[LTP] [PATCH v6 2/4] tst_test.c: Add tst_cmd_present check if a command is present

[Petr Vorel]

> Hi!
> > 1) tst_brk for not checked command at the end.
> > 2) tst_brk for NULL otherwise we get segfault. Maybe check also for empty
> > command before attempting to do anything else.

> Technically the empty command will end up in the tst_brk() at the end,
> so I would add only the NULL check.

Sure, just the message will be a bit confusing:
'' not checked.

That's why I checked also for empty command.
Anyway, v7 is out with cmd[0] == '\0' as well. But it can be removed during
merge.
https://lore.kernel.org/ltp/20260109061716.20258-3-wegao@suse.com/

> > index bf92e6cc92..cf67c2a9f0 100644
> > --- lib/tst_test.c
> > +++ lib/tst_test.c
> > @@ -1379,12 +1379,16 @@ bool tst_cmd_present(const char *cmd)
> >  {
> >  	struct tst_cmd *pcmd = tst_test->needs_cmds;

> > +	if (!cmd || cmd[0] == '\0')
> > +		tst_brk(TBROK, "Invalid cmd");
> > +
> >  	while (pcmd->cmd) {
> >  		if (!strcmp(pcmd->cmd, cmd))
> >  			return pcmd->present;

> >  		pcmd++;
> >  	}
> > +	tst_brk(TBROK, "'%s' not checked", cmd);
> >  	return false;
> >  }


> > 3) Shouldn't we use tst_brk_() to point to the code in the test instead of
> > tst_brk()? If yes, we probably want to have tst_cmd_present_() as library
> > internal (static, with const char *file, const int lineno parameters) which
> > calls tst_brk_() and tst_cmd_present() as public (the only with signature in
> > header).

> I do not think that this is that important to have, the
> tst_cmd_present() calls are going to be quite rare and hence easy to
> locate in the sources.

OK, that simplifies things.

Kind regards,
Petr