Re: [LTP] [PATCH] waitpid01: Test all standard deadly signals

[Petr Vorel <pvorel@suse.cz>]

Hi Martin,

> On 05. 02. 24 18:34, Petr Vorel wrote:
> > Hi Martin,

> > first this change (merged as ed5ccf6c1 ("waitpid01: Test all standard deadly
> > signals") is failing on 6.6 on Debian:

> > ed5ccf6c1 ("waitpid01: Test all standard deadly signals")

> > waitpid01.c:88: TINFO: Testing child: raise(sig)
> > waitpid01.c:109: TPASS: waitpid() returned correct pid 236064
> > waitpid01.c:118: TPASS: WIFSIGNALED() set in status
> > waitpid01.c:126: TPASS: WTERMSIG() == SIGIOT/SIGABRT
> > waitpid01.c:143: TPASS: Child dumped core as expected
> > waitpid01.c:109: TPASS: waitpid() returned correct pid 236071
> > waitpid01.c:118: TPASS: WIFSIGNALED() set in status
> > waitpid01.c:126: TPASS: WTERMSIG() == SIGALRM
> > waitpid01.c:109: TPASS: waitpid() returned correct pid 236072
> > waitpid01.c:118: TPASS: WIFSIGNALED() set in status
> > waitpid01.c:126: TPASS: WTERMSIG() == SIGBUS
> > waitpid01.c:143: TPASS: Child dumped core as expected
> > waitpid01.c:109: TPASS: waitpid() returned correct pid 236082
> > waitpid01.c:113: TFAIL: WIFSIGNALED() not set in status (exited with 0)
> > ...
> > waitpid01.c:88: TINFO: Testing child: kill(getpid(), sig)
> > waitpid01.c:109: TPASS: waitpid() returned correct pid 236165
> > waitpid01.c:118: TPASS: WIFSIGNALED() set in status
> > waitpid01.c:126: TPASS: WTERMSIG() == SIGIOT/SIGABRT
> > waitpid01.c:143: TPASS: Child dumped core as expected
> > waitpid01.c:109: TPASS: waitpid() returned correct pid 236176
> > waitpid01.c:118: TPASS: WIFSIGNALED() set in status
> > waitpid01.c:126: TPASS: WTERMSIG() == SIGALRM
> > waitpid01.c:109: TPASS: waitpid() returned correct pid 236177
> > waitpid01.c:118: TPASS: WIFSIGNALED() set in status
> > waitpid01.c:126: TPASS: WTERMSIG() == SIGBUS
> > waitpid01.c:143: TPASS: Child dumped core as expected
> > waitpid01.c:109: TPASS: waitpid() returned correct pid 236187
> > waitpid01.c:113: TFAIL: WIFSIGNALED() not set in status (exited with 0)
> > ...

> > But it works on 6.8.0-rc1-2.gf4ba5db-default Tumbleweed, which I tested it
> > before.

> I've tested the patch on kernel v6.6.1 on Tumbleweed and all testcases pass.
> The failure shows that SIGFPE doesn't trigger coredump for some reason,
> that's either a system bug or some parent process left over a SIGFPE
> handler. Could you try the Debian test again with this patch?

> diff --git a/testcases/kernel/syscalls/waitpid/waitpid01.c
> b/testcases/kernel/syscalls/waitpid/waitpid01.c
> index 367f7875c..a42b7e8ac 100644
> --- a/testcases/kernel/syscalls/waitpid/waitpid01.c
> +++ b/testcases/kernel/syscalls/waitpid/waitpid01.c
> @@ -94,6 +94,9 @@ static void run(unsigned int n)
>         int status;
>         const struct testcase *tc = testcase_list + n;

> +       if (tc->sig != SIGKILL)
> +               SAFE_SIGNAL(tc->sig, SIG_DFL);
> +

Ah, reset the signal, thanks!

It works on the current master, but because different patch would broke your
waiting waitpid01 patch, so that I'll review and merge it first.

Will you send this as a patch or should I merge this as your Suggested-by: ?

Kind regards,
Petr

>         pid = SAFE_FORK();
>         if (!pid)
>                 variant_list[tst_variant].func(tc->sig);

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp