| Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
|
|
the default build
|
|
|
|
|
|
|
|
HAVE_DECL_PT_GETREGSET
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
-D__ANDROID_API__=$ANDROID_API_V in standalone compilation mode
|
|
|
|
Closes google/honggfuzz#159
|
|
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
All CPU targets have matching names expect arm64 (vs aarch64). Create
a softlink for this case so that we don't need create more target naming
conventions in makefiles.
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
When user clones repo without submodules, by default most
git versions appear to create just an empty directory for them. As
such just checking if directory present to init submodules is not
enough. Checks have been upgraded to use the present of
projects .git dir as an indicator for missing submodule initialization.
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
If repo was cloned with recursive submodule fetch, the client
hooks were not registered from build scripts. As such split
the missing submodule update from client hooks register
logic.
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
Each build script is dealing with a separate submodule. As
such don't recursively fetch them all in one step, since not
all client hooks will be registered.
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
If user updates dep submodules (capstone, libunwind)
although forgets to make a clean build, old .a files persist
the update resulting into an invalid setup. Avoid such case
by registering post-checkout git hooks for the two submodules.
As soon as the submodules are updated the client git hooks
ensure the previous .a files will be deleted forcing the user to
rebuild with new source code version.
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
We're applying a set of patches to upstream libunwind to be
compatible with some Android specific internals. We want the
repo to be left in a clean state to accept future submodule
updates without requiring user manual steps. As such after
static library for selected CPU is built, reset local applied
changes.
Output .a files persist across git reset so users can have them
available across different CPU target builds.
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
libunwind imports have to be re-ordered to pick the correct
"_Ux86_64_setcontext" definition. Small typo fixes too.
Signed-off-by: Anestis Bechtsoudis <anestis@census-labs.com>
|
|
|
|
Fix numerous Android bugs in libunwind for ARM64
|
|
|
|
* Fork upstream libunwind and cross-compile with Android NDK
* libunwind is also used for proc symbol resolve instead of bfd
* Replace BDF lib with capstone for disassembling actions
* Engineer build & patch scripts to support Android ARM, ARM64, x86, x86_64
* Rename linux/ptrace.* to linux/ptrace_utils.* since it conflicts with Android platform includes for ptrace API
* Wrap register & instruction sizes under macros to reflect sizes for supported CPU arch (makes debugging easier)
* Add / Implement missing definitions / functions for Android
* Improve Android build process
|