FAIL: nanosleep.gen
===================

nanosleep: /usr/src/debug/strace/6.9/tests/nanosleep.c:114: main: Assertion `k_nanosleep(&req.ts, &rem.ts) == -1' failed.
Aborted
nanosleep({tv_sec=0, tv_nsec=789985}, NULL) = 0
nanosleep(NULL, 0x7ffe7d8e98f0) = -1 EFAULT (Bad address)
nanosleep({tv_sec=0, tv_nsec=789985}, 0x7ffe7d8e98f0) = 0
nanosleep({tv_sec=0, tv_nsec=1000000000}, 0x7ffe7d8e98f0) = -1 EINVAL (Invalid argument)
nanosleep({tv_sec=3735928559, tv_nsec=4207869677}, 0x7ffe7d8e98f0) = -1 EINVAL (Invalid argument)
nanosleep({tv_sec=-3819351491602432273, tv_nsec=13464652301697220333}, 0x7ffe7d8e98f0) = -1 EINVAL (Invalid argument)
nanosleep({tv_sec=-1, tv_nsec=18446744073709551615}, 0x7ffe7d8e98f0) = -1 EINVAL (Invalid argument)
nanosleep({tv_sec=0, tv_nsec=999999999}, 0x7ffe7d8e98f0) = 0
--- SIGALRM {si_signo=SIGALRM, si_code=SI_KERNEL} ---
--- SIGABRT {si_signo=SIGABRT, si_code=SI_TKILL, si_pid=14133, si_uid=0} ---
+++ killed by SIGABRT +++
nanosleep.gen.test: failed test: ../../src/strace -e trace=nanosleep -a20 ../nanosleep failed with code 134
FAIL nanosleep.gen.test (exit status: 1)

Smells like a load-sensitive test that we should just skip to me...

Has not happened in 6 months.

https://autobuilder.yoctoproject.org/valkyrie/#/builders/61/builds/1180/steps/12/logs/stdio

Failed ptests:{'strace': ['nanosleep.gen.test', 'nanosleep.gen']}

qemuarm64-ptest ubuntu2410-vk-arm1

Reopening.

https://autobuilder.yoctoproject.org/valkyrie/#/builders/73/builds/1605/steps/12/logs/stdio

qemux86-64-ptest fedora40-vk-2

https://autobuilder.yoctoproject.org/valkyrie/#/builders/73/builds/1857/steps/12/logs/stdio

Failed ptests:{'strace': ['nanosleep.gen.test', 'nanosleep.gen']}

qemux86-64-ptest ubuntu2204-vk-1

Move to M3 and CC Osama in case he has time to either skip the test, extend the timeout, or talk to upstream about how strace should not have shortish timeouts by default since testing environments may be heavily CPU loaded.

Thanks Randy for CC my, I will take a look.

I can't find this bug anymore in recent builds after upgrading strace to 6.16

https://autobuilder.yoctoproject.org/valkyrie/#/builders/73/builds/2129/steps/12/logs/stdio


Can you please re-check?

BR,
Osama

Hi Osama,

Thanks for checking. The bugs with the "Whiteboard" label of AB-INT mostly or 
only happen in the YP autobuilder environment. We deliberately run 2 builds per machine to get good throughput and we run the ptests in parallel as well. 
We do try to limit the build load using:
   https://git.yoctoproject.org/yocto-autobuilder-helper/tree/config.json
which pulls in:
   https://git.openembedded.org/openembedded-core/tree/meta/conf/fragments/yocto-autobuilder/autobuilder-resource-constraints.conf
so you may have trouble reproducing the error.

The test seems to be generated since it's in nanosleep.gen.test and from a quick look at: tests/gen_tests.in

Assuming that it's generated, and looking for nanosleep in 
strace.git/master, 
I see only test/nanosleep.c and note that it hasn't has a significant commit in Aug 2023:

❯ gl -4 tests/nanosleep.c 
144176806   2023-09-01   Update copyright headers
7178658ea   2023-08-31   tests: avoid libc definition for time_t in direct syscalls
8fa8a8886   2021-02-16   Update copyright headers
9d9e66299   2020-12-31   Change my email address


Anyway, as you can see it's a rare failure, we tried to just assume that it went away one but it came back so the recommendation is to extend the timeout or skip the test by patching the code. It would be interesting to know what the return code of 134 is - EINTR, maybe?
https://man7.org/linux/man-pages/man2/nanosleep.2.html

That's all the time I have tonight... any ideas?

Hi Randy,

Thanks for your check. I'm going to check it more in details and create a patch to extend the timeout.

BR

Created attachment 5137 [details]
proposed fix

Hi Randy,

Please find the patch to skip that test, I found already 
a patch that skips load-sensitive tests, so I added nanosleep to it.

BR,
Osama