You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We should reject the configuration at the cmake or make step. Doing it in cmake may be easier, but doing it at the make step (ideally through a #warning directive so that it's independent of the build system) would have the advantage that it would work even if config.h is changed after running cmake or when using other build systems.
The text was updated successfully, but these errors were encountered:
This should be fairly easy to do at the source level:
#if defined(__has_feature)
#if __has_feature(memory_sanitizer)
#warning "This function doesn't work with Memory Sanitizer"
#endif
#endif
We could also try to make the tests actually pass using __attribute__((no_sanitize("memory"))) on some functions (but I'm not sure if it still marks memory written by those functions as initialised) or the sanitizer special case list, though last time I tried the latter approach I couldn't get anywhere.
Steps to reproduce (on
mbedtls-2.7.0
or 2.8):Several tests fail. In fact we know about this —
all.sh
saysWe should reject the configuration at the
cmake
ormake
step. Doing it incmake
may be easier, but doing it at themake
step (ideally through a#warning
directive so that it's independent of the build system) would have the advantage that it would work even ifconfig.h
is changed after runningcmake
or when using other build systems.The text was updated successfully, but these errors were encountered: