[Mesa-dev] [PATCH 1/2] configure.ac: enable building GLES1 and GLES2 by default

Emil Velikov emil.l.velikov at gmail.com
Sat May 30 05:30:11 PDT 2015

On 28 May 2015 at 21:48, Matt Turner <mattst88 at gmail.com> wrote:
> On Thu, May 28, 2015 at 5:07 AM, Emil Velikov <emil.l.velikov at gmail.com> wrote:
>> On 27/05/15 16:59, Matt Turner wrote:
>>> On Wed, May 27, 2015 at 4:53 AM, Emil Velikov <emil.l.velikov at gmail.com> wrote:
>>>> On 27 May 2015 at 11:23, Dave Airlie <airlied at gmail.com> wrote:
>>>>>> Wow, I hadn't expected such a hateful comment on GLES1.
>>>>>> Does anyone else want to convince me that GLES1 should burn in hell?
>>>>> So I dug around,
>>>>> commit 4c06853833996d990eb76b195ca5d6838c6f3d6b
>>>>> Author: Adam Jackson <ajax at redhat.com>
>>>>> Date:   Wed May 8 18:03:21 2013 -0400
>>>>>     Switch to Mesa master (pre 9.2)
>>>>>     - Fix llvmpipe on big-endian and enable llvmpipe everywhere
>>>>>     - Build vdpau drivers for r600/radeonsi/nouveau
>>>>>     - Enable hardware floating-point texture support
>>>>>     - Drop GLESv1, nothing's using it, let's not start
>>>>> So at least in Fedora 2 years ago, we realised there was no GLES1
>>>>> users in the distro,
>>>>> and we didn't want to encourage any.
>>>>> I suppose some users might exist outside the classic Linux distro
>>>>> world. i.e. android, embedded land.
>>>> At least three other distros have GLES1 - Arch, Debian and OpenSUSE.
>>>> So imho one should just leave the decision to depreciate/kill it off
>>>> to the distros ?
>>> No one is suggesting deleting GLES1.
>> Ack - I'm aware of that. I'm not sure which part of my statement led
>> came out as the opposite. If you can point me out as such that'll be
>> appreciated.
> The part I quoted specifically -- "So imho one should just leave the
> decision to depreciate/kill it off to the distros ?"
> Since the thread is about defaults... I have no idea what you mean.
Indeed my wording there wasn't quite right. Dully noted.

>>> I think distributions are going to inevitably ship things they have no
>>> reason to, regardless of our defaults, often because they don't know.
>>> The best we can do is guide their hands. If GLES1 is unused, we should
>>> note that in configure.ac to better communicate it to the
>>> distributions.
>> No objection on the hand-holding, but the idea that Mesa knows which
>> distro does not ship/have GLES1 compatible software sounds a bit
>> strange. So I was pointing out that some (be that by mistake or not)
>> still have/use it.
> Let me give an example from my experience as a Gentoo developer:
> Cairo has a number of backends -- Image, Skia, Xlib, PDF, SVG, OpenGL,
> GLESv2, DirectFB, OpenVG, DRM, Cogl, etc. Users don't have to
> understand what any of these are in order to want them, and they'll
> file bugs requesting we enable them [1]. Unless distribution
> maintainers are aware that the features are not useful, they'll likely
> just turn them on.
> Cairo recently added some text to configure.ac to tell users that the
> OpenGL and GLESv2 backends are experimental:
> --- The OpenGL surface backend feature is still under active development and
> --- is included in this release only as a preview. It does NOT fully work yet
> --- and incompatible changes may yet be made to OpenGL surface backend
> --- specific API.
I believe that such warnings are a great idea, apart from leaving the
option disabled by default.

>> At the end of the day, do as you please. The argument of "distro X
>> defaults to Y, so mesa should do the same", just seems strange to me.
> I don't think anyone is making that argument either...
Dave's initial reply come out precisely as such. Maybe I misread it ?

>> Similar to Marek, I do wonder about the hostility towards GLES1. Sure it
>> is a bit old, and the API is not the most flexible/useful but there are
>> dozens (hundreds?) of projects that fit the criteria :)
> You seem to be making the argument that we should build GLESv1 by
> default because distributions ship it. I'm claiming that distributions
> shipping it are doing so without understanding that there are zero
> known uses of GLESv1 on desktop Linux.
mesa-demos and piglit has some GLES1 "uses" ;-)

But seriously, imho a feature should be enabled by default unless 1)
its build requirements are not met or 2) is known (severely) broken,
incomplete, experimental or considered "a toy". Distributions will
likely continue to ship/do things as originally regardless of changes
like this one.

Hope that gives some insight and I'm not the only one that thinks so.
But if people feel otherwise so be it - I'm not going to pick on this
topic any more.


More information about the mesa-dev mailing list