[Piglit] Piglit and decision making

[Kenneth Graunke]

On Friday, May 30, 2014 11:58:23 AM Michel Dänzer wrote:
[snip]
> Ken, I think this was really uncalled for. Ilia didn't complain or ask
> for anything, but *posted a patch* to make piglit handle the old, broken
> results files more gracefully.

Yes, and I definitely respect how Ilia responds constructively to these 
things.  When similar issues of workflow breakage have come up in the past, 
not everyone has.  I suppose I'm guilty of that now as well.

[snip]
> I haven't got around to testing your patch, but I fully support its
> intent. Please don't let nay-sayers demotivate you.
> 
> We should always try to keep piglit handling old results files as
> gracefully as possible.

See, this is the problem.  You believe this is an important goal, while Matt 
and I, at least until hearing Ilia's concerns, did not think this was a 
worthwhile goal.  Even if we agreed on retaining compatibility, we probably 
wouldn't agree on "how long".  From Marek's post, it sounds like 3 months is a 
minimum for his workflow.  My workflow is totally different.

We can have a similar debate about other topics.  For example, substantial 
renaming of tests.  It wouldn't affect me much, and one could argue keeps 
Piglit organized.  It would probably make life hard for you and Ilia.

In order to make progress as a project, we need to be able to make decisions 
about things like this.  So what is our process for that?

Mesa feels much simpler to me: the people affected are generally paying close 
attention, so they participate in discussions.  In those discussions, certain 
people's voices carry weight - generally people defer to Brian, or Eric in the 
compiler, or Marek in a lot of places.  It usually only takes a small group of 
people, and they agree on a solution.

Piglit is much more of a free-for-all.  Nobody is "in charge", and we've 
tended to operate on ad-hoc consensus.  But, everybody is busy with their 
"real" work, so the people that will be affected may not even see the 
discussion, much less participate.  Maybe only one side of a debate is 
present...or maybe no one.  After a long wait, someone commits code, and 
suddenly everyone with a different usage model notices, because Piglit is 
totally broken for them.

Sometimes that results in constructive patches, and sometimes it results in 
revert wars and very frustrated people.  Remember the "Is 'Not Run' a change?" 
debacle?  There were several months where Piglit was totally broken for one 
subset of the community or another (maybe it still is, I don't know).  The 
core issue was that we couldn't come to a consensus on what the right behavior 
was.  Half of us wanted one thing, and half wanted another, and maybe a 
quarter of us were even paying attention when patches went in.  So...back and 
forth.

I'm not sure how to improve our process.  Maybe we need to explicitly Cc 
people we think may be affected more often.  I think part of the trouble is 
that the changes which break people's workflow often seem like "no brainers" 
to the authors, so they don't think to ask.

I'm open to ideas and suggestions.  If I'm the only one frustrated, feel free 
to write this off and call me crazy.  But I suspect I'm not alone in this.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.freedesktop.org/archives/piglit/attachments/20140530/7102f889/attachment.sig>