[Mesa-dev] [PATCH] autotools: Deprecate the use of autotools

[Ilia Mirkin]

On Wed, Dec 19, 2018 at 11:03 AM Jason Ekstrand <jason at jlekstrand.net> wrote:
>
> On Wed, Dec 19, 2018 at 9:32 AM Ilia Mirkin <imirkin at alum.mit.edu> wrote:
>>
>> On Wed, Dec 19, 2018 at 10:25 AM Matt Turner <mattst88 at gmail.com> wrote:
>> >
>> > On Wed, Dec 19, 2018 at 8:06 AM Ilia Mirkin <imirkin at alum.mit.edu> wrote:
>> > >
>> > > On Wed, Dec 19, 2018 at 1:01 AM Matt Turner <mattst88 at gmail.com> wrote:
>> > > > WTF would you have us do?
>> > >
>> > > Same thing as for any change with an impact this wide --
>> > >
>> > > 1. Identify stakeholders. In this case, probably the sub-project
>> > > maintainers, major contributors, and a smattering of distro
>> > > maintainers.
>> > > 2. Make them happy, or at least get them, as a group, to agree that
>> > > it's "good enough".
>> >
>> > So we're trying to get better coverage than what you're suggesting.
>> >
>> > > 3. Apply.
>> > >
>> > > This is the point at which you can make autotools less visible. We're
>> > > not at that point yet.
>> >
>> > Ilia, it's an extra flag. I think you'll survive.
>>
>> It's an advertising strategy for meson (hello world, check this out,
>> it's going to be the default soon). It can be done at the final stage.
>> We're not at that stage.
>>
>> We're at the stage of "hello community, we'd like to replace
>> autotools", and the community coming back to you with feedback.
>
>
> I disagree.  I think we were at that stage 6-8 months ago and a bunch of the community didn't come back with feedback until we sent a patch to delete autotools.  Identify stakeholders?  Done; the distros are all cool with it or have already switched.  Agree it's "good enough"?  We thought we'd done that and then people raised issues at the 11th hour.  Even with those issues, the ones that are real issues with meson are all in-progress to fix.  The others are just "make it look like autotools so I can pretend meson doesn't exist".  Please pardon my frustration but we thought we'd done our due diligence and it wasn't until we took a step very much like this one that we actually got feedback from certain people such as yourself.  To say that we're only now getting to the "hello community, we'd like to replace autotools" stage is a bit disingenuous.
>
> That said, that doesn't mean I think this patch is the right way to go.  I think the referenced e-mail conversaion has flushed out enough of the remaining issues that we need to fix a couple of remaining things in meson and then just delete autotools.  Maybe this means that autotools stays around for one more release but then I think we should just can it without bothering with the extra deprecation step.

First -- I want you (and Dylan, and others who are pushing this) to
know that I understand your frustration. Making big changes is a giant
pain. Not only is the change itself difficult (aka 80% of the work),
but then you have to herd all the cats to make it all happen. And cats
don't like to be herded (which is why 20% of the work takes 80% of the
time).

Second -- you're not just getting to "hello community" -- you've been
there for a while. But it's not a signpost to move past (like an
announcement might be), it's a stage to complete. The community has to
be happy. You're saying the concerns are last-second, but I've been
seeing these complaints going on for a while (stuff about saving env
vars, inability to see how you configured something, etc). I was
expecting these would all be addressed before I had to have another
look. And then all of a sudden, "let's drop autotools!" and variants
thereof.

What you've been doing thus far is getting all the yaysayers to be
happy (people who are enthusiastic about the change, such as yourself)
-- making sure it all basically works, etc. Now you have to get the
naysayers to be happy, like me, who are pretty happy with the status
quo, and see limited/no benefit in the change. The way you do that is
to make the new system no worse than the old one. Given that some
people are interested, the naysayers aren't going to just shut it
down, but their concerns should be addressed or ruled invalid.

I've outlined what I'm looking for in a replacement to autotools in
another thread [which you've responded to]. I'd recommend reaching out
to some of the other stakeholders directly and getting their take.
(And perhaps privately in case some fear anything public may affect
their jobs, as I think most stakeholders are employed to contribute to
mesa.)

  -ilia