[Mesa-dev] [PATCH 00/12] bmake inspired fixes

[Emil Velikov]

On 3 August 2015 at 17:17, Matt Turner <mattst88 at gmail.com> wrote:
> On Fri, Jul 17, 2015 at 11:53 AM, Emil Velikov <emil.l.velikov at gmail.com> wrote:
>> On 17 July 2015 at 19:09, Matt Turner <mattst88 at gmail.com> wrote:
>>> On Fri, Jul 17, 2015 at 10:29 AM, Emil Velikov <emil.l.velikov at gmail.com> wrote:
>>>> Hello all,
>>>>
>>>> A few days ago I realised that BSD make (bmake) is available in the
>>>> Archlinux repos, so I decided to give it a try for drm & mesa.
>>>>
>>>> While the former was working (minus a small patch) mesa is not so lucky.
>>>>
>>>> This series attempts to remove the GNU make idioms, with the first two
>>>> being the base essential for a successful build from tarball.
>>>
>>> ... why should we care about non-GNU make? GNU make has nice features
>>> that we want to use and we use them. I don't see the benefit.
>>>
>> A few reasons:
>>  - It will allow the OpenBSD people to use upstream mesa and devote
>> that time to something more useful ?
>
> Mesa builds on OpenBSD already, as far as I know. The build system
> isn't holding back contributions.
>
They use an in-house bmake compatible system rather. So as they hit a
bug, it's hard to establish if it's due to their build or not. That,
plus the serious rework they need to do in their build, contributes as
to why they're not updating mesa as frequently.
Would be great to spare them those obstacles, even if they choose to
be slightly different ;-)

> I still don't follow how making the build system compatible with
> non-GNU make is beneficial.
>
Let try this from another angle. Even if there is zero benefit, do you
foresee any issues with making it compatible ? Afaics it won't make
anyone's job harder - I/Jonathan will send a quick every so often and
things will just work for everyone. Or maybe there is some subtlety
that I'm missing ?

As mentioned before - there seems to be only one pattern "at fault",
plus it's been addressed with the series.

>>  - Mostly a single pattern/issue/thinko seems to be at fault.
>>  - The rules already look a bit shaky :-)
>
> I don't understand what these mean.
Imho a handful of the Makefiles in src/mapi src/mesa/ are inconsistent
(and confusing) comparing to their dri/glx/egl/gallium counterparts.

The lex/bison/python rules being a good example. With these we provide
explicit info (expand $<) and provide a more consistent look. If they
look harder to read/grasp/etc. just say so and I'll update things
accordingly.

Thanks
Emil