Re: [LTP] [PATCH 1/2] tst_test: Add $LTP_SINGLE_ITERATION to limit variant

[Petr Vorel <pvorel@suse.cz>]

> On Mon, Feb 24, 2025 at 6:21 PM Petr Vorel <pvorel@suse.cz> wrote:

> > Hi Li,

> > > > Hi all,

> > > > Andrea wanted to run only single variant.

> > > Could you tell me what the benefit is? why do we need this?

> > @Andrea ^

> > IMHO similar to LTP_SINGLE_FS_TYPE - speedup. Although I admit
> > LTP_SINGLE_FS_TYPE allows much bigger cleanup.

> > > And back to the patch itself, if LTP_SINGLE_ITERATION is used, which
> > > variant will be chosen to run by default? Is that expected?

> > By default all variants are tested.


> Sorry, here I wanted to ask:
>  Which variant will be selected when LTP_SINGLE_VARIANT is set?
>  Is the selected variant the expected one?

Well, test variant can mean anything. It started as libc variant or raw syscall,
but getdents.h has 4 variants (raw syscall SYS_getdents, raw syscall
SYS_getdents64, libc getdents() and libc getdents64()), futex_utils.h has 2
conditional variants variants syscall with old kernel spec and syscall time64
with kernel spec (often only one variant will be set), tests which are use
select_var.h have 5 variants, ...


> Since the order of test variants is not standardized and is determined
> by the author while test writing. Once we used LTP_SINGLE_VARIANT
> it still does not guarantee the executed variant is the same for all tests.
> Unless people only use it in manual debugging mode.

As Andrea wrote, the goal is to use it for debugging of a particular test.
Yes, it will not be useful to run whole syscalls with it defined.

Kind regards,
Petr

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