Re: glibc 2.34 failed tests

[Adhemerval Zanella <adhemerval.zanella@linaro.org>]

On 04/08/2021 09:29, Michael Brunnbauer wrote:
> 
> hi
> 
> On Wed, Aug 04, 2021 at 08:56:17AM -0300, Adhemerval Zanella wrote:
>>> FAIL: elf/tst-dlopenfail
>>
>> This is a new failure, could you post the .out output?
>>
>>> FAIL: malloc/tst-safe-linking
> 
> Those two failures occured with glibc 2.33 but not with 2.34 on my current 
> machine.
> 
>>> FAIL: malloc/tst-tcfree2
>>
>> This is another new failure., could you post the .out output as well?
> 
> FAIL: tcache double free
> Expected signal 'Aborted' from child, got none
> 
>>> FAIL: nss/tst-nss-files-hosts-multi
>>> FAIL: resolv/tst-bug18665
>>> FAIL: resolv/tst-resolv-search
>>
>> I recall this might fail depending of how the system network is configured.
>> What .out file shows?
> 
> It's an ipv4 only system. Here are the outputs:
> 
> cat nss/tst-nss-files-hosts-multi.out
> Timed out: killed the child process
> Termination time: 2021-08-03T15:50:02.888711343
> Last write to standard output: 2021-08-03T15:49:22.870384035
> 
> cat resolv/tst-bug18665.out
> warning: unshare (CLONE_NEWUTS) failed: Invalid argument
> warning: could not enter network namespace
> Timed out: killed the child process
> Termination time: 2021-08-03T15:48:22.269459412
> Last write to standard output: 2021-08-03T15:48:02.250298068
> 
> cat resolv/tst-resolv-search.out
> warning: unshare (CLONE_NEWUTS) failed: Invalid argument
> warning: could not enter network namespace
> Timed out: killed the child process
> Termination time: 2021-08-03T15:48:52.829291960
> Last write to standard output: 2021-08-03T15:48:32.810330616

This is due the missing kernel support for network namespace, I think we
should set the tests as unsupported in this case.

> 
> 
>>> cat ./rt/tst-mqueue10.out
>>> error: tst-mqueue10.c:46: not true: q != (mqd_t) -1
>>> error: 1 test failures
>>
>> I think this might be related to the fact rt/tst-mqueue10.c uses the
>> "/tst-mqueue2-" as template (it might interfere with rt/tst-mqueue2.c
>> itself).
> 
> I opened a ticket for this one: https://sourceware.org/bugzilla/show_bug.cgi?id=28188

Thanks, but the only reason I see to this fail is if the kernel 
misses POSIX message queue support.  Could you check if your kernel
has CONFIG_POSIX_MQUEUE enabled? 

> 
>>> cat ./time/tst-itimer.out
>>> tst-itimer.c:76: numeric comparison failure (widths 64 and 32)
>>>    left: 10000 (0x2710); from: it_old.it_interval.tv_usec
>>>   right: 20 (0x14); from: 20
>>> tst-itimer.c:90: numeric comparison failure
>>>    left: 20 (0x14); from: it.it_interval.tv_usec
>>>   right: 10000 (0x2710); from: it_old.it_interval.tv_usec
>>> tst-itimer.c:118: numeric comparison failure
>>>    left: 2748779068 (0xa3d70a3c); from: it_old.it_interval.tv_sec
>>>   right: 8589934591 (0x1ffffffff); from: 0x1ffffffffull
>>> tst-itimer.c:119: numeric comparison failure (widths 64 and 32)
>>>    left: 0 (0x0); from: it_old.it_interval.tv_usec
>>>   right: 20 (0x14); from: 20
>>> tst-itimer.c:141: numeric comparison failure
>>>    left: 8589934591 (0x1ffffffff); from: it.it_interval.tv_sec
>>>   right: 2748779068 (0xa3d70a3c); from: it_old.it_interval.tv_sec
>>> tst-itimer.c:142: numeric comparison failure
>>>    left: 20 (0x14); from: it.it_interval.tv_usec
>>>   right: 0 (0x0); from: it_old.it_interval.tv_usec
>>> tst-itimer.c:76: numeric comparison failure (widths 64 and 32)
>>>    left: 10000 (0x2710); from: it_old.it_interval.tv_usec
>>>   right: 20 (0x14); from: 20
>>> tst-itimer.c:90: numeric comparison failure
>>>    left: 20 (0x14); from: it.it_interval.tv_usec
>>>   right: 10000 (0x2710); from: it_old.it_interval.tv_usec
>>> tst-itimer.c:118: numeric comparison failure
>>>    left: 2748779068 (0xa3d70a3c); from: it_old.it_interval.tv_sec
>>>   right: 8589934591 (0x1ffffffff); from: 0x1ffffffffull
>>> tst-itimer.c:119: numeric comparison failure (widths 64 and 32)
>>>    left: 0 (0x0); from: it_old.it_interval.tv_usec
>>>   right: 20 (0x14); from: 20
>>> tst-itimer.c:141: numeric comparison failure
>>>    left: 8589934591 (0x1ffffffff); from: it.it_interval.tv_sec
>>>   right: 2748779068 (0xa3d70a3c); from: it_old.it_interval.tv_sec
>>> tst-itimer.c:142: numeric comparison failure
>>>    left: 20 (0x14); from: it.it_interval.tv_usec
>>>   right: 0 (0x0); from: it_old.it_interval.tv_usec
>>> error: 12 test failures
>>
>> I almost sure this is the same issue I saw on an older kernel under KVM [1]
>>
>> [1] https://patchwork.sourceware.org/project/glibc/patch/20210719163846.2954193-3-adhemerval.zanella@linaro.org/
> 
> My machine is not a virtual one though. Here is the ticket for this one:
> 
> https://sourceware.org/bugzilla/show_bug.cgi?id=28189

Thanks, but I do not think KVM is the only thing that changes the clock
precision here.