llvm 3.0 anonymous structure support and metadata updates

4 files changed, 22 insertions, 3 deletions

diff --git a/src/analyzer.js b/src/analyzer.js
index 824e7903..8e8f8ac7 100644
--- a/src/analyzer.js
+++ b/src/analyzer.js
@@ -107,6 +107,8 @@ function analyzer(data) {
                               // check that we never allocate with this (either as a child structure
                               // in the analyzer, or in calcSize in alloca).
       var subType = check[2];
+      addTypeInternal(subType, data); // needed for anonymous structure definitions (see below)
+
       Types.types[nonPointing] = {
         name_: nonPointing,
         fields: range(num).map(function() { return subType }),
@@ -124,6 +126,18 @@ function analyzer(data) {
       return;
     }
 
+    // anonymous structure definition, for example |{ i32, i8*, void ()*, i32 }|
+    if (type[0] == '{') {
+      Types.types[type] = {
+        name_: type,
+        fields: splitTokenList(tokenize(type.substr(2, type.length-4)).tokens).map(function(segment) {
+          return segment[0].text;
+        }),
+        lineNum: '?'
+      };
+      return;
+    }
+
     if (isPointerType(type)) return;
     if (['['].indexOf(type) != -1) return;
     Types.types[type] = {
diff --git a/src/intertyper.js b/src/intertyper.js
index 83a49645..56443d20 100644
--- a/src/intertyper.js
+++ b/src/intertyper.js
@@ -2,6 +2,9 @@
 // to be processed by the later stages.
 
 var tokenizer; // TODO: Clean this up/out
+function tokenize(text) {
+  return tokenizer.processItem({ lineText: text }, true);
+}
 
 //! @param parseFunctions We parse functions only on later passes, since we do not
 //!                       want to parse all of them at once, and have all their
diff --git a/src/modules.js b/src/modules.js
index f613c20b..9765fa8d 100644
--- a/src/modules.js
+++ b/src/modules.js
@@ -31,7 +31,8 @@ var Debugging = {
     var form3a = new RegExp(/^!(\d+) = metadata !{i32 \d+, metadata !\d+, i32 \d+, i32 \d+, metadata !(\d+), i32 \d+} ; \[ DW_TAG_lexical_block \]$/);
     var form3ab = new RegExp(/^!(\d+) = metadata !{i32 \d+, i32 \d+, metadata !(\d+), .*$/);
     var form3ac = new RegExp(/^!(\d+) = metadata !{i32 \d+, metadata !\d+, metadata !"[^"]+", metadata !(\d+)[^\[]* ; \[ DW_TAG_.*$/);
-    var form3b = new RegExp(/^!(\d+) = metadata !{i32 \d+, metadata !"([^"]+)", metadata !"([^"]+)", metadata !\d+} ; \[ DW_TAG_file_type \]$/);
+    var form3ad = new RegExp(/^!(\d+) = metadata !{i32 \d+, i32 \d+, null, metadata !"[^"]*", metadata !"[^"]*", metadata !"[^"]*", metadata !(\d+),.*$/);
+    var form3b = new RegExp(/^!(\d+) = metadata !{i32 \d+, metadata !"([^"]+)", metadata !"([^"]+)", (metadata !\d+|null)} ; \[ DW_TAG_file_type \]$/);
     var form3c = new RegExp(/^!(\d+) = metadata !{\w+\d* !?(\d+)[^\d].*$/);
     var form4 = new RegExp(/^!llvm.dbg.[\w\.]+ = .*$/);
     var form5 = new RegExp(/^!(\d+) = metadata !{.*$/);
@@ -79,7 +80,7 @@ var Debugging = {
         metadataToParentMetadata[calc[1]] = calc[3];
         return ';'; // return an empty line, to keep line numbers of subsequent lines the same
       }
-      calc = form3a.exec(line) || form3ab.exec(line) || form3ac.exec(line);
+      calc = form3a.exec(line) || form3ab.exec(line) || form3ac.exec(line) || form3ad.exec(line);
       if (calc) {
         metadataToParentMetadata[calc[1]] = calc[2];
         return ';';
diff --git a/src/parseTools.js b/src/parseTools.js
index a3347139..78b88ea8 100644
--- a/src/parseTools.js
+++ b/src/parseTools.js
@@ -114,6 +114,7 @@ function isStructPointerType(type) {
 function isStructType(type) {
   if (isPointerType(type)) return false;
   if (new RegExp(/^\[\d+\ x\ (.*)\]/g).test(type)) return true; // [15 x ?] blocks. Like structs
+  if (new RegExp(/{ [^}]* }/g).test(type)) return true; // { i32, i8 } etc. - anonymous struct types
   // See comment in isStructPointerType()
   return !Runtime.isNumberType(type) && type[0] == '%';
 }
@@ -148,7 +149,7 @@ function isFunctionType(type) {
   if (pointingLevels(type) !== 1) return false;
   var text = removeAllPointing(parts.slice(1).join(' '));
   if (!text) return false;
-  return isType(parts[0]) && isFunctionDef({ text: text, item: tokenizer.processItem({ lineText: text.substr(1, text.length-2) }, true) });
+  return isType(parts[0]) && isFunctionDef({ text: text, item: tokenize(text.substr(1, text.length-2), true) });
 }
 
 function isType(type) { // TODO!