Bug#1040817: glibc: Please ignore some tests on sparc64

[Aurelien Jarno <aurel32@debian.org>]

Hi,

On 2023-07-11 06:17, John Paul Adrian Glaubitz wrote:
> Source: glibc
> Version: 2.37-5
> Severity: normal
> User: debian-sparc@lists.debian.org
> Usertags: sparc64
> X-Debbugs-Cc: debian-sparc@lists.debian.org
> 
> Hi!
> 
> The list of currently failing tests on sparc64 is:
> 
> FAIL: elf/tst-audit24a
> FAIL: elf/tst-audit24b
> FAIL: elf/tst-audit24c
> FAIL: elf/tst-audit24d
> FAIL: elf/tst-rtld-run-static
> FAIL: nptl/tst-cancel24-static
> FAIL: nptl/tst-cancel30
> FAIL: socket/tst-socket-timestamp
> FAIL: stdlib/isomac
> 
> According to upstream, the following audit tests are not going to be
> fixed soon since the SPARC ABI makes it more difficult:
> 
> FAIL: elf/tst-audit24a
> FAIL: elf/tst-audit24b
> FAIL: elf/tst-audit24c
> FAIL: elf/tst-audit24d

Ok.

> These are going to be fixed upstream soon, the fixes are supposedly
> trivial:
> 
> FAIL: elf/tst-rtld-run-static
> FAIL: nptl/tst-cancel24-static
> FAIL: nptl/tst-cancel30

Great.

> This test is supposedly a kernel issue:
> 
> FAIL: socket/tst-socket-timestamp
> 
> And this one allegedly not related to sparc64:
> 
> FAIL: stdlib/isomac

What do you mean by "allegedly not related to sparc64"? This failure
only appears on sparc*. The sparc32 has the following comment in
debian/testsuite-xfail-debian.mk to ignore the failure:

  # Even if configured using --with-long-double-128, the biarch32 compiler
  # on sparc64 defaults to 64-bit doubles, causing the failure below. This
  # should be fixed by the following gcc patch:
  # http://gcc.gnu.org/ml/gcc-patches/2013-12/msg00318.html
  test-xfail-stdlib/isomac = yes

Can you please check if there is a similar issue with the GCC
configuration on sparc64?

> So, my suggestion would be to ignore the following tests for now:
> 
> FAIL: elf/tst-audit24a
> FAIL: elf/tst-audit24b
> FAIL: elf/tst-audit24c
> FAIL: elf/tst-audit24d
> FAIL: socket/tst-socket-timestamp
> FAIL: stdlib/isomac

Ok.

> And looking at the testsuite results with 32-bit tests enabled [1], it looks like
> the failures are the same. So, I think we can just ignore the above tests and then
> re-enable testing on 32 bit as well.

It appears that none of those fails on sparc32. Looking at it again, it
even appears that the sparc32 build passed the testsuite without issue,
so there was no need to disable it.

Cheers
Aurelien

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurelien@aurel32.net                     http://aurel32.net