-----Original Message-----
> I followed up the latest upstream changes,
> the implementation has been improved and simply uses std::string as the variable
type('fold_storage').
> Furthermore, crash tools will also have minor changes. How about the following
changes?
hmm, personally I'd prefer to simply backport only the two small patches
6a780b676637 and 2ccee230f830. The gdb commit 5f9febe0f6ef doesn't fix any
bug, the two patches are enough to fix the issue for now.
Unfortunately it doesn't work, so I have to look into the details of why it failed.
Maybe it could be related
to the build issues or gdb, or other defects? It's strange.
BTW: Have you reproduced this issue with my flags?
Yes, I have reproduced the issue with a hand-built crash with crash.spec and
the flags.
1. modify crash.spec and build
cp %{SOURCE1} .
-make RPMPKG="%{version}-%{release}" CFLAGS="%{optflags}"
CXXFLAGS="%{optflags}" LDFLAGS="%{build_ldflags}"
+make RPMPKG=8.0.0-5.el9 'CFLAGS=-O2 ...
# rpmbuild -ba crash.spec
2. reproduce
# BUILD/crash-8.0.0/crash
...
crash> p /usr/include/c++/11/string_view:234:...
3. modify crash.spec and build with the patch
+Patch7: gdb-assert.patch
+%patch7 -p1
# rpmbuild -ba crash.spec
4. cannot reproduce
# BUILD/crash-8.0.0.fix/crash
...
crash> p
Display all 174031 possibilities? (y or n)
Thanks,
Kazu