[PATCH] libstdc++: fix pointer type exception catch [PR105387]

[Jakob Hasse]

This is a patch for the bug 105387 reported in bugzilla: 105387 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105387. This report should contain all the necessary information about the issue. But the patch there was just a preliminary one.
I created a proper patch with the fix and a test case, based on TAG releases/gcc-11.2.0 (7ca388565af176bd4efd4f8db1e5e9e11e98ef45). The changelog is part of the commit message in the patch.

# TESTS
First, I tested only the test case with and without the actual fix, gcc configuration:  ../configure -prefix=/usr --enable-languages=c,c++ --enable-multiarch --host=x86_64-linux-gnu --build=x86_64-linux-gnu --target=x86_64-linux-gnu --enable-cxx-flags=-fno-rtti
		=== libstdc++ Summary ===

# of expected passes		1
# of unexpected failures	1

With the fix:
		=== libstdc++ Summary ===

# of expected passes		2

Then, I tested without RTTI. The gcc configuration was: ../configure -prefix=/usr --enable-languages=c,c++ --enable-multiarch --host=x86_64-linux-gnu --build=x86_64-linux-gnu --target=x86_64-linux-gnu --enable-cxx-flags=-fno-rtti
Without the patch:
		=== libstdc++ Summary ===

# of expected passes		14006
# of unexpected failures	114
# of expected failures		104
# of unresolved testcases	84
# of unsupported tests		663

With the patch:
		=== libstdc++ Summary ===

# of expected passes		14007
# of unexpected failures	113
# of expected failures		104
# of unresolved testcases	84
# of unsupported tests		664

Number of unexpected failures went down by one and the number of unsupported tests went  up by one. 17_intro/headers/c++1998/49745.cc suddenly passes, 22_locale/time_get/get_date/wchar_t/4.cc is suddenly unsupported. I don't know why.

I also tested the entire libstdc++ test suite with RTTI enabled, i.e.: ../configure -prefix=/usr --enable-languages=c,c++ --enable-multiarch --host=x86_64-linux-gnu --build=x86_64-linux-gnu --target=x86_64-linux-gnu --enable-cxx-flags=-frtti

Without patch:
		=== libstdc++ Summary ===

# of expected passes		14201
# of unexpected failures	3
# of expected failures		104
# of unsupported tests		663

With patch:
		=== libstdc++ Summary ===

# of expected passes		14202
# of unexpected failures	2
# of expected failures		104
# of unsupported tests		664

Same symptoms as when running the entire libstdc++ test suite without RTTI.
I'm not sure why the tests behave different, they seem unrelated.

I would be happy to receive feedback, in particular if parts have to be improved or additional information is necessary. I'm completely new to gcc, I'd appreciate concrete recommendations what to do.

Thanks,
Jakob Hasse
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-libstdc-v3-Check-for-ptr-in-__pbase_type_info-__do_c.patch
Type: text/x-patch
Size: 4386 bytes
Desc: 0001-libstdc-v3-Check-for-ptr-in-__pbase_type_info-__do_c.patch
URL: <https://gcc.gnu.org/pipermail/libstdc++/attachments/20220503/80094a99/attachment-0001.bin>