This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] elf: Add tests with a local IFUNC resolver [BZ #23937]
11.03.2019 14:55 Florian Weimer <fweimer@redhat.com> wrote:
> [...]
> I think this is GCC PR81128, see this comment:
>
> <https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81128#c4>
>
> The return value reported suggests that: it does look like an address in
> the executable.
Now I can confirm that GCC 7.2.1-1 which was the newest version
available in Fedora without this patch applied fails while GCC 7.2.1-2
which was the first version with this patch applied passes the tests.
Which, as far as I understand, means that I may have even an older
version of a compiler as long as it has this bug fixed. There
would be no problem with GCC 6.2 or 6.4 except that Fedora does not
distribute them anymore and has never built a patched version.
The tests correctly fail notifying that there is a problem in GCC.
I have another issue which is little off-topic but also little
similar. While playing with different Fedora releases I noticed
that these two tests:
resolv/tst-resolv-ai_idn
resolv/tst-resolv-ai_idn-latin1
may or may not fail depending on the version of libidn2 installed.
Again, there is no specific version which we may recommend as a
minimum. Sometimes newer versions may fail which suggests to me
that it's a matter of some bugs (and whether they are patched) than
any specific version number. Is it again the same case: the test
detects a problem in an external library and the recommendation is
to install the patched build, and patching is more important than
the version number? Shortly, is this anything we should deal with
as glibc or is the current behavior the best we want to provide?
Regards,
Rafal