Always compare C++ typeinfo (based on libstdc++ implementation).

[marxin]

Summary

Resending https://reviews.llvm.org/D56485

As seen here:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88684#c4

GCC libstdc++v3 implementation is __GXX_MERGED_TYPEINFO_NAMES=0 for all targets.
Thus it should be always compared string-wise. And one needs to check for name[0] != '*' as seen in typeinfo
implementation:

libstdc++-v3/libsupc++/typeinfo

[marxin]

There hasn't been any reply yet. May I understand that as approval?

[dexonsmith]

I’d like @kubamracek and/or @Bigcheese to take a look first.

[Bigcheese]

I do not believe this change is correct for MacOS or iOS. The original bug does not reproduce on MacOS, and on arm64 iOS you always need to compare strings. Could you preserve the existing behavior for those platforms?

[marxin]

In D58028#1398872, @Bigcheese wrote:

I do not believe this change is correct for MacOS or iOS. The original bug does not reproduce on MacOS, and on arm64 iOS you always need to compare strings. Could you preserve the existing behavior for those platforms?

I can if you want. Have you tested that on MacOS or iOS with:

-stdlib=libstdc++

?

[Bigcheese]

In D58028#1402149, @marxin wrote:

I can if you want. Have you tested that on MacOS or iOS with:

-stdlib=libstdc++

?

I have not, but this behavior depends on how clang emits the symbols and how the platform dynamic linker works, not on the standard library. We do not use libstdc++, so this needs to be correct for other implementations too.

[marxin]

Updated version which preserves behavior on MACOS and IOS. For other targets SANITIZER_NON_UNIQUE_TYPEINFO is set to true.

[marxin]

Gentle ping.

[Bigcheese]

Thanks for the update. This still looks incorrect in that SANITIZER_NON_UNIQUE_TYPEINFO should be 1 on defined(__arm64__) && SANITIZER_IOS and 0 otherwise on SANITIZER_MAC.

I believe the following is correct.

#if SANITIZER_MAC && !(defined(__arm64__) && SANITIZER_IOS)
# define SANITIZER_NON_UNIQUE_TYPEINFO 0
#else
# define SANITIZER_NON_UNIQUE_TYPEINFO 1
#endif

[marxin]

Update version where I address last comment.

[marxin]

In D58028#1414723, @marxin wrote:

Update version where I address last comment.

May I please remind that?

[Bigcheese]

LGTM.

lib/sanitizer_common/sanitizer_platform.h

 ///  * 1800: Microsoft Visual Studio 2013 / 12.0
 ///  * 1900: Microsoft Visual Studio 2015 / 14.0
 #ifdef _MSC_VER
 # define MSC_PREREQ(version) (_MSC_VER >= (version))
 #else
 # define MSC_PREREQ(version) 0
 #endif
 
-#if defined(__arm64__) && SANITIZER_IOS
-# define SANITIZER_NON_UNIQUE_TYPEINFO 1
-#else
+#if SANITIZER_MAC && !(defined(__arm64__) && SANITIZER_IOS)
 # define SANITIZER_NON_UNIQUE_TYPEINFO 0
+#else
+# define SANITIZER_NON_UNIQUE_TYPEINFO 1
 #endif
 
 // On linux, some architectures had an ABI transition from 64-bit long double
 // (ie. same as double) to 128-bit long double.  On those, glibc symbols
 // involving long doubles come in two versions, and we need to pass the
 // correct one to dlvsym when intercepting them.
 #if SANITIZER_LINUX && (SANITIZER_S390 || SANITIZER_PPC32 || SANITIZER_PPC64V1)
 #define SANITIZER_NLDBL_VERSION "GLIBC_2.4"

lib/ubsan/ubsan_type_hash_itanium.cc

 
 /// \brief Determine whether \p Derived has a \p Base base class subobject at
 /// offset \p Offset.
 static bool isDerivedFromAtOffset(const abi::__class_type_info *Derived,
                                   const abi::__class_type_info *Base,
                                   sptr Offset) {
   if (Derived->__type_name == Base->__type_name ||
       (SANITIZER_NON_UNIQUE_TYPEINFO &&
+       Derived->__type_name[0] != '*' &&
        !internal_strcmp(Derived->__type_name, Base->__type_name)))
     return Offset == 0;
 
   if (const abi::__si_class_type_info *SI =
         dynamic_cast<const abi::__si_class_type_info*>(Derived))
     return isDerivedFromAtOffset(SI->__base_type, Base, Offset);
 
   const abi::__vmi_class_type_info *VTI =