[LTP] [PATCH 1/2] lib: multiply the timeout if detect slow kconfigsD

[Li Wang]

On Wed, Jan 8, 2025 at 12:49 AM Petr Vorel <pvorel@suse.cz> wrote:

> > Hi!
> > > Ah, you mean we multiply the overall test time limit results->timeout,
> > > right?
> > > e.g.
>
> > >     results->timeout =  (default_30s_timeout + tst_test->timeout) *
> > > TIMEOUT_MUL + tst_test->runtime * RUNTIME_MUL;
> > >     if (tst_has_slow_kconfig())
> > >             results->timeout *= 4;
>
> > That would work too, but since the runtime will be always constant (once
> > only the test that call tst_remaning_runtime() use runtime and the rest
> > of the tests is moved to timeout) we may as well multiply the timeout
> > part.
>
> > However this detail does not matter that much, the most imporatant part
> > is the clear separation of the guessed upper bound and the actual
> > runtime that is used to controll how long should the loop in the test
> > spin.
>
> Thanks this for explanation.
>
> BTW do you plan to send a patch soon? If not I would prefer to temporarily
> revert 2da30df24e (or just *= 4), which causes some tests to timeout.
>

Petr, Cyril, all,

I am now working on the new patch, but it's not guaranteed the new one
will be ready for the coming release.

It's fine to apply this one, at least it addresses the side effects and
extend
timeout time for most tests.



-- 
Regards,
Li Wang