This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] signal: Use correct type for si_band in siginfo_t [BZ #23562]
On 22/10/2018 12:07, Joseph Myers wrote:
> This has broken the conform/ tests for SPARC, so that you now get:
>
> FAIL: conform/POSIX2008/signal.h/conform
> FAIL: conform/POSIX2008/sys/wait.h/conform
> FAIL: conform/UNIX98/signal.h/conform
> FAIL: conform/UNIX98/sys/wait.h/conform
> FAIL: conform/XOPEN2K/signal.h/conform
> FAIL: conform/XOPEN2K/sys/wait.h/conform
> FAIL: conform/XOPEN2K8/signal.h/conform
> FAIL: conform/XOPEN2K8/sys/wait.h/conform
> FAIL: conform/XPG42/signal.h/conform
> FAIL: conform/XPG42/sys/wait.h/conform
>
> because si_band no longer has the POSIX-specified type.
>
> For 32-bit SPARC you should be able to use long int and so avoid the
> failures, as that won't change the layout. For 64-bit SPARC this would be
> a case where the kernel having the wrong type means you need an
> appropriately conditioned and commented XFAIL in signal.h-data and
> sys/wait.h-data (and appropriately conditioned and commented
> conformtest-xfail-conds definition in a relevant sysdeps Makefile to
> define the condition that's used in the conform/ data changes).
>
I think we should XFAIL for both 32-bit and 64-bit, siginfo_t is passed
directly by kernel in signal handlers and passed unmodified on sigtimedwait,
sigwait, and sigwaitinfo syscalls.