Re: [PATCH 2/2] intel-gpu-tools: Re-use igt_subtest_init for simple tests

[Daniel Vetter <daniel@ffwll.ch>]

On Wed, Jul 09, 2014 at 01:23:46PM +0000, Gore, Tim wrote:
> Some comments on Daniels' comments inline
> 
> > -----Original Message-----
> > From: Daniel Vetter [mailto:daniel.vetter@ffwll.ch] On Behalf Of Daniel
> > Not sure whether forcing the reuse of igt_subtest_init makes a lot of sense
> > since it requires a lot of follow-up changes all over. I'd just add a bit of
> > argparsing here with the required special cases, i.e.
> > - exit without doing anything for --list-subtests
> > - exit with 77 when anything is specified with --run-subtests
> > 
>       What I wanted to do here was start removing the distinction between single
>      Tests and multiple test; this distinction seems somewhat artificial and doesn't
>       seem to add much. 
>      igt_subtest_init does pretty much everything we want for single tests
>      so I thought it made sense to re-use it. Perhaps the name should change,
>      although this would lead to more knock on changes.

Well under the hood subtests work with longjmps when skipping/failing
tests while simple tests directly call exit. The reason for that is that
historically we've started with simple binaries and slowly grew the igt
infrastructure on top of them. I don't really see much benefits in
converting the last stragglers.

> > Also I prefer if the piglit changes come together with this patch so that we
> > can roll it all out together.
> 
>     What piglet changes do you refer to? I have not made any piglit changes.

For me the only trouble with this disdinction is that people consistently
place tests in the wrong Makefile target. Hence I'd like to see those two
Makefile targets being unified (which requires adjustements in piglit,
too) to validate this work. Iirc Thomas is also working on this.

Cheers, Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch