Re: [NOMERGE] glibc: update to 2.31.

[voidlinux-github@inbox.vuxu.org]

[-- Attachment #1: Type: text/plain, Size: 1588 bytes --]

New comment by ailiop-git on void-packages repository

https://github.com/void-linux/void-packages/pull/18737#issuecomment-581163700

Comment:
> I just noticed sshd breaking due to this upgrade, so please hold off merging.

The sshd is breaking due to its seccomp filter not explicitly allowing the clock_nanosleep and clock_nanosleep_time64 syscalls which are being called due to glibc commit 2e44b10b42 ("sysdeps/clock_nanosleep: Use clock_nanosleep_time64 if avaliable"), so logins fail as the child process is being killed.

This is fixed in openssh upstream with commit b1c82f4b ("seccomp: Allow clock_nanosleep() in sandbox."), which will be part of the next release (see also [1]). 

I've backported this for now, but I'm afraid that this means there will be many other packages breaking due to using seccomp sandboxes (note that this is especially valid on 64-bit archs, as glibc is now invoking clock_nanosleep_time64 which has a different syscall nr).

There's also at least one other issue, glibc 2.31 has removed the __glibc_has_include macro (commit bfa864e164 ("Don't use a custom wrapper macro around __has_include (bug 25189)."), and this breaks compilation of sources that include stat.h/statx.h, as gcc does fixincludes on those headers (it will use the gcc fixed-up /usr/lib/gcc/x86_64-unknown-linux-gnu/9.2/include-fixed/bits/statx.h while glibc-2.31 has removed the macro, and fail). See also [2], this seems to be in progress with gcc upstream.

[1] https://bugzilla.mindrot.org/show_bug.cgi?id=3093
[2] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91085