(path_from_root('tests', 'parseInt', 'output.txt'), 'r').read() self.do_run(src, expected) def test_transtrcase(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('musl libc needs ta2') test_path = path_from_root('tests', 'core', 'test_transtrcase') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_printf(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('i64 mode 1 requires ta2') self.banned_js_engines = [NODE_JS, V8_ENGINE] # SpiderMonkey and V8 do different things to float64 typed arrays, un-NaNing, etc. src = open(path_from_root('tests', 'printf', 'test.c'), 'r').read() expected = [open(path_from_root('tests', 'printf', 'output.txt'), 'r').read(), open(path_from_root('tests', 'printf', 'output_i64_1.txt'), 'r').read()] self.do_run(src, expected) def test_printf_2(self): test_path = path_from_root('tests', 'core', 'test_printf_2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_vprintf(self): test_path = path_from_root('tests', 'core', 'test_vprintf') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_vsnprintf(self): if self.emcc_args is None: return self.skip('needs i64 math') test_path = path_from_root('tests', 'core', 'test_vsnprintf') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_printf_more(self): test_path = path_from_root('tests', 'core', 'test_printf_more') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_perrar(self): test_path = path_from_root('tests', 'core', 'test_perrar') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_atoX(self): if self.emcc_args is None: return self.skip('requires ta2') test_path = path_from_root('tests', 'core', 'test_atoX') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_strstr(self): if self.run_name.startswith('s_'): return self.skip('musl libc strstr() assumes that C strings can be loaded via i16 and i32 loads.') test_path = path_from_root('tests', 'core', 'test_strstr') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_fnmatch(self): if self.emcc_args is None: return self.skip('requires linking in libc++') test_path = path_from_root('tests', 'core', 'fnmatch') src, output = (test_path + s for s in ('.c', '.out')) self.do_run_from_file(src, output) def test_sscanf(self): if self.emcc_args is None: return self.skip('needs emcc for libc') if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for accurate math') test_path = path_from_root('tests', 'core', 'test_sscanf') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_2(self): # doubles if Settings.USE_TYPED_ARRAYS == 2: for ftype in ['float', 'double']: src = r''' #include int main(){ char strval1[] = "1.2345678901"; char strval2[] = "1.23456789e5"; char strval3[] = "1.23456789E5"; char strval4[] = "1.2345678e-5"; char strval5[] = "1.2345678E-5"; double dblval = 1.2345678901; double tstval; sscanf(strval1, "%lf", &tstval); if(dblval != tstval) printf("FAIL: Values are not equal: %lf %lf\n", dblval, tstval); else printf("Pass: %lf %lf\n", tstval, dblval); sscanf(strval2, "%lf", &tstval); dblval = 123456.789; if(dblval != tstval) printf("FAIL: Values are not equal: %lf %lf\n", dblval, tstval); else printf("Pass: %lf %lf\n", tstval, dblval); sscanf(strval3, "%lf", &tstval); dblval = 123456.789; if(dblval != tstval) printf("FAIL: Values are not equal: %lf %lf\n", dblval, tstval); else printf("Pass: %lf %lf\n", tstval, dblval); sscanf(strval4, "%lf", &tstval); dblval = 0.000012345678; if(dblval != tstval) printf("FAIL: Values are not equal: %lf %lf\n", dblval, tstval); else printf("Pass: %lf %lf\n", tstval, dblval); sscanf(strval5, "%lf", &tstval); dblval = 0.000012345678; if(dblval != tstval) printf("FAIL: Values are not equal: %lf %lf\n", dblval, tstval); else printf("Pass: %lf %lf\n", tstval, dblval); return 0; } ''' if ftype == 'float': self.do_run(src.replace('%lf', '%f').replace('double', 'float'), '''Pass: 1.234568 1.234568 Pass: 123456.789063 123456.789063 Pass: 123456.789063 123456.789063 Pass: 0.000012 0.000012 Pass: 0.000012 0.000012''') else: self.do_run(src, '''Pass: 1.234568 1.234568 Pass: 123456.789000 123456.789000 Pass: 123456.789000 123456.789000 Pass: 0.000012 0.000012 Pass: 0.000012 0.000012''') def test_sscanf_n(self): test_path = path_from_root('tests', 'core', 'test_sscanf_n') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_whitespace(self): test_path = path_from_root('tests', 'core', 'test_sscanf_whitespace') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_other_whitespace(self): Settings.SAFE_HEAP = 0 # use i16s in printf test_path = path_from_root('tests', 'core', 'test_sscanf_other_whitespace') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_3(self): # i64 if not Settings.USE_TYPED_ARRAYS == 2: return self.skip('64-bit sscanf only supported in ta2') test_path = path_from_root('tests', 'core', 'test_sscanf_3') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_4(self): test_path = path_from_root('tests', 'core', 'test_sscanf_4') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_5(self): test_path = path_from_root('tests', 'core', 'test_sscanf_5') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_6(self): test_path = path_from_root('tests', 'core', 'test_sscanf_6') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_skip(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip("need ta2 for full i64") test_path = path_from_root('tests', 'core', 'test_sscanf_skip') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_caps(self): test_path = path_from_root('tests', 'core', 'test_sscanf_caps') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_hex(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('requires ta2') test_path = path_from_root('tests', 'core', 'test_sscanf_hex') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_sscanf_float(self): test_path = path_from_root('tests', 'core', 'test_sscanf_float') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_langinfo(self): src = open(path_from_root('tests', 'langinfo', 'test.c'), 'r').read() expected = open(path_from_root('tests', 'langinfo', 'output.txt'), 'r').read() self.do_run(src, expected, extra_emscripten_args=['-H', 'libc/langinfo.h']) def test_files(self): self.banned_js_engines = [SPIDERMONKEY_ENGINE] # closure can generate variables called 'gc', which pick up js shell stuff if self.emcc_args is not None and '-O2' in self.emcc_args: self.emcc_args += ['--closure', '1'] # Use closure here, to test we don't break FS stuff self.emcc_args = filter(lambda x: x != '-g', self.emcc_args) # ensure we test --closure 1 --memory-init-file 1 (-g would disable closure) self.emcc_args += ["-s", "CHECK_HEAP_ALIGN=0"] # disable heap align check here, it mixes poorly with closure Settings.CORRECT_SIGNS = 1 # Just so our output is what we expect. Can flip them both. post = ''' def process(filename): src = \'\'\' var Module = { 'noFSInit': true, 'preRun': function() { FS.createLazyFile('/', 'test.file', 'test.file', true, false); // Test FS_* exporting Module['FS_createDataFile']('/', 'somefile.binary', [100, 200, 50, 25, 10, 77, 123], true, false); // 200 becomes -56, since signed chars are used in memory var test_files_input = 'hi there!'; var test_files_input_index = 0; FS.init(function() { return test_files_input.charCodeAt(test_files_input_index++) || null; }); } }; \'\'\' + open(filename, 'r').read() open(filename, 'w').write(src) ''' other = open(os.path.join(self.get_dir(), 'test.file'), 'w') other.write('some data'); other.close() src = open(path_from_root('tests', 'files.cpp'), 'r').read() mem_file = 'src.cpp.o.js.mem' try_delete(mem_file) self.do_run(src, ('size: 7\ndata: 100,-56,50,25,10,77,123\nloop: 100 -56 50 25 10 77 123 \ninput:hi there!\ntexto\n$\n5 : 10,30,20,11,88\nother=some data.\nseeked=me da.\nseeked=ata.\nseeked=ta.\nfscanfed: 10 - hello\nok.\ntexte\n', 'size: 7\ndata: 100,-56,50,25,10,77,123\nloop: 100 -56 50 25 10 77 123 \ninput:hi there!\ntexto\ntexte\n$\n5 : 10,30,20,11,88\nother=some data.\nseeked=me da.\nseeked=ata.\nseeked=ta.\nfscanfed: 10 - hello\nok.\n'), post_build=post, extra_emscripten_args=['-H', 'libc/fcntl.h']) if self.emcc_args and '--memory-init-file' in self.emcc_args: assert os.path.exists(mem_file) def test_files_m(self): # Test for Module.stdin etc. Settings.CORRECT_SIGNS = 1 post = ''' def process(filename): src = \'\'\' var Module = { data: [10, 20, 40, 30], stdin: function() { return Module.data.pop() || null }, stdout: function(x) { Module.print('got: ' + x) } }; \'\'\' + open(filename, 'r').read() open(filename, 'w').write(src) ''' src = r''' #include #include int main () { char c; fprintf(stderr, "isatty? %d,%d,%d\n", isatty(fileno(stdin)), isatty(fileno(stdout)), isatty(fileno(stderr))); while ((c = fgetc(stdin)) != EOF) { putc(c+5, stdout); } return 0; } ''' def clean(out, err): return '\n'.join(filter(lambda line: 'warning' not in line, (out + err).split('\n'))) self.do_run(src, ('got: 35\ngot: 45\ngot: 25\ngot: 15\nisatty? 0,0,1\n', 'isatty? 0,0,1\ngot: 35\ngot: 45\ngot: 25\ngot: 15\n'), post_build=post, output_nicerizer=clean) def test_mount(self): src = open(path_from_root('tests', 'fs', 'test_mount.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_fwrite_0(self): test_path = path_from_root('tests', 'core', 'test_fwrite_0') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_fgetc_ungetc(self): src = open(path_from_root('tests', 'stdio', 'test_fgetc_ungetc.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_fgetc_unsigned(self): if self.emcc_args is None: return self.skip('requires emcc') src = r''' #include int main() { FILE *file = fopen("file_with_byte_234.txt", "rb"); int c = fgetc(file); printf("*%d\n", c); } ''' open('file_with_byte_234.txt', 'wb').write('\xea') self.emcc_args += ['--embed-file', 'file_with_byte_234.txt'] self.do_run(src, '*234\n') def test_fgets_eol(self): if self.emcc_args is None: return self.skip('requires emcc') src = r''' #include char buf[32]; int main() { char *r = "SUCCESS"; FILE *f = fopen("eol.txt", "r"); while (fgets(buf, 32, f) != NULL) { if (buf[0] == '\0') { r = "FAIL"; break; } } printf("%s\n", r); fclose(f); return 0; } ''' open('eol.txt', 'wb').write('\n') self.emcc_args += ['--embed-file', 'eol.txt'] self.do_run(src, 'SUCCESS\n') def test_fscanf(self): if self.emcc_args is None: return self.skip('requires emcc') open(os.path.join(self.get_dir(), 'three_numbers.txt'), 'w').write('''-1 0.1 -.1''') src = r''' #include #include #include int main() { float x = FLT_MAX, y = FLT_MAX, z = FLT_MAX; FILE* fp = fopen("three_numbers.txt", "r"); if (fp) { int match = fscanf(fp, " %f %f %f ", &x, &y, &z); printf("match = %d\n", match); printf("x = %0.1f, y = %0.1f, z = %0.1f\n", x, y, z); } else { printf("failed to open three_numbers.txt\n"); } return 0; } ''' self.emcc_args += ['--embed-file', 'three_numbers.txt'] self.do_run(src, 'match = 3\nx = -1.0, y = 0.1, z = -0.1\n') def test_fileno(self): if self.emcc_args is None: return self.skip('requires emcc') open(os.path.join(self.get_dir(), 'empty.txt'), 'w').write('') src = r''' #include #include int main() { FILE* fp = fopen("empty.txt", "r"); if (fp) { printf("%d\n", fp); printf("%d\n", fileno(fp)); printf("%d\n", fileno((FILE*)42)); // nonexistent stream } else { printf("failed to open empty.txt\n"); } return 0; } ''' self.emcc_args += ['--embed-file', 'empty.txt'] self.do_run(src, '4\n3\n-1\n') def test_readdir(self): src = open(path_from_root('tests', 'dirent', 'test_readdir.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_readdir_empty(self): src = open(path_from_root('tests', 'dirent', 'test_readdir_empty.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_stat(self): src = open(path_from_root('tests', 'stat', 'test_stat.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_stat_chmod(self): src = open(path_from_root('tests', 'stat', 'test_chmod.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_stat_mknod(self): src = open(path_from_root('tests', 'stat', 'test_mknod.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_fcntl(self): add_pre_run = ''' def process(filename): src = open(filename, 'r').read().replace( '// {{PRE_RUN_ADDITIONS}}', "FS.createDataFile('/', 'test', 'abcdef', true, true);" ) open(filename, 'w').write(src) ''' src = open(path_from_root('tests', 'fcntl', 'src.c'), 'r').read() expected = open(path_from_root('tests', 'fcntl', 'output.txt'), 'r').read() self.do_run(src, expected, post_build=add_pre_run, extra_emscripten_args=['-H', 'libc/fcntl.h']) def test_fcntl_open(self): src = open(path_from_root('tests', 'fcntl-open', 'src.c'), 'r').read() expected = open(path_from_root('tests', 'fcntl-open', 'output.txt'), 'r').read() self.do_run(src, expected, force_c=True, extra_emscripten_args=['-H', 'libc/fcntl.h']) def test_fcntl_misc(self): add_pre_run = ''' def process(filename): src = open(filename, 'r').read().replace( '// {{PRE_RUN_ADDITIONS}}', "FS.createDataFile('/', 'test', 'abcdef', true, true);" ) open(filename, 'w').write(src) ''' src = open(path_from_root('tests', 'fcntl-misc', 'src.c'), 'r').read() expected = open(path_from_root('tests', 'fcntl-misc', 'output.txt'), 'r').read() self.do_run(src, expected, post_build=add_pre_run, extra_emscripten_args=['-H', 'libc/fcntl.h']) def test_poll(self): add_pre_run = ''' def process(filename): src = open(filename, 'r').read().replace( '// {{PRE_RUN_ADDITIONS}}', \'\'\' var dummy_device = FS.makedev(64, 0); FS.registerDevice(dummy_device, {}); FS.createDataFile('/', 'file', 'abcdef', true, true); FS.mkdev('/device', dummy_device); \'\'\' ) open(filename, 'w').write(src) ''' test_path = path_from_root('tests', 'core', 'test_poll') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, post_build=add_pre_run, extra_emscripten_args=['-H', 'libc/fcntl.h,poll.h']) def test_statvfs(self): test_path = path_from_root('tests', 'core', 'test_statvfs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_libgen(self): test_path = path_from_root('tests', 'core', 'test_libgen') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_utime(self): src = open(path_from_root('tests', 'utime', 'test_utime.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_utf(self): self.banned_js_engines = [SPIDERMONKEY_ENGINE] # only node handles utf well Settings.EXPORTED_FUNCTIONS = ['_main', '_malloc'] test_path = path_from_root('tests', 'core', 'test_utf') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_utf32(self): if self.emcc_args is None: return self.skip('need libc for wcslen()') if not self.is_emscripten_abi(): return self.skip('this test uses inline js, which requires asmjs-unknown-emscripten') self.do_run(open(path_from_root('tests', 'utf32.cpp')).read(), 'OK.') self.do_run(open(path_from_root('tests', 'utf32.cpp')).read(), 'OK.', args=['-fshort-wchar']) def test_wprintf(self): if self.emcc_args is None: return self.skip('requires libcxx') test_path = path_from_root('tests', 'core', 'test_wprintf') src, output = (test_path + s for s in ('.c', '.out')) self.do_run_from_file(src, output) def test_direct_string_constant_usage(self): if self.emcc_args is None: return self.skip('requires libcxx') test_path = path_from_root('tests', 'core', 'test_direct_string_constant_usage') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_std_cout_new(self): if self.emcc_args is None: return self.skip('requires emcc') test_path = path_from_root('tests', 'core', 'test_std_cout_new') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_istream(self): if self.emcc_args is None: return self.skip('requires libcxx') test_path = path_from_root('tests', 'core', 'test_istream') src, output = (test_path + s for s in ('.in', '.out')) for linkable in [0]:#, 1]: print linkable Settings.LINKABLE = linkable # regression check for issue #273 self.do_run_from_file(src, output) def test_fs_base(self): Settings.INCLUDE_FULL_LIBRARY = 1 try: addJS = ''' def process(filename): import tools.shared as shared src = open(filename, 'r').read().replace('FS.init();', '').replace( # Disable normal initialization, replace with ours '// {{PRE_RUN_ADDITIONS}}', open(shared.path_from_root('tests', 'filesystem', 'src.js'), 'r').read()) open(filename, 'w').write(src) ''' src = 'int main() {return 0;}\n' expected = open(path_from_root('tests', 'filesystem', 'output.txt'), 'r').read() self.do_run(src, expected, post_build=addJS, extra_emscripten_args=['-H', 'libc/fcntl.h,libc/sys/unistd.h,poll.h,libc/math.h,libc/langinfo.h,libc/time.h']) finally: Settings.INCLUDE_FULL_LIBRARY = 0 def test_fs_nodefs_rw(self): if self.emcc_args is None: return self.skip('requires emcc') if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') src = open(path_from_root('tests', 'fs', 'test_nodefs_rw.c'), 'r').read() self.do_run(src, 'success', force_c=True, js_engines=[NODE_JS]) def test_fs_trackingdelegate(self): src = path_from_root('tests', 'fs', 'test_trackingdelegate.c') out = path_from_root('tests', 'fs', 'test_trackingdelegate.out') self.do_run_from_file(src, out) def test_fs_writeFile(self): if self.emcc_args is None: return self.skip('requires emcc') self.emcc_args += ['-s', 'DISABLE_EXCEPTION_CATCHING=1'] # see issue 2334 src = path_from_root('tests', 'fs', 'test_writeFile.cc') out = path_from_root('tests', 'fs', 'test_writeFile.out') self.do_run_from_file(src, out) def test_unistd_access(self): self.clear() if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') for fs in ['MEMFS', 'NODEFS']: src = open(path_from_root('tests', 'unistd', 'access.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'access.out'), 'r').read() Building.COMPILER_TEST_OPTS += ['-D' + fs] self.do_run(src, expected, js_engines=[NODE_JS]) def test_unistd_curdir(self): if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') src = open(path_from_root('tests', 'unistd', 'curdir.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'curdir.out'), 'r').read() self.do_run(src, expected) def test_unistd_close(self): src = open(path_from_root('tests', 'unistd', 'close.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'close.out'), 'r').read() self.do_run(src, expected) def test_unistd_confstr(self): src = open(path_from_root('tests', 'unistd', 'confstr.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'confstr.out'), 'r').read() self.do_run(src, expected, extra_emscripten_args=['-H', 'libc/unistd.h']) def test_unistd_ttyname(self): if self.run_name.startswith('s_'): return self.skip('musl libc strstr() assumes that C strings can be loaded via i16 and i32 loads.') src = open(path_from_root('tests', 'unistd', 'ttyname.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_unistd_dup(self): src = open(path_from_root('tests', 'unistd', 'dup.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'dup.out'), 'r').read() self.do_run(src, expected) def test_unistd_pathconf(self): src = open(path_from_root('tests', 'unistd', 'pathconf.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'pathconf.out'), 'r').read() self.do_run(src, expected) def test_unistd_truncate(self): self.clear() if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') for fs in ['MEMFS', 'NODEFS']: src = open(path_from_root('tests', 'unistd', 'truncate.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'truncate.out'), 'r').read() Building.COMPILER_TEST_OPTS += ['-D' + fs] self.do_run(src, expected, js_engines=[NODE_JS]) def test_unistd_swab(self): src = open(path_from_root('tests', 'unistd', 'swab.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'swab.out'), 'r').read() self.do_run(src, expected) def test_unistd_isatty(self): src = open(path_from_root('tests', 'unistd', 'isatty.c'), 'r').read() self.do_run(src, 'success', force_c=True) def test_unistd_sysconf(self): src = open(path_from_root('tests', 'unistd', 'sysconf.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'sysconf.out'), 'r').read() self.do_run(src, expected) def test_unistd_login(self): src = open(path_from_root('tests', 'unistd', 'login.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'login.out'), 'r').read() self.do_run(src, expected) def test_unistd_unlink(self): self.clear() if self.emcc_args is None: return self.skip('requires emcc') if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') for fs in ['MEMFS', 'NODEFS']: src = open(path_from_root('tests', 'unistd', 'unlink.c'), 'r').read() Building.COMPILER_TEST_OPTS += ['-D' + fs] self.do_run(src, 'success', force_c=True, js_engines=[NODE_JS]) def test_unistd_links(self): self.clear() if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') for fs in ['MEMFS', 'NODEFS']: if WINDOWS and fs == 'NODEFS': print >> sys.stderr, 'Skipping NODEFS part of this test for test_unistd_links on Windows, since it would require administrative privileges.' # Also, other detected discrepancies if you do end up running this test on NODEFS: # test expects /, but Windows gives \ as path slashes. # Calling readlink() on a non-link gives error 22 EINVAL on Unix, but simply error 0 OK on Windows. continue src = open(path_from_root('tests', 'unistd', 'links.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'links.out'), 'r').read() Building.COMPILER_TEST_OPTS += ['-D' + fs] self.do_run(src, expected, js_engines=[NODE_JS]) def test_unistd_sleep(self): src = open(path_from_root('tests', 'unistd', 'sleep.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'sleep.out'), 'r').read() self.do_run(src, expected) def test_unistd_io(self): self.clear() if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') if self.run_name == 'o2': return self.skip('non-asm optimized builds can fail with inline js') if self.emcc_args is None: return self.skip('requires emcc') for fs in ['MEMFS', 'NODEFS']: src = open(path_from_root('tests', 'unistd', 'io.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'io.out'), 'r').read() Building.COMPILER_TEST_OPTS += ['-D' + fs] self.do_run(src, expected, js_engines=[NODE_JS]) def test_unistd_misc(self): if self.emcc_args is None: return self.skip('requires emcc') if not self.is_emscripten_abi(): return self.skip('asmjs-unknown-emscripten needed for inline js') for fs in ['MEMFS', 'NODEFS']: src = open(path_from_root('tests', 'unistd', 'misc.c'), 'r').read() expected = open(path_from_root('tests', 'unistd', 'misc.out'), 'r').read() Building.COMPILER_TEST_OPTS += ['-D' + fs] self.do_run(src, expected, js_engines=[NODE_JS]) def test_uname(self): test_path = path_from_root('tests', 'core', 'test_uname') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_env(self): src = open(path_from_root('tests', 'env', 'src.c'), 'r').read() expected = open(path_from_root('tests', 'env', 'output.txt'), 'r').read() self.do_run(src, expected) def test_systypes(self): src = open(path_from_root('tests', 'systypes', 'src.c'), 'r').read() expected = open(path_from_root('tests', 'systypes', 'output.txt'), 'r').read() self.do_run(src, expected) def test_getloadavg(self): test_path = path_from_root('tests', 'core', 'test_getloadavg') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_nl_types(self): test_path = path_from_root('tests', 'core', 'test_nl_types') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_799(self): src = open(path_from_root('tests', '799.cpp'), 'r').read() self.do_run(src, '''Set PORT family: 0, port: 3979 Get PORT family: 0 PORT: 3979 ''') def test_ctype(self): # The bit fiddling done by the macros using __ctype_b_loc requires this. Settings.CORRECT_SIGNS = 1 src = open(path_from_root('tests', 'ctype', 'src.c'), 'r').read() expected = open(path_from_root('tests', 'ctype', 'output.txt'), 'r').read() self.do_run(src, expected) def test_strcasecmp(self): test_path = path_from_root('tests', 'core', 'test_strcasecmp') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_atomic(self): test_path = path_from_root('tests', 'core', 'test_atomic') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_atomic_cxx(self): if self.emcc_args is None: return self.skip('needs emcc') if os.environ.get('EMCC_FAST_COMPILER') == '0': return self.skip('needs fastcomp') test_path = path_from_root('tests', 'core', 'test_atomic_cxx') src, output = (test_path + s for s in ('.cpp', '.txt')) Building.COMPILER_TEST_OPTS += ['-std=c++11'] self.do_run_from_file(src, output) def test_phiundef(self): test_path = path_from_root('tests', 'core', 'test_phiundef') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_netinet_in(self): src = open(path_from_root('tests', 'netinet', 'in.cpp'), 'r').read() expected = open(path_from_root('tests', 'netinet', 'in.out'), 'r').read() self.do_run(src, expected) # libc++ tests def test_iostream(self): if Settings.QUANTUM_SIZE == 1: return self.skip("we don't support libcxx in q1") if self.emcc_args is None: return self.skip('needs ta2 and emcc') src = ''' #include int main() { std::cout << "hello world" << std::endl << 77 << "." << std::endl; return 0; } ''' # FIXME: should not have so many newlines in output here self.do_run(src, 'hello world\n77.\n') def test_stdvec(self): if self.emcc_args is None: return self.skip('requires emcc') test_path = path_from_root('tests', 'core', 'test_stdvec') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_reinterpreted_ptrs(self): if self.emcc_args is None: return self.skip('needs emcc and libc') test_path = path_from_root('tests', 'core', 'test_reinterpreted_ptrs') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_jansson(self): return self.skip('currently broken') if Settings.USE_TYPED_ARRAYS != 2: return self.skip('requires ta2') if Settings.SAFE_HEAP: return self.skip('jansson is not safe-heap safe') src = ''' #include #include #include int main() { const char* jsonString = "{\\"key\\": \\"value\\",\\"array\\": [\\"array_item1\\",\\"array_item2\\",\\"array_item3\\"],\\"dict\\":{\\"number\\": 3,\\"float\\": 2.2}}"; json_error_t error; json_t *root = json_loadb(jsonString, strlen(jsonString), 0, &error); if(!root) { printf("Node `root` is `null`."); return 0; } if(!json_is_object(root)) { printf("Node `root` is no object."); return 0; } printf("%s\\n", json_string_value(json_object_get(root, "key"))); json_t *array = json_object_get(root, "array"); if(!array) { printf("Node `array` is `null`."); return 0; } if(!json_is_array(array)) { printf("Node `array` is no array."); return 0; } for(size_t i=0; i struct one_const { long a; }; struct two_consts { long a; long b; }; union some_consts { struct one_const one; struct two_consts two; }; union some_consts my_consts = {{ 1 }}; struct one_const addr_of_my_consts = { (long)(&my_consts) }; int main(void) { printf("%li\n", !!addr_of_my_consts.a); return 0; } ''', '1') ### 'Medium' tests def test_fannkuch(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('musl libc needs ta2') try: if self.run_name == 'slow2' or self.run_name == 'slow2asm': old_target = os.environ.get('EMCC_LLVM_TARGET') or '' os.environ['EMCC_LLVM_TARGET'] = "asmjs-unknown-emscripten" # testing for asm-emscripten target on non-fastcomp results = [ (1,0), (2,1), (3,2), (4,4), (5,7), (6,10), (7, 16), (8,22) ] for i, j in results: src = open(path_from_root('tests', 'fannkuch.cpp'), 'r').read() self.do_run(src, 'Pfannkuchen(%d) = %d.' % (i,j), [str(i)], no_build=i>1) finally: if self.run_name == 'slow2' or self.run_name == 'slow2asm': if old_target: os.environ['EMCC_LLVM_TARGET'] = old_target else: del os.environ['EMCC_LLVM_TARGET'] def test_raytrace(self): if self.emcc_args is None: return self.skip('requires emcc') if Settings.USE_TYPED_ARRAYS == 2: return self.skip('Relies on double value rounding, extremely sensitive') src = open(path_from_root('tests', 'raytrace.cpp'), 'r').read().replace('double', 'float') output = open(path_from_root('tests', 'raytrace.ppm'), 'r').read() self.do_run(src, output, ['3', '16'])#, build_ll_hook=self.do_autodebug) def test_fasta(self): if self.emcc_args is None: return self.skip('requires emcc') results = [ (1,'''GG*ctt**tgagc*'''), (20,'''GGCCGGGCGCGGTGGCTCACGCCTGTAATCCCAGCACTTT*cttBtatcatatgctaKggNcataaaSatgtaaaDcDRtBggDtctttataattcBgtcg**tacgtgtagcctagtgtttgtgttgcgttatagtctatttgtggacacagtatggtcaaa**tgacgtcttttgatctgacggcgttaacaaagatactctg*'''), (50,'''GGCCGGGCGCGGTGGCTCACGCCTGTAATCCCAGCACTTTGGGAGGCCGAGGCGGGCGGA*TCACCTGAGGTCAGGAGTTCGAGACCAGCCTGGCCAACAT*cttBtatcatatgctaKggNcataaaSatgtaaaDcDRtBggDtctttataattcBgtcg**tactDtDagcctatttSVHtHttKtgtHMaSattgWaHKHttttagacatWatgtRgaaa**NtactMcSMtYtcMgRtacttctWBacgaa**agatactctgggcaacacacatacttctctcatgttgtttcttcggacctttcataacct**ttcctggcacatggttagctgcacatcacaggattgtaagggtctagtggttcagtgagc**ggaatatcattcgtcggtggtgttaatctatctcggtgtagcttataaatgcatccgtaa**gaatattatgtttatttgtcggtacgttcatggtagtggtgtcgccgatttagacgtaaa**ggcatgtatg*''') ] for precision in [0, 1, 2]: Settings.PRECISE_F32 = precision for t in ['float', 'double']: print precision, t src = open(path_from_root('tests', 'fasta.cpp'), 'r').read().replace('double', t) for i, j in results: self.do_run(src, j, [str(i)], lambda x, err: x.replace('\n', '*'), no_build=i>1) shutil.copyfile('src.cpp.o.js', '%d_%s.js' % (precision, t)) def test_whets(self): if not Settings.ASM_JS: return self.skip('mainly a test for asm validation here') self.do_run(open(path_from_root('tests', 'whets.cpp')).read(), 'Single Precision C Whetstone Benchmark') def test_dlmalloc(self): if self.emcc_args is None: self.emcc_args = [] # dlmalloc auto-inclusion is only done if we use emcc self.banned_js_engines = [NODE_JS] # slower, and fail on 64-bit Settings.CORRECT_SIGNS = 1 Settings.TOTAL_MEMORY = 128*1024*1024 # needed with typed arrays src = open(path_from_root('system', 'lib', 'dlmalloc.c'), 'r').read() + '\n\n\n' + open(path_from_root('tests', 'dlmalloc_test.c'), 'r').read() self.do_run(src, '*1,0*', ['200', '1']) self.do_run(src, '*400,0*', ['400', '400'], no_build=True) # Linked version src = open(path_from_root('tests', 'dlmalloc_test.c'), 'r').read() self.do_run(src, '*1,0*', ['200', '1'], extra_emscripten_args=['-m']) self.do_run(src, '*400,0*', ['400', '400'], extra_emscripten_args=['-m'], no_build=True) if self.emcc_args == []: # TODO: do this in other passes too, passing their opts into emcc # emcc should build in dlmalloc automatically, and do all the sign correction etc. for it try_delete(os.path.join(self.get_dir(), 'src.cpp.o.js')) output = Popen([PYTHON, EMCC, path_from_root('tests', 'dlmalloc_test.c'), '-s', 'TOTAL_MEMORY=' + str(128*1024*1024), '-o', os.path.join(self.get_dir(), 'src.cpp.o.js')], stdout=PIPE, stderr=self.stderr_redirect).communicate() self.do_run('x', '*1,0*', ['200', '1'], no_build=True) self.do_run('x', '*400,0*', ['400', '400'], no_build=True) # The same for new and all its variants src = open(path_from_root('tests', 'new.cpp')).read() for new, delete in [ ('malloc(100)', 'free'), ('new char[100]', 'delete[]'), ('new Structy', 'delete'), ('new int', 'delete'), ('new Structy[10]', 'delete[]'), ]: self.do_run(src.replace('{{{ NEW }}}', new).replace('{{{ DELETE }}}', delete), '*1,0*') def test_dlmalloc_partial(self): if self.emcc_args is None: return self.skip('only emcc will link in dlmalloc') # present part of the symbols of dlmalloc, not all src = open(path_from_root('tests', 'new.cpp')).read().replace('{{{ NEW }}}', 'new int').replace('{{{ DELETE }}}', 'delete') + ''' void * operator new(size_t size) { printf("new %d!\\n", size); return malloc(size); } ''' self.do_run(src, 'new 4!\n*1,0*') def test_dlmalloc_partial_2(self): if self.emcc_args is None or 'SAFE_HEAP' in str(self.emcc_args) or 'CHECK_HEAP_ALIGN' in str(self.emcc_args): return self.skip('only emcc will link in dlmalloc, and we do unsafe stuff') # present part of the symbols of dlmalloc, not all. malloc is harder to link than new which is weak. test_path = path_from_root('tests', 'core', 'test_dlmalloc_partial_2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_libcxx(self): if self.emcc_args is None: return self.skip('requires emcc') self.do_run(open(path_from_root('tests', 'hashtest.cpp')).read(), 'june -> 30\nPrevious (in alphabetical order) is july\nNext (in alphabetical order) is march') self.do_run(''' #include #include int main() { std::set *fetchOriginatorNums = new std::set(); fetchOriginatorNums->insert(171); printf("hello world\\n"); return 0; } ''', 'hello world'); def test_typeid(self): test_path = path_from_root('tests', 'core', 'test_typeid') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_static_variable(self): if self.emcc_args is None: Settings.SAFE_HEAP = 0 # LLVM mixes i64 and i8 in the guard check test_path = path_from_root('tests', 'core', 'test_static_variable') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_fakestat(self): test_path = path_from_root('tests', 'core', 'test_fakestat') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_mmap(self): if self.emcc_args is None: return self.skip('requires emcc') Settings.TOTAL_MEMORY = 128*1024*1024 test_path = path_from_root('tests', 'core', 'test_mmap') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) self.do_run_from_file(src, output, force_c=True) def test_mmap_file(self): if self.emcc_args is None: return self.skip('requires emcc') for extra_args in [[], ['--no-heap-copy']]: self.emcc_args += ['--embed-file', 'data.dat'] + extra_args open(self.in_dir('data.dat'), 'w').write('data from the file ' + ('.' * 9000)) src = open(path_from_root('tests', 'mmap_file.c')).read() self.do_run(src, '*\ndata from the file .\nfrom the file ......\n*\n') def test_cubescript(self): if self.emcc_args is None: return self.skip('requires emcc') if self.run_name == 'o2': self.emcc_args += ['--closure', '1'] # Use closure here for some additional coverage Building.COMPILER_TEST_OPTS = filter(lambda x: x != '-g', Building.COMPILER_TEST_OPTS) # remove -g, so we have one test without it by default if self.emcc_args is None: Settings.SAFE_HEAP = 0 # Has some actual loads of unwritten-to places, in the C++ code... # Overflows happen in hash loop Settings.CORRECT_OVERFLOWS = 1 Settings.CHECK_OVERFLOWS = 0 if Settings.USE_TYPED_ARRAYS == 2: Settings.CORRECT_SIGNS = 1 self.do_run(path_from_root('tests', 'cubescript'), '*\nTemp is 33\n9\n5\nhello, everyone\n*', main_file='command.cpp') assert 'asm1' in test_modes if self.run_name == 'asm1': print 'verifing postsets' generated = open('src.cpp.o.js').read() generated = re.sub(r'\n+[ \n]*\n+', '\n', generated) main = generated[generated.find('function runPostSets'):] main = main[:main.find('\n}')] assert main.count('\n') <= 7, ('must not emit too many postSets: %d' % main.count('\n')) + ' : ' + main def test_simd(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('needs ta2') if Settings.ASM_JS: Settings.ASM_JS = 2 # does not validate test_path = path_from_root('tests', 'core', 'test_simd') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_simd2(self): if Settings.ASM_JS: Settings.ASM_JS = 2 # does not validate test_path = path_from_root('tests', 'core', 'test_simd2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_simd3(self): return self.skip('FIXME: this appears to be broken') if Settings.USE_TYPED_ARRAYS != 2: return self.skip('needs ta2') if Settings.ASM_JS: Settings.ASM_JS = 2 # does not validate test_path = path_from_root('tests', 'core', 'test_simd3') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_gcc_unmangler(self): if os.environ.get('EMCC_FAST_COMPILER') == '0': Settings.NAMED_GLOBALS = 1 # test coverage for this; fastcomp never names globals Building.COMPILER_TEST_OPTS += ['-I' + path_from_root('third_party'), '-Wno-warn-absolute-paths'] self.do_run(open(path_from_root('third_party', 'gcc_demangler.c')).read(), '*d_demangle(char const*, int, unsigned int*)*', args=['_ZL10d_demanglePKciPj']) #### Code snippet that is helpful to search for nonportable optimizations #### #global LLVM_OPT_OPTS #for opt in ['-aa-eval', '-adce', '-always-inline', '-argpromotion', '-basicaa', '-basiccg', '-block-placement', '-break-crit-edges', '-codegenprepare', '-constmerge', '-constprop', '-correlated-propagation', '-count-aa', '-dce', '-deadargelim', '-deadtypeelim', '-debug-aa', '-die', '-domfrontier', '-domtree', '-dse', '-extract-blocks', '-functionattrs', '-globaldce', '-globalopt', '-globalsmodref-aa', '-gvn', '-indvars', '-inline', '-insert-edge-profiling', '-insert-optimal-edge-profiling', '-instcombine', '-instcount', '-instnamer', '-internalize', '-intervals', '-ipconstprop', '-ipsccp', '-iv-users', '-jump-threading', '-lazy-value-info', '-lcssa', '-lda', '-libcall-aa', '-licm', '-lint', '-live-values', '-loop-deletion', '-loop-extract', '-loop-extract-single', '-loop-index-split', '-loop-reduce', '-loop-rotate', '-loop-unroll', '-loop-unswitch', '-loops', '-loopsimplify', '-loweratomic', '-lowerinvoke', '-lowersetjmp', '-lowerswitch', '-mem2reg', '-memcpyopt', '-memdep', '-mergefunc', '-mergereturn', '-module-debuginfo', '-no-aa', '-no-profile', '-partial-inliner', '-partialspecialization', '-pointertracking', '-postdomfrontier', '-postdomtree', '-preverify', '-prune-eh', '-reassociate', '-reg2mem', '-regions', '-scalar-evolution', '-scalarrepl', '-sccp', '-scev-aa', '-simplify-libcalls', '-simplify-libcalls-halfpowr', '-simplifycfg', '-sink', '-split-geps', '-sretpromotion', '-strip', '-strip-dead-debug-info', '-strip-dead-prototypes', '-strip-debug-declare', '-strip-nondebug', '-tailcallelim', '-tailduplicate', '-targetdata', '-tbaa']: # LLVM_OPT_OPTS = [opt] # try: # self.do_run(path_from_root(['third_party']), '*d_demangle(char const*, int, unsigned int*)*', args=['_ZL10d_demanglePKciPj'], main_file='gcc_demangler.c') # print opt, "ok" # except: # print opt, "FAIL" def test_lua(self): if self.emcc_args is None: return self.skip('requires emcc') if Settings.QUANTUM_SIZE == 1: return self.skip('TODO: make this work') for aggro in ([0, 1] if Settings.ASM_JS and '-O2' in self.emcc_args else [0]): print aggro Settings.AGGRESSIVE_VARIABLE_ELIMINATION = aggro self.do_run('', 'hello lua world!\n17\n1\n2\n3\n4\n7', args=['-e', '''print("hello lua world!");print(17);for x = 1,4 do print(x) end;print(10-3)'''], libraries=self.get_library('lua', [os.path.join('src', 'lua'), os.path.join('src', 'liblua.a')], make=['make', 'generic'], configure=None), includes=[path_from_root('tests', 'lua')], output_nicerizer=lambda string, err: (string + err).replace('\n\n', '\n').replace('\n\n', '\n')) def get_freetype(self): Settings.DEAD_FUNCTIONS += ['_inflateEnd', '_inflate', '_inflateReset', '_inflateInit2_'] return self.get_library('freetype', os.path.join('objs', '.libs', 'libfreetype.a')) def test_freetype(self): if self.emcc_args is None: return self.skip('requires emcc') if Settings.QUANTUM_SIZE == 1: return self.skip('TODO: Figure out and try to fix') assert 'asm2g' in test_modes if self.run_name == 'asm2g': Settings.ALIASING_FUNCTION_POINTERS = 1 - Settings.ALIASING_FUNCTION_POINTERS # flip for some more coverage here if Settings.CORRECT_SIGNS == 0: Settings.CORRECT_SIGNS = 1 # Not sure why, but needed post = ''' def process(filename): import tools.shared as shared # Embed the font into the document src = open(filename, 'r').read().replace( '// {{PRE_RUN_ADDITIONS}}', "FS.createDataFile('/', 'font.ttf', %s, true, false);" % str( map(ord, open(shared.path_from_root('tests', 'freetype', 'LiberationSansBold.ttf'), 'rb').read()) ) ) open(filename, 'w').write(src) ''' # Not needed for js, but useful for debugging shutil.copyfile(path_from_root('tests', 'freetype', 'LiberationSansBold.ttf'), os.path.join(self.get_dir(), 'font.ttf')) # Main for outlining in [0, 5000]: Settings.OUTLINING_LIMIT = outlining print >> sys.stderr, 'outlining:', outlining self.do_run(open(path_from_root('tests', 'freetype', 'main.c'), 'r').read(), open(path_from_root('tests', 'freetype', 'ref.txt'), 'r').read(), ['font.ttf', 'test!', '150', '120', '25'], libraries=self.get_freetype(), includes=[path_from_root('tests', 'freetype', 'include')], post_build=post) # github issue 324 print '[issue 324]' self.do_run(open(path_from_root('tests', 'freetype', 'main_2.c'), 'r').read(), open(path_from_root('tests', 'freetype', 'ref_2.txt'), 'r').read(), ['font.ttf', 'w', '32', '32', '25'], libraries=self.get_freetype(), includes=[path_from_root('tests', 'freetype', 'include')], post_build=post) print '[issue 324 case 2]' self.do_run(open(path_from_root('tests', 'freetype', 'main_3.c'), 'r').read(), open(path_from_root('tests', 'freetype', 'ref_3.txt'), 'r').read(), ['font.ttf', 'W', '32', '32', '0'], libraries=self.get_freetype(), includes=[path_from_root('tests', 'freetype', 'include')], post_build=post) print '[issue 324 case 3]' self.do_run('', open(path_from_root('tests', 'freetype', 'ref_4.txt'), 'r').read(), ['font.ttf', 'ea', '40', '32', '0'], no_build=True) def test_sqlite(self): # gcc -O3 -I/home/alon/Dev/emscripten/tests/sqlite -ldl src.c if self.emcc_args is None: return self.skip('Very slow without ta2, and we would also need to include dlmalloc manually without emcc') if not self.is_emscripten_abi(): return self.skip('fails on x86 due to a legalization issue on llvm 3.3') if Settings.QUANTUM_SIZE == 1: return self.skip('TODO FIXME') self.banned_js_engines = [NODE_JS] # OOM in older node Settings.CORRECT_SIGNS = 1 Settings.CORRECT_OVERFLOWS = 0 Settings.CORRECT_ROUNDINGS = 0 if self.emcc_args is None: Settings.SAFE_HEAP = 0 # uses time.h to set random bytes, other stuff Settings.DISABLE_EXCEPTION_CATCHING = 1 Settings.FAST_MEMORY = 4*1024*1024 Settings.EXPORTED_FUNCTIONS += ['_sqlite3_open', '_sqlite3_close', '_sqlite3_exec', '_sqlite3_free', '_callback']; if Settings.ASM_JS == 1 and '-g' in self.emcc_args: print "disabling inlining" # without registerize (which -g disables), we generate huge amounts of code Settings.INLINING_LIMIT = 50 #Settings.OUTLINING_LIMIT = 60000 self.do_run(r''' #define SQLITE_DISABLE_LFS #define LONGDOUBLE_TYPE double #define SQLITE_INT64_TYPE long long int #define SQLITE_THREADSAFE 0 ''' + open(path_from_root('tests', 'sqlite', 'sqlite3.c'), 'r').read() + open(path_from_root('tests', 'sqlite', 'benchmark.c'), 'r').read(), open(path_from_root('tests', 'sqlite', 'benchmark.txt'), 'r').read(), includes=[path_from_root('tests', 'sqlite')], force_c=True) def test_zlib(self): if not Settings.USE_TYPED_ARRAYS == 2: return self.skip('works in general, but cached build will be optimized and fail, so disable this') if Settings.ASM_JS: self.banned_js_engines = [NODE_JS] # TODO investigate if self.emcc_args is not None and '-O2' in self.emcc_args and 'ASM_JS=0' not in self.emcc_args: # without asm, closure minifies Math.imul badly self.emcc_args += ['--closure', '1'] # Use closure here for some additional coverage Settings.CORRECT_SIGNS = 1 self.do_run(open(path_from_root('tests', 'zlib', 'example.c'), 'r').read(), open(path_from_root('tests', 'zlib', 'ref.txt'), 'r').read(), libraries=self.get_library('zlib', os.path.join('libz.a'), make_args=['libz.a']), includes=[path_from_root('tests', 'zlib')], force_c=True) def test_the_bullet(self): # Called thus so it runs late in the alphabetical cycle... it is long if self.emcc_args is None: return self.skip('requires emcc') if Building.LLVM_OPTS and self.emcc_args is None: Settings.SAFE_HEAP = 0 # Optimizations make it so we do not have debug info on the line we need to ignore Settings.DEAD_FUNCTIONS = ['__ZSt9terminatev'] # Note: this is also a good test of per-file and per-line changes (since we have multiple files, and correct specific lines) if Settings.SAFE_HEAP: # Ignore bitfield warnings Settings.SAFE_HEAP = 3 Settings.SAFE_HEAP_LINES = ['btVoronoiSimplexSolver.h:40', 'btVoronoiSimplexSolver.h:41', 'btVoronoiSimplexSolver.h:42', 'btVoronoiSimplexSolver.h:43'] asserts = Settings.ASSERTIONS for use_cmake in [False, True]: # If false, use a configure script to configure Bullet build. print 'cmake', use_cmake # Windows cannot run configure sh scripts. if WINDOWS and not use_cmake: continue Settings.ASSERTIONS = 2 if use_cmake else asserts # extra testing for ASSERTIONS == 2 def test(): self.do_run(open(path_from_root('tests', 'bullet', 'Demos', 'HelloWorld', 'HelloWorld.cpp'), 'r').read(), [open(path_from_root('tests', 'bullet', 'output.txt'), 'r').read(), # different roundings open(path_from_root('tests', 'bullet', 'output2.txt'), 'r').read(), open(path_from_root('tests', 'bullet', 'output3.txt'), 'r').read()], libraries=get_bullet_library(self, use_cmake), includes=[path_from_root('tests', 'bullet', 'src')]) test() assert 'asm2g' in test_modes if self.run_name == 'asm2g' and not use_cmake and os.environ.get('EMCC_FAST_COMPILER') == '0': # Test forced alignment print >> sys.stderr, 'testing FORCE_ALIGNED_MEMORY' old = open('src.cpp.o.js').read() Settings.FORCE_ALIGNED_MEMORY = 1 test() new = open('src.cpp.o.js').read() print len(old), len(new), old.count('tempBigInt'), new.count('tempBigInt') assert len(old) > len(new) assert old.count('tempBigInt') > new.count('tempBigInt') def test_poppler(self): if self.emcc_args is None: return self.skip('very slow, we only do this in emcc runs') Settings.CORRECT_OVERFLOWS = 1 Settings.CORRECT_SIGNS = 1 Building.COMPILER_TEST_OPTS += [ '-I' + path_from_root('tests', 'freetype', 'include'), '-I' + path_from_root('tests', 'poppler', 'include'), '-Wno-warn-absolute-paths' ] Settings.INVOKE_RUN = 0 # We append code that does run() ourselves # See post(), below input_file = open(os.path.join(self.get_dir(), 'paper.pdf.js'), 'w') input_file.write(str(map(ord, open(path_from_root('tests', 'poppler', 'paper.pdf'), 'rb').read()))) input_file.close() post = ''' def process(filename): # To avoid loading this large file to memory and altering it, we simply append to the end src = open(filename, 'a') src.write( \'\'\' FS.createDataFile('/', 'paper.pdf', eval(Module.read('paper.pdf.js')), true, false); Module.callMain(Module.arguments); Module.print("Data: " + JSON.stringify(FS.root.contents['filename-1.ppm'].contents.map(function(x) { return unSign(x, 8) }))); \'\'\' ) src.close() ''' #fontconfig = self.get_library('fontconfig', [os.path.join('src', '.libs', 'libfontconfig.a')]) # Used in file, but not needed, mostly freetype = self.get_freetype() poppler = self.get_library('poppler', [os.path.join('utils', 'pdftoppm.o'), os.path.join('utils', 'parseargs.o'), os.path.join('poppler', '.libs', 'libpoppler.a')], env_init={ 'FONTCONFIG_CFLAGS': ' ', 'FONTCONFIG_LIBS': ' ' }, configure_args=['--disable-libjpeg', '--disable-libpng', '--disable-poppler-qt', '--disable-poppler-qt4', '--disable-cms', '--disable-cairo-output', '--disable-abiword-output', '--enable-shared=no']) # Combine libraries combined = os.path.join(self.get_dir(), 'poppler-combined.bc') Building.link(poppler + freetype, combined) self.do_ll_run(combined, map(ord, open(path_from_root('tests', 'poppler', 'ref.ppm'), 'r').read()).__str__().replace(' ', ''), args='-scale-to 512 paper.pdf filename'.split(' '), post_build=post) #, build_ll_hook=self.do_autodebug) def test_openjpeg(self): if self.emcc_args is None: return self.skip('needs libc for getopt') Building.COMPILER_TEST_OPTS = filter(lambda x: x != '-g', Building.COMPILER_TEST_OPTS) # remove -g, so we have one test without it by default if Settings.USE_TYPED_ARRAYS == 2: Settings.CORRECT_SIGNS = 1 else: Settings.CORRECT_SIGNS = 2 Settings.CORRECT_SIGNS_LINES = ["mqc.c:566", "mqc.c:317"] post = ''' def process(filename): import tools.shared as shared original_j2k = shared.path_from_root('tests', 'openjpeg', 'syntensity_lobby_s.j2k') src = open(filename, 'r').read().replace( '// {{PRE_RUN_ADDITIONS}}', "FS.createDataFile('/', 'image.j2k', %s, true, false);" % shared.line_splitter(str( map(ord, open(original_j2k, 'rb').read()) )) ).replace( '// {{POST_RUN_ADDITIONS}}', "Module.print('Data: ' + JSON.stringify(FS.analyzePath('image.raw').object.contents));" ) open(filename, 'w').write(src) ''' shutil.copy(path_from_root('tests', 'openjpeg', 'opj_config.h'), self.get_dir()) lib = self.get_library('openjpeg', [os.path.sep.join('codec/CMakeFiles/j2k_to_image.dir/index.c.o'.split('/')), os.path.sep.join('codec/CMakeFiles/j2k_to_image.dir/convert.c.o'.split('/')), os.path.sep.join('codec/CMakeFiles/j2k_to_image.dir/__/common/color.c.o'.split('/')), os.path.join('bin', 'libopenjpeg.so.1.4.0')], configure=['cmake', '.'], #configure_args=['--enable-tiff=no', '--enable-jp3d=no', '--enable-png=no'], make_args=[]) # no -j 2, since parallel builds can fail # We use doubles in JS, so we get slightly different values than native code. So we # check our output by comparing the average pixel difference def image_compare(output, err): # Get the image generated by JS, from the JSON.stringify'd array m = re.search('\[[\d, -]*\]', output) try: js_data = eval(m.group(0)) except AttributeError: print 'Failed to find proper image output in: ' + output raise js_data = map(lambda x: x if x >= 0 else 256+x, js_data) # Our output may be signed, so unsign it # Get the correct output true_data = open(path_from_root('tests', 'openjpeg', 'syntensity_lobby_s.raw'), 'rb').read() # Compare them assert(len(js_data) == len(true_data)) num = len(js_data) diff_total = js_total = true_total = 0 for i in range(num): js_total += js_data[i] true_total += ord(true_data[i]) diff_total += abs(js_data[i] - ord(true_data[i])) js_mean = js_total/float(num) true_mean = true_total/float(num) diff_mean = diff_total/float(num) image_mean = 83.265 #print '[image stats:', js_mean, image_mean, true_mean, diff_mean, num, ']' assert abs(js_mean - image_mean) < 0.01 assert abs(true_mean - image_mean) < 0.01 assert diff_mean < 0.01 return output self.emcc_args += ['--minify', '0'] # to compare the versions def do_test(): self.do_run(open(path_from_root('tests', 'openjpeg', 'codec', 'j2k_to_image.c'), 'r').read(), 'Successfully generated', # The real test for valid output is in image_compare '-i image.j2k -o image.raw'.split(' '), libraries=lib, includes=[path_from_root('tests', 'openjpeg', 'libopenjpeg'), path_from_root('tests', 'openjpeg', 'codec'), path_from_root('tests', 'openjpeg', 'common'), os.path.join(self.get_build_dir(), 'openjpeg')], force_c=True, post_build=post, output_nicerizer=image_compare)#, build_ll_hook=self.do_autodebug) do_test() # some test coverage for EMCC_DEBUG 1 and 2 if self.emcc_args and '-O2' in self.emcc_args and 'EMCC_DEBUG' not in os.environ and '-g' in self.emcc_args: shutil.copyfile('src.c.o.js', 'release.js') try: os.environ['EMCC_DEBUG'] = '1' print '2' do_test() shutil.copyfile('src.c.o.js', 'debug1.js') os.environ['EMCC_DEBUG'] = '2' print '3' do_test() shutil.copyfile('src.c.o.js', 'debug2.js') finally: del os.environ['EMCC_DEBUG'] for debug in [1,2]: def clean(text): text = text.replace('\n\n', '\n').replace('\n\n', '\n').replace('\n\n', '\n').replace('\n\n', '\n').replace('\n\n', '\n').replace('{\n}', '{}') return '\n'.join(sorted(text.split('\n'))) sizes = len(open('release.js').read()), len(open('debug%d.js' % debug).read()) print >> sys.stderr, debug, 'sizes', sizes assert abs(sizes[0] - sizes[1]) < 0.001*sizes[0], sizes # we can't check on identical output, compilation is not 100% deterministic (order of switch elements, etc.), but size should be ~identical print >> sys.stderr, 'debug check %d passed too' % debug try: os.environ['EMCC_FORCE_STDLIBS'] = '1' print 'EMCC_FORCE_STDLIBS' do_test() finally: del os.environ['EMCC_FORCE_STDLIBS'] print >> sys.stderr, 'EMCC_FORCE_STDLIBS ok' try_delete(CANONICAL_TEMP_DIR) else: print >> sys.stderr, 'not doing debug check' def test_python(self): if self.emcc_args is None: return self.skip('requires emcc') if Settings.QUANTUM_SIZE == 1: return self.skip('TODO: make this work') if not self.is_emscripten_abi(): return self.skip('fails on not asmjs-unknown-emscripten') # FIXME #Settings.EXPORTED_FUNCTIONS += ['_PyRun_SimpleStringFlags'] # for the demo if self.is_emscripten_abi(): bitcode = path_from_root('tests', 'python', 'python.asmjs-unknown-emscripten.bc') else: bitcode = path_from_root('tests', 'python', 'python.small.bc') for lto in [0, 1]: if lto == 1: self.emcc_args += ['--llvm-lto', '1'] print self.emcc_args self.do_ll_run(bitcode, 'hello python world!\n[0, 2, 4, 6]\n5\n22\n5.470000', args=['-S', '-c' '''print "hello python world!"; print [x*2 for x in range(4)]; t=2; print 10-3-t; print (lambda x: x*2)(11); print '%f' % 5.47''']) def test_lifetime(self): if self.emcc_args is None: return self.skip('test relies on emcc opts') self.do_ll_run(path_from_root('tests', 'lifetime.ll'), 'hello, world!\n') if '-O1' in self.emcc_args or '-O2' in self.emcc_args: assert 'a18' not in open(os.path.join(self.get_dir(), 'src.cpp.o.js')).read(), 'lifetime stuff and their vars must be culled' # Test cases in separate files. Note that these files may contain invalid .ll! # They are only valid enough for us to read for test purposes, not for llvm-as # to process. def test_cases(self): if Building.LLVM_OPTS: return self.skip("Our code is not exactly 'normal' llvm assembly") emcc_args = self.emcc_args try: os.environ['EMCC_LEAVE_INPUTS_RAW'] = '1' Settings.CHECK_OVERFLOWS = 0 for name in glob.glob(path_from_root('tests', 'cases', '*.ll')): shortname = name.replace('.ll', '') if '' not in shortname: continue if os.environ.get('EMCC_FAST_COMPILER') != '0' and os.path.basename(shortname) in [ 'aliasbitcast', 'structparam', 'extendedprecision', 'issue_39', 'phinonexist', 'oob_ta2', 'phiself', 'invokebitcast', # invalid ir 'structphiparam', 'callwithstructural_ta2', 'callwithstructural64_ta2', 'structinparam', # pnacl limitations in ExpandStructRegs '2xi40', # pnacl limitations in ExpandGetElementPtr 'quoted', # current fastcomp limitations FIXME 'atomicrmw_unaligned', # TODO XXX 'emptyasm_aue' # we don't support inline asm ]: continue if '_ta2' in shortname and not Settings.USE_TYPED_ARRAYS == 2: print self.skip('case "%s" only relevant for ta2' % shortname) continue if '_noasm' in shortname and Settings.ASM_JS: print self.skip('case "%s" not relevant for asm.js' % shortname) continue if '_eua' in shortname and not self.is_emscripten_abi(): print self.skip('case "%s" not relevant for not asmjs-unknown-emscripten target' % shortname) continue if '_fastcomp' in shortname and not os.environ.get('EMCC_FAST_COMPILER') != '0': print self.skip('case "%s" not relevant for non-fastcomp' % shortname) continue self.emcc_args = emcc_args if os.path.exists(shortname + '.emcc'): if not self.emcc_args: continue self.emcc_args = self.emcc_args + json.loads(open(shortname + '.emcc').read()) print >> sys.stderr, "Testing case '%s'..." % shortname output_file = path_from_root('tests', 'cases', shortname + '.txt') if Settings.QUANTUM_SIZE == 1: q1_output_file = path_from_root('tests', 'cases', shortname + '_q1.txt') if os.path.exists(q1_output_file): output_file = q1_output_file if os.path.exists(output_file): output = open(output_file, 'r').read() else: output = 'hello, world!' if output.rstrip() != 'skip': self.do_ll_run(path_from_root('tests', 'cases', name), output) # Optional source checking, a python script that gets a global generated with the source src_checker = path_from_root('tests', 'cases', shortname + '.py') if os.path.exists(src_checker): generated = open('src.cpp.o.js').read() exec(open(src_checker).read()) finally: del os.environ['EMCC_LEAVE_INPUTS_RAW'] self.emcc_args = emcc_args def test_fuzz(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip('needs ta2') Building.COMPILER_TEST_OPTS += ['-I' + path_from_root('tests', 'fuzz'), '-Wno-warn-absolute-paths'] def run_all(x): print x for name in glob.glob(path_from_root('tests', 'fuzz', '*.c')) + glob.glob(path_from_root('tests', 'fuzz', '*.cpp')): #if os.path.basename(name) != '4.c': continue if 'newfail' in name: continue if os.environ.get('EMCC_FAST_COMPILER') == '0' and os.path.basename(name) in [ '18.cpp', '15.c', '21.c' ]: continue # works only in fastcomp if x == 'lto' and self.run_name == 'default' and os.path.basename(name) in [ '19.c' ]: continue # LLVM LTO bug print name self.do_run(open(path_from_root('tests', 'fuzz', name)).read(), open(path_from_root('tests', 'fuzz', name + '.txt')).read(), force_c=name.endswith('.c')) run_all('normal') self.emcc_args += ['--llvm-lto', '1'] run_all('lto') # Autodebug the code def do_autodebug(self, filename): output = Popen([PYTHON, AUTODEBUGGER, filename+'.o.ll', filename+'.o.ll.ll'], stdout=PIPE, stderr=self.stderr_redirect).communicate()[0] assert 'Success.' in output, output self.prep_ll_run(filename, filename+'.o.ll.ll', force_recompile=True) # rebuild .bc # TODO: use code in do_autodebug_post for this # Autodebug the code, after LLVM opts. Will only work once! def do_autodebug_post(self, filename): if not hasattr(self, 'post'): print 'Asking for post re-call' self.post = True return True print 'Autodebugging during post time' delattr(self, 'post') output = Popen([PYTHON, AUTODEBUGGER, filename+'.o.ll', filename+'.o.ll.ll'], stdout=PIPE, stderr=self.stderr_redirect).communicate()[0] assert 'Success.' in output, output shutil.copyfile(filename + '.o.ll.ll', filename + '.o.ll') Building.llvm_as(filename) Building.llvm_dis(filename) def test_autodebug(self): if Building.LLVM_OPTS: return self.skip('LLVM opts mess us up') Building.COMPILER_TEST_OPTS += ['--llvm-opts', '0'] # Run a test that should work, generating some code self.test_structs() filename = os.path.join(self.get_dir(), 'src.cpp') self.do_autodebug(filename) # Compare to each other, and to expected output self.do_ll_run(path_from_root('tests', filename+'.o.ll.ll'), '''AD:-1,1''') assert open('stdout').read().startswith('AD:-1'), 'We must note when we enter functions' # Test using build_ll_hook src = ''' #include char cache[256], *next = cache; int main() { cache[10] = 25; next[20] = 51; int x = cache[10]; double y = 11.52; printf("*%d,%d,%.2f*\\n", x, cache[20], y); return 0; } ''' self.do_run(src, '''AD:-1,1''', build_ll_hook=self.do_autodebug) def test_corruption(self): if Settings.ASM_JS: return self.skip('cannot use corruption checks in asm') if Settings.USE_TYPED_ARRAYS != 2: return self.skip('needs ta2 for actual test') Settings.CORRUPTION_CHECK = 1 src = r''' #include #include #include int main(int argc, char **argv) { int size = 1024*argc; char *buffer = (char*)malloc(size); #if CORRUPT memset(buffer, argc, size+15); #else memset(buffer, argc, size); #endif for (int x = 0; x < size; x += argc*3) buffer[x] = x/3; int ret = 0; for (int x = 0; x < size; x++) ret += buffer[x]; free(buffer); printf("All ok, %d\n", ret); } ''' for corrupt in [1]: self.do_run(src.replace('CORRUPT', str(corrupt)), 'Heap corruption detected!' if corrupt else 'All ok, 4209') def test_corruption_2(self): if Settings.ASM_JS: return self.skip('cannot use corruption checks in asm') if Settings.USE_TYPED_ARRAYS != 2: return self.skip('needs ta2 for actual test') Settings.SAFE_HEAP = 1 Settings.CORRUPTION_CHECK = 1 # test for free(0), malloc(0), etc. test_path = path_from_root('tests', 'core', 'test_corruption_2') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_corruption_3(self): if Settings.ASM_JS: return self.skip('cannot use corruption checks in asm') if Settings.USE_TYPED_ARRAYS != 2: return self.skip('needs ta2 for actual test') Settings.CORRUPTION_CHECK = 1 # realloc test_path = path_from_root('tests', 'core', 'test_corruption_3') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) ### Integration tests def test_ccall(self): if self.emcc_args is not None and '-O2' in self.emcc_args: self.emcc_args += ['--closure', '1'] # Use closure here, to test we export things right post = ''' def process(filename): src = \'\'\' var Module = { 'noInitialRun': true }; \'\'\' + open(filename, 'r').read() + \'\'\' Module.addOnExit(function () { Module.print('*'); var ret; ret = Module['ccall']('get_int', 'number'); Module.print([typeof ret, ret]); ret = ccall('get_float', 'number'); Module.print([typeof ret, ret.toFixed(2)]); ret = ccall('get_string', 'string'); Module.print([typeof ret, ret]); ret = ccall('print_int', null, ['number'], [12]); Module.print(typeof ret); ret = ccall('print_float', null, ['number'], [14.56]); Module.print(typeof ret); ret = ccall('print_string', null, ['string'], ["cheez"]); Module.print(typeof ret); ret = ccall('print_string', null, ['array'], [[97, 114, 114, 45, 97, 121, 0]]); Module.print(typeof ret); ret = ccall('multi', 'number', ['number', 'number', 'number', 'string'], [2, 1.4, 3, 'more']); Module.print([typeof ret, ret]); var p = ccall('malloc', 'pointer', ['number'], [4]); setValue(p, 650, 'i32'); ret = ccall('pointer', 'pointer', ['pointer'], [p]); Module.print([typeof ret, getValue(ret, 'i32')]); Module.print('*'); // part 2: cwrap var multi = Module['cwrap']('multi', 'number', ['number', 'number', 'number', 'string']); Module.print(multi(2, 1.4, 3, 'atr')); Module.print(multi(8, 5.4, 4, 'bret')); Module.print('*'); // part 3: avoid stack explosion for (var i = 0; i < TOTAL_STACK/60; i++) { ccall('multi', 'number', ['number', 'number', 'number', 'string'], [0, 0, 0, '123456789012345678901234567890123456789012345678901234567890']); } Module.print('stack is ok.'); }); Module.callMain(); \'\'\' open(filename, 'w').write(src) ''' Settings.EXPORTED_FUNCTIONS += ['_get_int', '_get_float', '_get_string', '_print_int', '_print_float', '_print_string', '_multi', '_pointer', '_malloc'] test_path = path_from_root('tests', 'core', 'test_ccall') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output, post_build=post) def test_pgo(self): if Settings.ASM_JS: return self.skip('PGO does not work in asm mode') def run_all(name, src): print name def test(expected, args=[], no_build=False): self.do_run(src, expected, args=args, no_build=no_build) return open(self.in_dir('src.cpp.o.js')).read() # Sanity check that it works and the dead function is emitted js = test('*9*') assert 'function _unused(' in js # Run with PGO, see that unused is true to its name Settings.PGO = 1 test("*9*\n-s DEAD_FUNCTIONS='[\"_unused\"]'") Settings.PGO = 0 # Kill off the dead function, still works and it is not emitted Settings.DEAD_FUNCTIONS = ['_unused'] js = test('*9*') assert 'function _unused($' not in js # no compiled code assert 'function _unused(' in js # lib-generated stub Settings.DEAD_FUNCTIONS = [] # Run the same code with argc that uses the dead function, see abort test(('dead function: unused'), args=['a', 'b'], no_build=True) # Normal stuff run_all('normal', r''' #include extern "C" { int used(int x) { if (x == 0) return -1; return used(x/3) + used(x/17) + x%5; } int unused(int x) { if (x == 0) return -1; return unused(x/4) + unused(x/23) + x%7; } } int main(int argc, char **argv) { printf("*%d*\n", argc == 3 ? unused(argv[0][0] + 1024) : used(argc + 1555)); return 0; } ''') # Call by function pointer run_all('function pointers', r''' #include extern "C" { int used(int x) { if (x == 0) return -1; return used(x/3) + used(x/17) + x%5; } int unused(int x) { if (x == 0) return -1; return unused(x/4) + unused(x/23) + x%7; } } typedef int (*ii)(int); int main(int argc, char **argv) { ii pointers[256]; for (int i = 0; i < 256; i++) { pointers[i] = (i == 3) ? unused : used; } printf("*%d*\n", pointers[argc](argc + 1555)); return 0; } ''') def test_asm_pgo(self): if not Settings.ASM_JS: return self.skip('this is a test for PGO for asm (NB: not *in* asm)') if os.environ.get('EMCC_FAST_COMPILER') != '0': return self.skip('todo in fastcomp') src = open(path_from_root('tests', 'hello_libcxx.cpp')).read() output = 'hello, world!' self.do_run(src, output) shutil.move(self.in_dir('src.cpp.o.js'), self.in_dir('normal.js')) Settings.ASM_JS = 0 Settings.PGO = 1 self.do_run(src, output) Settings.ASM_JS = 1 Settings.PGO = 0 shutil.move(self.in_dir('src.cpp.o.js'), self.in_dir('pgo.js')) pgo_output = run_js(self.in_dir('pgo.js')).split('\n')[1] open('pgo_data.rsp', 'w').write(pgo_output) # with response file self.emcc_args += ['@pgo_data.rsp'] self.do_run(src, output) self.emcc_args.pop() shutil.move(self.in_dir('src.cpp.o.js'), self.in_dir('pgoed.js')) before = len(open('normal.js').read()) after = len(open('pgoed.js').read()) assert after < 0.90 * before, [before, after] # expect a size reduction # with response in settings element itself open('dead_funcs', 'w').write(pgo_output[pgo_output.find('['):-1]) self.emcc_args += ['-s', 'DEAD_FUNCTIONS=@' + self.in_dir('dead_funcs')] self.do_run(src, output) self.emcc_args.pop() self.emcc_args.pop() shutil.move(self.in_dir('src.cpp.o.js'), self.in_dir('pgoed2.js')) assert open('pgoed.js').read() == open('pgoed2.js').read() # with relative response in settings element itself open('dead_funcs', 'w').write(pgo_output[pgo_output.find('['):-1]) self.emcc_args += ['-s', 'DEAD_FUNCTIONS=@dead_funcs'] self.do_run(src, output) self.emcc_args.pop() self.emcc_args.pop() shutil.move(self.in_dir('src.cpp.o.js'), self.in_dir('pgoed2.js')) assert open('pgoed.js').read() == open('pgoed2.js').read() def test_exported_response(self): if self.emcc_args is None: return self.skip('requires emcc') src = r''' #include #include extern "C" { int other_function() { return 5; } } int main() { printf("waka!\n"); return 0; } ''' open('exps', 'w').write('["_main","_other_function"]') self.emcc_args += ['-s', 'EXPORTED_FUNCTIONS=@exps'] self.do_run(src, '''waka!''') assert 'other_function' in open('src.cpp.o.js').read() def test_add_function(self): if self.emcc_args is None: return self.skip('requires emcc') Settings.INVOKE_RUN = 0 Settings.RESERVED_FUNCTION_POINTERS = 1 src = r''' #include #include int main(int argc, char **argv) { int fp = atoi(argv[1]); printf("fp: %d\n", fp); void (*f)(int) = reinterpret_cast(fp); f(7); return 0; } ''' open(os.path.join(self.get_dir(), 'post.js'), 'w').write(''' var newFuncPtr = Runtime.addFunction(function(num) { Module.print('Hello ' + num + ' from JS!'); }); Module.callMain([newFuncPtr.toString()]); ''') self.emcc_args += ['--post-js', 'post.js'] self.do_run(src, '''Hello 7 from JS!''') if Settings.ASM_JS: Settings.RESERVED_FUNCTION_POINTERS = 0 self.do_run(src, '''Finished up all reserved function pointers. Use a higher value for RESERVED_FUNCTION_POINTERS.''') generated = open('src.cpp.o.js').read() assert 'jsCall' not in generated Settings.RESERVED_FUNCTION_POINTERS = 1 Settings.ALIASING_FUNCTION_POINTERS = 1 - Settings.ALIASING_FUNCTION_POINTERS # flip the test self.do_run(src, '''Hello 7 from JS!''') def test_demangle_stacks(self): if Settings.ASM_JS: return self.skip('spidermonkey has stack trace issues') test_path = path_from_root('tests', 'core', 'test_demangle_stacks') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_embind(self): if self.emcc_args is None: return self.skip('requires emcc') Building.COMPILER_TEST_OPTS += ['--bind'] src = r''' #include #include using namespace emscripten; int main() { val Math = val::global("Math"); // two ways to call Math.abs printf("abs(-10): %d\n", Math.call("abs", -10)); printf("abs(-11): %d\n", Math["abs"](-11).as()); return 0; } ''' self.do_run(src, 'abs(-10): 10\nabs(-11): 11'); def test_embind_2(self): if self.emcc_args is None: return self.skip('requires emcc') if self.run_name == 'slow2asm': return self.skip('embind/asm.js requires fastcomp') Building.COMPILER_TEST_OPTS += ['--bind', '--post-js', 'post.js'] open('post.js', 'w').write(''' Module.print('lerp ' + Module.lerp(1, 2, 0.66) + '.'); ''') src = r''' #include #include #include using namespace emscripten; float lerp(float a, float b, float t) { return (1 - t) * a + t * b; } EMSCRIPTEN_BINDINGS(my_module) { function("lerp", &lerp); } ''' self.do_run(src, 'lerp 1.66'); def test_scriptaclass(self): if self.emcc_args is None: return self.skip('requires emcc') Settings.EXPORT_BINDINGS = 1 header_filename = os.path.join(self.get_dir(), 'header.h') header = ''' struct ScriptMe { int value; ScriptMe(int val); int getVal(); // XXX Sadly, inlining these will result in LLVM not // producing any code for them (when just building // as a library) void mulVal(int mul); }; ''' h = open(header_filename, 'w') h.write(header) h.close() src = ''' #include "header.h" ScriptMe::ScriptMe(int val) : value(val) { } int ScriptMe::getVal() { return value; } void ScriptMe::mulVal(int mul) { value *= mul; } ''' # Way 1: use demangler and namespacer script_src = ''' var sme = Module._.ScriptMe.__new__(83); // malloc(sizeof(ScriptMe)), ScriptMe::ScriptMe(sme, 83) / new ScriptMe(83) (at addr sme) Module._.ScriptMe.mulVal(sme, 2); // ScriptMe::mulVal(sme, 2) sme.mulVal(2) Module.print('*' + Module._.ScriptMe.getVal(sme) + '*'); _free(sme); Module.print('*ok*'); ''' post = ''' def process(filename): Popen([PYTHON, DEMANGLER, filename], stdout=open(filename + '.tmp', 'w')).communicate() Popen([PYTHON, NAMESPACER, filename, filename + '.tmp'], stdout=open(filename + '.tmp2', 'w')).communicate() src = open(filename, 'r').read().replace( '// {{MODULE_ADDITIONS}', 'Module["_"] = ' + open(filename + '.tmp2', 'r').read().replace('var ModuleNames = ', '').rstrip() + ';\n\n' + script_src + '\n\n' + '// {{MODULE_ADDITIONS}' ) open(filename, 'w').write(src) ''' # XXX disable due to possible v8 bug -- self.do_run(src, '*166*\n*ok*', post_build=post) if self.emcc_args is not None and '-O2' in self.emcc_args and 'ASM_JS=0' not in self.emcc_args: # without asm, closure minifies Math.imul badly self.emcc_args += ['--closure', '1'] # Use closure here, to test we export things right # Way 2: use CppHeaderParser header = ''' #include class Parent { protected: int value; public: Parent(int val); Parent(Parent *p, Parent *q); // overload constructor int getVal() { return value; }; // inline should work just fine here, unlike Way 1 before void mulVal(int mul); }; class Child1 : public Parent { public: Child1() : Parent(7) { printf("Child1:%d\\n", value); }; Child1(int val) : Parent(val*2) { value -= 1; printf("Child1:%d\\n", value); }; int getValSqr() { return value*value; } int getValSqr(int more) { return value*value*more; } int getValTimes(int times=1) { return value*times; } }; class Child2 : public Parent { public: Child2() : Parent(9) { printf("Child2:%d\\n", value); }; int getValCube() { return value*value*value; } static void printStatic() { printf("*static*\\n"); } virtual void virtualFunc() { printf("*virtualf*\\n"); } virtual void virtualFunc2() { printf("*virtualf2*\\n"); } static void runVirtualFunc(Child2 *self) { self->virtualFunc(); }; private: void doSomethingSecret() { printf("security breached!\\n"); }; // we should not be able to do this }; ''' open(header_filename, 'w').write(header) basename = os.path.join(self.get_dir(), 'bindingtest') output = Popen([PYTHON, BINDINGS_GENERATOR, basename, header_filename], stdout=PIPE, stderr=self.stderr_redirect).communicate()[0] #print output assert 'Traceback' not in output, 'Failure in binding generation: ' + output src = ''' #include "header.h" Parent::Parent(int val) : value(val) { printf("Parent:%d\\n", val); } Parent::Parent(Parent *p, Parent *q) : value(p->value + q->value) { printf("Parent:%d\\n", value); } void Parent::mulVal(int mul) { value *= mul; } #include "bindingtest.cpp" ''' post2 = ''' def process(filename): src = open(filename, 'a') src.write(open('bindingtest.js').read() + '\\n\\n') src.close() ''' def post3(filename): script_src_2 = ''' var sme = new Module.Parent(42); sme.mulVal(2); Module.print('*') Module.print(sme.getVal()); Module.print('c1'); var c1 = new Module.Child1(); Module.print(c1.getVal()); c1.mulVal(2); Module.print(c1.getVal()); Module.print(c1.getValSqr()); Module.print(c1.getValSqr(3)); Module.print(c1.getValTimes()); // default argument should be 1 Module.print(c1.getValTimes(2)); Module.print('c1 v2'); c1 = new Module.Child1(8); // now with a parameter, we should handle the overloading automatically and properly and use constructor #2 Module.print(c1.getVal()); c1.mulVal(2); Module.print(c1.getVal()); Module.print(c1.getValSqr()); Module.print(c1.getValSqr(3)); Module.print('c2') var c2 = new Module.Child2(); Module.print(c2.getVal()); c2.mulVal(2); Module.print(c2.getVal()); Module.print(c2.getValCube()); var succeeded; try { succeeded = 0; Module.print(c2.doSomethingSecret()); // should fail since private succeeded = 1; } catch(e) {} Module.print(succeeded); try { succeeded = 0; Module.print(c2.getValSqr()); // function from the other class succeeded = 1; } catch(e) {} Module.print(succeeded); try { succeeded = 0; c2.getValCube(); // sanity succeeded = 1; } catch(e) {} Module.print(succeeded); Module.Child2.prototype.printStatic(); // static calls go through the prototype // virtual function c2.virtualFunc(); Module.Child2.prototype.runVirtualFunc(c2); c2.virtualFunc2(); // extend the class from JS var c3 = new Module.Child2; Module.customizeVTable(c3, [{ original: Module.Child2.prototype.virtualFunc, replacement: function() { Module.print('*js virtualf replacement*'); } }, { original: Module.Child2.prototype.virtualFunc2, replacement: function() { Module.print('*js virtualf2 replacement*'); } }]); c3.virtualFunc(); Module.Child2.prototype.runVirtualFunc(c3); c3.virtualFunc2(); c2.virtualFunc(); // original should remain the same Module.Child2.prototype.runVirtualFunc(c2); c2.virtualFunc2(); Module.print('*ok*'); ''' code = open(filename).read() src = open(filename, 'w') src.write('var Module = {};\n') # name Module src.write(code) src.write(script_src_2 + '\n') src.close() Settings.RESERVED_FUNCTION_POINTERS = 20 self.do_run(src, '''* 84 c1 Parent:7 Child1:7 7 14 196 588 14 28 c1 v2 Parent:16 Child1:15 15 30 900 2700 c2 Parent:9 Child2:9 9 18 5832 0 0 1 *static* *virtualf* *virtualf* *virtualf2*''' + (''' Parent:9 Child2:9 *js virtualf replacement* *js virtualf replacement* *js virtualf2 replacement* *virtualf* *virtualf* *virtualf2*''') + ''' *ok* ''', post_build=(post2, post3)) def test_scriptaclass_2(self): if self.emcc_args is None: return self.skip('requires emcc') Settings.EXPORT_BINDINGS = 1 header_filename = os.path.join(self.get_dir(), 'header.h') header = ''' #include #include class StringUser { char *s; int i; public: StringUser(char *string, int integer) : s(strdup(string)), i(integer) {} void Print(int anotherInteger, char *anotherString) { printf("|%s|%d|%s|%d|\\n", s, i, anotherString, anotherInteger); } void CallOther(StringUser *fr) { fr->Print(i, s); } }; ''' open(header_filename, 'w').write(header) basename = os.path.join(self.get_dir(), 'bindingtest') output = Popen([PYTHON, BINDINGS_GENERATOR, basename, header_filename], stdout=PIPE, stderr=self.stderr_redirect).communicate()[0] #print output assert 'Traceback' not in output, 'Failure in binding generation: ' + output src = ''' #include "header.h" #include "bindingtest.cpp" ''' post = ''' def process(filename): src = open(filename, 'a') src.write(open('bindingtest.js').read() + '\\n\\n') src.write(\'\'\' var user = new Module.StringUser("hello", 43); user.Print(41, "world"); \'\'\') src.close() ''' self.do_run(src, '|hello|43|world|41|', post_build=post) def test_webidl(self): if self.emcc_args is None: return self.skip('requires emcc') output = Popen([PYTHON, path_from_root('tools', 'webidl_binder.py'), path_from_root('tests', 'webidl', 'test.idl'), 'glue']).communicate()[0] assert os.path.exists('glue.cpp') assert os.path.exists('glue.js') self.emcc_args += ['--post-js', 'glue.js', '--post-js', path_from_root('tests', 'webidl', 'post.js')] shutil.copyfile(path_from_root('tests', 'webidl', 'test.h'), self.in_dir('test.h')) shutil.copyfile(path_from_root('tests', 'webidl', 'test.cpp'), self.in_dir('test.cpp')) src = open('test.cpp').read() self.do_run(src, open(path_from_root('tests', 'webidl', 'output.txt')).read()) def test_typeinfo(self): if os.environ.get('EMCC_FAST_COMPILER') != '0': return self.skip('fastcomp does not support RUNTIME_TYPE_INFO') if self.emcc_args is not None and self.emcc_args != []: return self.skip('full LLVM opts optimize out all the code that uses the type') Settings.RUNTIME_TYPE_INFO = 1 if Settings.QUANTUM_SIZE != 4: return self.skip('We assume normal sizes in the output here') src = ''' #include struct UserStruct { int x; char y; short z; }; struct Encloser { short x; UserStruct us; int y; }; int main() { Encloser e; e.us.y = 5; printf("*ok:%d*\\n", e.us.y); return 0; } ''' post = ''' def process(filename): src = open(filename, 'r').read().replace( '// {{POST_RUN_ADDITIONS}}', \'\'\' if (Runtime.typeInfo) { Module.print('|' + Runtime.typeInfo.UserStruct.fields + '|' + Runtime.typeInfo.UserStruct.flatIndexes + '|'); var t = Runtime.generateStructInfo(['x', { us: ['x', 'y', 'z'] }, 'y'], 'Encloser') Module.print('|' + [t.x, t.us.x, t.us.y, t.us.z, t.y] + '|'); Module.print('|' + JSON.stringify(Runtime.generateStructInfo(['x', 'y', 'z'], 'UserStruct')) + '|'); } else { Module.print('No type info.'); } \'\'\' ) open(filename, 'w').write(src) ''' self.do_run(src, '*ok:5*\n|i32,i8,i16|0,4,6|\n|0,4,8,10,12|\n|{"__size__":8,"x":0,"y":4,"z":6}|', post_build=post) # Make sure that without the setting, we don't spam the .js with the type info Settings.RUNTIME_TYPE_INFO = 0 self.do_run(src, 'No type info.', post_build=post) ### Tests for tools def test_safe_heap(self): if not Settings.SAFE_HEAP: return self.skip('We need SAFE_HEAP to test SAFE_HEAP') if Settings.USE_TYPED_ARRAYS == 2: return self.skip('It is ok to violate the load-store assumption with TA2') if Building.LLVM_OPTS: return self.skip('LLVM can optimize away the intermediate |x|') src = ''' #include #include int main() { int *x = (int*)malloc(sizeof(int)); *x = 20; float *y = (float*)x; printf("%f\\n", *y); printf("*ok*\\n"); return 0; } ''' try: self.do_run(src, '*nothingatall*') except Exception, e: # This test *should* fail, by throwing this exception assert 'Assertion failed: Load-store consistency assumption failure!' in str(e), str(e) # And we should not fail if we disable checking on that line Settings.SAFE_HEAP = 3 Settings.SAFE_HEAP_LINES = ["src.cpp:7"] self.do_run(src, '*ok*') # But if we disable the wrong lines, we still fail Settings.SAFE_HEAP_LINES = ["src.cpp:99"] try: self.do_run(src, '*nothingatall*') except Exception, e: # This test *should* fail, by throwing this exception assert 'Assertion failed: Load-store consistency assumption failure!' in str(e), str(e) # And reverse the checks with = 2 Settings.SAFE_HEAP = 2 Settings.SAFE_HEAP_LINES = ["src.cpp:99"] self.do_run(src, '*ok*') Settings.SAFE_HEAP = 1 # Linking multiple files should work too module = ''' #include #include void callFunc() { int *x = (int*)malloc(sizeof(int)); *x = 20; float *y = (float*)x; printf("%f\\n", *y); } ''' module_name = os.path.join(self.get_dir(), 'module.cpp') open(module_name, 'w').write(module) main = ''' #include #include extern void callFunc(); int main() { callFunc(); int *x = (int*)malloc(sizeof(int)); *x = 20; float *y = (float*)x; printf("%f\\n", *y); printf("*ok*\\n"); return 0; } ''' main_name = os.path.join(self.get_dir(), 'main.cpp') open(main_name, 'w').write(main) Building.emcc(module_name, ['-g']) Building.emcc(main_name, ['-g']) all_name = os.path.join(self.get_dir(), 'all.bc') Building.link([module_name + '.o', main_name + '.o'], all_name) try: self.do_ll_run(all_name, '*nothingatall*') except Exception, e: # This test *should* fail, by throwing this exception assert 'Assertion failed: Load-store consistency assumption failure!' in str(e), str(e) # And we should not fail if we disable checking on those lines Settings.SAFE_HEAP = 3 Settings.SAFE_HEAP_LINES = ["module.cpp:7", "main.cpp:9"] self.do_ll_run(all_name, '*ok*') # But we will fail if we do not disable exactly what we need to - any mistake leads to error for lines in [["module.cpp:22", "main.cpp:9"], ["module.cpp:7", "main.cpp:29"], ["module.cpp:127", "main.cpp:449"], ["module.cpp:7"], ["main.cpp:9"]]: Settings.SAFE_HEAP_LINES = lines try: self.do_ll_run(all_name, '*nothingatall*') except Exception, e: # This test *should* fail, by throwing this exception assert 'Assertion failed: Load-store consistency assumption failure!' in str(e), str(e) def test_debug(self): if '-g' not in Building.COMPILER_TEST_OPTS: Building.COMPILER_TEST_OPTS.append('-g') if self.emcc_args is not None: if '-O1' in self.emcc_args or '-O2' in self.emcc_args or '-O3' in self.emcc_args: return self.skip('optimizations remove LLVM debug info') src = ''' #include #include void checker(int x) { x += 20; assert(x < 15); // this is line 7! } int main() { checker(10); return 0; } ''' try: self.do_run(src, '*nothingatall*') except Exception, e: # This test *should* fail assert 'Assertion failed: x < 15' in str(e), str(e) lines = open('src.cpp.o.js', 'r').readlines() lines = filter(lambda line: '___assert_fail(' in line or '___assert_func(' in line, lines) found_line_num = any(('//@line 7 "' in line) for line in lines) found_filename = any(('src.cpp"\n' in line) for line in lines) assert found_line_num, 'Must have debug info with the line number' assert found_filename, 'Must have debug info with the filename' def test_source_map(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip("doesn't pass without typed arrays") if NODE_JS not in JS_ENGINES: return self.skip('sourcemapper requires Node to run') if '-g' not in Building.COMPILER_TEST_OPTS: Building.COMPILER_TEST_OPTS.append('-g') src = ''' #include #include __attribute__((noinline)) int foo() { printf("hi"); // line 6 return 1; // line 7 } int main() { printf("%d", foo()); // line 11 return 0; // line 12 } ''' dirname = self.get_dir() src_filename = os.path.join(dirname, 'src.cpp') out_filename = os.path.join(dirname, 'a.out.js') no_maps_filename = os.path.join(dirname, 'no-maps.out.js') with open(src_filename, 'w') as f: f.write(src) assert '-g4' not in Building.COMPILER_TEST_OPTS Building.emcc(src_filename, Settings.serialize() + self.emcc_args + Building.COMPILER_TEST_OPTS, out_filename) # the file name may find its way into the generated code, so make sure we # can do an apples-to-apples comparison by compiling with the same file name shutil.move(out_filename, no_maps_filename) with open(no_maps_filename) as f: no_maps_file = f.read() no_maps_file = re.sub(' *//[@#].*$', '', no_maps_file, flags=re.MULTILINE) Building.COMPILER_TEST_OPTS.append('-g4') def build_and_check(): import json Building.emcc(src_filename, Settings.serialize() + self.emcc_args + Building.COMPILER_TEST_OPTS, out_filename, stderr=PIPE) with open(out_filename) as f: out_file = f.read() # after removing the @line and @sourceMappingURL comments, the build # result should be identical to the non-source-mapped debug version. # this is worth checking because the parser AST swaps strings for token # objects when generating source maps, so we want to make sure the # optimizer can deal with both types. out_file = re.sub(' *//[@#].*$', '', out_file, flags=re.MULTILINE) def clean(code): code = re.sub(r'\n+[ \n]*\n+', '\n', code) code = code.replace('{\n}', '{}') return '\n'.join(sorted(code.split('\n'))) self.assertIdentical(clean(no_maps_file), clean(out_file)) map_filename = out_filename + '.map' data = json.load(open(map_filename, 'r')) self.assertPathsIdentical(out_filename, data['file']) self.assertPathsIdentical(src_filename, data['sources'][0]) self.assertTextDataIdentical(src, data['sourcesContent'][0]) mappings = json.loads(jsrun.run_js( path_from_root('tools', 'source-maps', 'sourcemap2json.js'), tools.shared.NODE_JS, [map_filename])) seen_lines = set() for m in mappings: self.assertPathsIdentical(src_filename, m['source']) seen_lines.add(m['originalLine']) # ensure that all the 'meaningful' lines in the original code get mapped assert seen_lines.issuperset([6, 7, 11, 12]) # EMCC_DEBUG=2 causes lots of intermediate files to be written, and so # serves as a stress test for source maps because it needs to correlate # line numbers across all those files. old_emcc_debug = os.environ.get('EMCC_DEBUG', None) os.environ.pop('EMCC_DEBUG', None) try: build_and_check() os.environ['EMCC_DEBUG'] = '2' build_and_check() finally: if old_emcc_debug is not None: os.environ['EMCC_DEBUG'] = old_emcc_debug else: os.environ.pop('EMCC_DEBUG', None) def test_exception_source_map(self): if Settings.USE_TYPED_ARRAYS != 2: return self.skip("doesn't pass without typed arrays") if '-g4' not in Building.COMPILER_TEST_OPTS: Building.COMPILER_TEST_OPTS.append('-g4') if NODE_JS not in JS_ENGINES: return self.skip('sourcemapper requires Node to run') src = ''' #include __attribute__((noinline)) void foo(int i) { if (i < 10) throw i; // line 5 } int main() { int i; scanf("%d", &i); foo(i); return 0; } ''' def post(filename): import json map_filename = filename + '.map' mappings = json.loads(jsrun.run_js( path_from_root('tools', 'source-maps', 'sourcemap2json.js'), tools.shared.NODE_JS, [map_filename])) with open(filename) as f: lines = f.readlines() for m in mappings: if m['originalLine'] == 5 and '__cxa_throw' in lines[m['generatedLine']-1]: # -1 to fix 0-start vs 1-start return assert False, 'Must label throw statements with line numbers' dirname = self.get_dir() self.build(src, dirname, os.path.join(dirname, 'src.cpp'), post_build=(None, post)) def test_emscripten_log(self): if Settings.ASM_JS: # XXX Does not work in SpiderMonkey since callstacks cannot be captured when running in asm.js, see https://bugzilla.mozilla.org/show_bug.cgi?id=947996 self.banned_js_engines = [SPIDERMONKEY_ENGINE] if self.emcc_args is None: return self.skip('This test needs libc.') if '-g' not in Building.COMPILER_TEST_OPTS: Building.COMPILER_TEST_OPTS.append('-g') self.do_run('#define RUN_FROM_JS_SHELL\n' + open(path_from_root('tests', 'emscripten_log', 'emscripten_log.cpp')).read(), "Success!") def test_linespecific(self): if Settings.ASM_JS: return self.skip('asm always has corrections on') if '-g' not in Building.COMPILER_TEST_OPTS: Building.COMPILER_TEST_OPTS.append('-g') if self.emcc_args: self.emcc_args += ['--llvm-opts', '0'] # llvm full opts make the expected failures here not happen Building.COMPILER_TEST_OPTS += ['--llvm-opts', '0'] Settings.CHECK_SIGNS = 0 Settings.CHECK_OVERFLOWS = 0 # Signs src = ''' #include #include int main() { int varey = 100; unsigned int MAXEY = -1; printf("*%d*\\n", varey >= MAXEY); // 100 >= -1? not in unsigned! } ''' Settings.CORRECT_SIGNS = 0 self.do_run(src, '*1*') # This is a fail - we expect 0 Settings.CORRECT_SIGNS = 1 self.do_run(src, '*0*') # Now it will work properly # And now let's fix just that one line Settings.CORRECT_SIGNS = 2 Settings.CORRECT_SIGNS_LINES = ["src.cpp:9"] self.do_run(src, '*0*') # Fixing the wrong line should not work Settings.CORRECT_SIGNS = 2 Settings.CORRECT_SIGNS_LINES = ["src.cpp:3"] self.do_run(src, '*1*') # And reverse the checks with = 2 Settings.CORRECT_SIGNS = 3 Settings.CORRECT_SIGNS_LINES = ["src.cpp:3"] self.do_run(src, '*0*') Settings.CORRECT_SIGNS = 3 Settings.CORRECT_SIGNS_LINES = ["src.cpp:9"] self.do_run(src, '*1*') Settings.CORRECT_SIGNS = 0 # Overflows src = ''' #include int main() { int t = 77; for (int i = 0; i < 30; i++) { t = t + t + t + t + t + 1; } printf("*%d,%d*\\n", t, t & 127); return 0; } ''' correct = '*186854335,63*' Settings.CORRECT_OVERFLOWS = 0 try: self.do_run(src, correct) raise Exception('UNEXPECTED-PASS') except Exception, e: assert 'UNEXPECTED' not in str(e), str(e) assert 'Expected to find' in str(e), str(e) Settings.CORRECT_OVERFLOWS = 1 self.do_run(src, correct) # Now it will work properly # And now let's fix just that one line Settings.CORRECT_OVERFLOWS = 2 Settings.CORRECT_OVERFLOWS_LINES = ["src.cpp:6"] self.do_run(src, correct) # Fixing the wrong line should not work Settings.CORRECT_OVERFLOWS = 2 Settings.CORRECT_OVERFLOWS_LINES = ["src.cpp:3"] try: self.do_run(src, correct) raise Exception('UNEXPECTED-PASS') except Exception, e: assert 'UNEXPECTED' not in str(e), str(e) assert 'Expected to find' in str(e), str(e) # And reverse the checks with = 2 Settings.CORRECT_OVERFLOWS = 3 Settings.CORRECT_OVERFLOWS_LINES = ["src.cpp:3"] self.do_run(src, correct) Settings.CORRECT_OVERFLOWS = 3 Settings.CORRECT_OVERFLOWS_LINES = ["src.cpp:6"] try: self.do_run(src, correct) raise Exception('UNEXPECTED-PASS') except Exception, e: assert 'UNEXPECTED' not in str(e), str(e) assert 'Expected to find' in str(e), str(e) Settings.CORRECT_OVERFLOWS = 0 # Roundings src = ''' #include #include int main() { TYPE x = -5; printf("*%d*", x/2); x = 5; printf("*%d*", x/2); float y = -5.33; x = y; printf("*%d*", x); y = 5.33; x = y; printf("*%d*", x); printf("\\n"); } ''' if Settings.USE_TYPED_ARRAYS != 2: # the errors here are very specific to non-i64 mode 1 Settings.CORRECT_ROUNDINGS = 0 self.do_run(src.replace('TYPE', 'long long'), '*-3**2**-6**5*') # JS floor operations, always to the negative. This is an undetected error here! self.do_run(src.replace('TYPE', 'int'), '*-2**2**-5**5*') # We get these right, since they are 32-bit and we can shortcut using the |0 trick self.do_run(src.replace('TYPE', 'unsigned int'), '*-2**2**-6**5*') Settings.CORRECT_ROUNDINGS = 1 Settings.CORRECT_SIGNS = 1 # To be correct here, we need sign corrections as well self.do_run(src.replace('TYPE', 'long long'), '*-2**2**-5**5*') # Correct self.do_run(src.replace('TYPE', 'int'), '*-2**2**-5**5*') # Correct self.do_run(src.replace('TYPE', 'unsigned int'), '*2147483645**2**-5**5*') # Correct Settings.CORRECT_SIGNS = 0 if Settings.USE_TYPED_ARRAYS != 2: # the errors here are very specific to non-i64 mode 1 Settings.CORRECT_ROUNDINGS = 2 Settings.CORRECT_ROUNDINGS_LINES = ["src.cpp:13"] # Fix just the last mistake self.do_run(src.replace('TYPE', 'long long'), '*-3**2**-5**5*') self.do_run(src.replace('TYPE', 'int'), '*-2**2**-5**5*') # Here we are lucky and also get the first one right self.do_run(src.replace('TYPE', 'unsigned int'), '*-2**2**-5**5*') # And reverse the check with = 2 if Settings.USE_TYPED_ARRAYS != 2: # the errors here are very specific to non-i64 mode 1 Settings.CORRECT_ROUNDINGS = 3 Settings.CORRECT_ROUNDINGS_LINES = ["src.cpp:999"] self.do_run(src.replace('TYPE', 'long long'), '*-2**2**-5**5*') self.do_run(src.replace('TYPE', 'int'), '*-2**2**-5**5*') Settings.CORRECT_SIGNS = 1 # To be correct here, we need sign corrections as well self.do_run(src.replace('TYPE', 'unsigned int'), '*2147483645**2**-5**5*') Settings.CORRECT_SIGNS = 0 def test_float_literals(self): self.do_run_from_file(path_from_root('tests', 'test_float_literals.cpp'), path_from_root('tests', 'test_float_literals.out')) def test_exit_status(self): if self.emcc_args is None: return self.skip('need emcc') src = r''' #include #include static void cleanup() { printf("cleanup\n"); } int main() { atexit(cleanup); // this atexit should still be called printf("hello, world!\n"); exit(118); // Unusual exit status to make sure it's working! } ''' open('post.js', 'w').write(''' Module.addOnExit(function () { Module.print('I see exit status: ' + EXITSTATUS); }); Module.callMain(); ''') self.emcc_args += ['-s', 'INVOKE_RUN=0', '--post-js', 'post.js'] self.do_run(src, 'hello, world!\ncleanup\nI see exit status: 118') def test_gc(self): if self.emcc_args == None: return self.skip('needs ta2') if Settings.ASM_JS: return self.skip('asm cannot support generic function table') Settings.GC_SUPPORT = 1 test_path = path_from_root('tests', 'core', 'test_gc') src, output = (test_path + s for s in ('.in', '.out')) self.do_run_from_file(src, output) def test_minmax(self): if self.emcc_args == None: return self.skip('needs emcc') if os.environ.get('EMCC_FAST_COMPILER') == '0': return self.skip('this test will not pass in the old compiler') self.do_run(open(path_from_root('tests', 'test_minmax.c')).read(), 'NAN != NAN\nSuccess!') def test_locale(self): if self.emcc_args is None: return self.skip('needs emcc') self.do_run_from_file(path_from_root('tests', 'test_locale.c'), path_from_root('tests', 'test_locale.out')) def test_64bit_return_value(self): # This test checks that the most significant 32 bits of a 64 bit long are correctly made available # to native JavaScript applications that wish to interact with compiled code returning 64 bit longs. # The MS 32 bits should be available in Runtime.getTempRet0() even when compiled with -O2 --closure 1 # Run with ./runner.py test_64bit_return_value # Compile test.c and wrap it in a native JavaScript binding so we can call our compiled function from JS. Popen([PYTHON, EMCC, path_from_root('tests', 'return64bit', 'test.c'), '--pre-js', path_from_root('tests', 'return64bit', 'testbindstart.js'), '--pre-js', path_from_root('tests', 'return64bit', 'testbind.js'), '--post-js', path_from_root('tests', 'return64bit', 'testbindend.js'), '-s', 'EXPORTED_FUNCTIONS=["_test"]', '-o', 'test.js', '-O2', '--closure', '1'], stdout=PIPE, stderr=PIPE).communicate() # Simple test program to load the test.js binding library and call the binding to the # C function returning the 64 bit long. open(os.path.join(self.get_dir(), 'testrun.js'), 'w').write(''' var test = require("./test.js"); test.runtest(); ''') # Run the test and confirm the output is as expected. if NODE_JS in JS_ENGINES: out = run_js('testrun.js', engine=NODE_JS, full_output=True) assert "low = 5678" in out assert "high = 1234" in out # Generate tests for everything def make_run(fullname, name=-1, compiler=-1, embetter=0, quantum_size=0, typed_arrays=0, emcc_args=None, env=None): if env is None: env = {} TT = type(fullname, (T,), dict(run_name = fullname, env = env)) def tearDown(self): super(TT, self).tearDown() for k, v in self.env.iteritems(): del os.environ[k] # clear global changes to Building Building.COMPILER_TEST_OPTS = [] Building.COMPILER = CLANG Building.LLVM_OPTS = 0 TT.tearDown = tearDown def setUp(self): super(TT, self).setUp() for k, v in self.env.iteritems(): assert k not in os.environ, k + ' should not be in environment' os.environ[k] = v global checked_sanity if not checked_sanity: print '(checking sanity from test runner)' # do this after we set env stuff check_sanity(force=True) checked_sanity = True Building.COMPILER_TEST_OPTS = ['-g'] os.chdir(self.get_dir()) # Ensure the directory exists and go there Building.COMPILER = compiler self.emcc_args = None if emcc_args is None else emcc_args[:] if self.emcc_args is not None: Settings.load(self.emcc_args) Building.LLVM_OPTS = 0 if '-O2' in self.emcc_args or '-O3' in self.emcc_args: Building.COMPILER_TEST_OPTS = [] # remove -g in -O2 tests, for more coverage #Building.COMPILER_TEST_OPTS += self.emcc_args for arg in self.emcc_args: if arg.startswith('-O'): Building.COMPILER_TEST_OPTS.append(arg) # so bitcode is optimized too, this is for cpp to ll else: try: key, value = arg.split('=') Settings[key] = value # forward -s K=V except: pass return # TODO: Move much of these to a init() function in shared.py, and reuse that Settings.ASM_JS = 0 Settings.USE_TYPED_ARRAYS = typed_arrays Settings.INVOKE_RUN = 1 Settings.RELOOP = 0 # we only do them in the "o2" pass Settings.MICRO_OPTS = embetter Settings.QUANTUM_SIZE = quantum_size Settings.ASSERTIONS = 1-embetter Settings.SAFE_HEAP = 1-embetter Settings.CHECK_OVERFLOWS = 1-embetter Settings.CORRECT_OVERFLOWS = 1-embetter Settings.CORRECT_SIGNS = 0 Settings.CORRECT_ROUNDINGS = 0 Settings.CORRECT_OVERFLOWS_LINES = CORRECT_SIGNS_LINES = CORRECT_ROUNDINGS_LINES = SAFE_HEAP_LINES = [] Settings.CHECK_SIGNS = 0 #1-embetter Settings.RUNTIME_TYPE_INFO = 0 Settings.DISABLE_EXCEPTION_CATCHING = 0 Settings.INCLUDE_FULL_LIBRARY = 0 Settings.BUILD_AS_SHARED_LIB = 0 Settings.RUNTIME_LINKED_LIBS = [] Settings.EMULATE_UNALIGNED_ACCESSES = int(Settings.USE_TYPED_ARRAYS == 2 and Building.LLVM_OPTS == 2) Settings.DOUBLE_MODE = 1 if Settings.USE_TYPED_ARRAYS and Building.LLVM_OPTS == 0 else 0 Settings.PRECISE_I64_MATH = 0 Settings.NAMED_GLOBALS = 0 if not embetter else 1 TT.setUp = setUp return TT # Main test modes default = make_run("default", compiler=CLANG, emcc_args=[]) asm1 = make_run("asm1", compiler=CLANG, emcc_args=["-O1"]) asm2 = make_run("asm2", compiler=CLANG, emcc_args=["-O2"]) asm3 = make_run("asm3", compiler=CLANG, emcc_args=["-O3"]) asm2f = make_run("asm2f", compiler=CLANG, emcc_args=["-O2", "-s", "PRECISE_F32=1"]) asm2g = make_run("asm2g", compiler=CLANG, emcc_args=["-O2", "-g", "-s", "ASSERTIONS=1", "--memory-init-file", "1", "-s", "SAFE_HEAP=1"]) # Legacy test modes - slow2 = make_run("slow2", compiler=CLANG, emcc_args=["-O2", "-s", "ASM_JS=0"], env={"EMCC_FAST_COMPILER": "0"}) slow2asm = make_run("slow2asm", compiler=CLANG, emcc_args=["-O2"], env={"EMCC_FAST_COMPILER": "0"}) # Legacy test modes - very old for compiler, quantum, embetter, typed_arrays in [ (CLANG, 4, 0, 0), (CLANG, 4, 1, 1), ]: fullname = 's_0_%d%s%s' % ( embetter, '' if quantum == 4 else '_q' + str(quantum), '' if typed_arrays in [0, 1] else '_t' + str(typed_arrays) ) locals()[fullname] = make_run(fullname, fullname, compiler, embetter, quantum, typed_arrays, env={"EMCC_FAST_COMPILER": "0"}) del T # T is just a shape for the specific subclasses, we don't test it itself