[Piglit] [PATCH 3/3] piglit: Add piglit command

[Daniel Vetter]

On Wed, Apr 30, 2014 at 03:40:16PM -0700, Dylan Baker wrote:
> On Wednesday, April 30, 2014 15:21:37 Jordan Justen wrote:
> > On Wed, Apr 30, 2014 at 1:58 PM, Dylan Baker <baker.dylan.c at gmail.com> wrote:
> > > On Wednesday, April 30, 2014 13:42:09 Jordan Justen wrote:
> > >> Anyway, piglit_cmd.py is the source that controls the piglit command,
> > >> so it can evolve to present something different.
> > >>
> > >> This patch seems like a reasonable step which would allow you to then
> > >> re-implement the unified interface as you see fit.
> > >
> > > It's not the interface I have a problem with, it's the implementation.
> > >
> > > This isn't what we want at all. There is zero code here I would want to
> > > use, and it creates a situation where the user interface will change,
> > > and we're left with unhappy users.
> > 
> > So, you are fine with the interface, but it is going to have to change?
> > 
> > And, our users (??) will be unhappy if we change the interface, but
> > you want to do something similar, and they'll be okay with that
> > change?
> 
> Let me rephrase. I'm good with the general idea
> Let my draw the scenario I'm worried about, people (including our QA
> team) beginusing this wrapper. We decide that another layer of argparse
> is simpler and easier to work with. There are minor differences which
> break users wrapper scripts and they become unhappy.

Note that our QA didn't update _any_ piglit based test list (for mesa,
opencl or igt) for about a month due to the addition of the non-verbose
mode breaking their scripts. And I'm trying to get them to integrate even
more tightly so that we can share e.g. dmesg filtering rules.

Of course part of the problem is the complete lack of automated tests and
sanity checks for their NIH infrastructre.

Just my 2c.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch