By pass address sanitizer for BadEuler1DSolver member function #261
Conversation
cpp/modmesh/modmesh.hpp
Outdated
| #if defined(USE_SANITIZER) && (defined(__clang__) || defined(__GNUC__)) | ||
| #define ATTRIBUTE_NO_SANITIZE_ADDRESS __attribute__((no_sanitize_address)) | ||
| #else | ||
| #define ATTRIBUTE_NO_SANITIZE_ADDRESS |
There was a problem hiding this comment.
we can leave a TODO here for Windows.
There was a problem hiding this comment.
Thanks for the advice!
I have left a TODO comment in 0132dce
|
I also noticed that there are tons of issues from asan... |
Agree. I'll help to organize these asan issues. If this PR is ready to merge, I think we can open an issue to track these asan issue, then this PR can be close. @tigercosmos, @yungyuc what do you think ? |
j8xixo12
force-pushed
the
sanitizer_enhancement
branch
from
a9ab6f7 to
0132dce
Compare
|
thanks. LGTM |
cpp/modmesh/modmesh.hpp
Outdated
| @@ -40,4 +40,11 @@ | |||
| #include <modmesh/mesh/mesh.hpp> | |||
| #include <modmesh/toggle/toggle.hpp> | |||
|
|
|||
| // TODO Add MSVC case once sanitizer can be default turned on for CI testing | |||
| #if defined(USE_SANITIZER) && (defined(__clang__) || defined(__GNUC__)) | |||
| #define ATTRIBUTE_NO_SANITIZE_ADDRESS __attribute__((no_sanitize_address)) | |||
There was a problem hiding this comment.
What about we name this kind of macros as ASAN_...?
There was a problem hiding this comment.
Thanks for the recommendations, I have changed the marco in 58ed2df
|
LGTM. Merging. |
@tigercosmos, we can continue the discussion of #257 here.
I have added
ElementBase::neg()andElementBase::xpos()to address sanitize whitelist.After bypass address sanity check, modmesh still got tons of memory detection.
Here is the final result snippet :
SUMMARY: AddressSanitizer: 166560 byte(s) leaked in 2070 allocation(s).I also attached the sanitizer log : asan_log.txt
reference : Google sanitizer - How to turn off instrumentation