[Mesa-dev] [PATCH 0/2] Disable the EGL state tracker for Linux/DRI builds

[Emil Velikov]

On 05/11/14 21:11, Jose Fonseca wrote:
>> How many people/companies use EGL for Windows/fbdev, how about OpenVG on
> any platform ?
> 
> I already said this privately to Marek when he was RFC'ing on this change: I'm fine if Linux-specific drivers abandon st/egl to focus solely on st/dri, but removing st/egl altogether seems unnecessary and short-sighted: EGL is a cross-platform API, Mesa is a cross-platform implementation of OpenGL and friends, so sooner or later people will want to have Mesa's EGL support on platforms others than Linux.
> 
> This is not hypothetical:
> - See https://bugs.freedesktop.org/show_bug.cgi?id=40920 for an example of a bug reported from an user using llvmpipe + egl + opengv on windows.
> - VMware doesn't currently ship or support EGL on Windows, but I suspect we eventually we'll want to support EGL on non-linux platforms.
> 
> Even if OpenVG is loosing popularity, but maybe Khronos will come up with another cross-platform graphics API (maybe OpenGL NG) that's tied to EGL.
> 
> So a cross-platform implementation of EGL is bound to be useful.
> 
> 
> I don't test, but I build egl-static and OpenVG on Windows nightly w/ llvmpipe.  It's like a superset of OSMesa, and it seems more useful, as it gives one more APIs than OSMesa, and through a standard API to create/bind contexts .
> 
> 
> In short, stop caring about st/egl on Linux, maybe even remove DRI support out st/egl if you must, but please don't go out of your way to break EGL on non-linux platforms.
> 
So let me justify why I brought this in the first place:

1. Over the last two years st/egl had the following patches
 * Build fixes & related - most of the patches (80%?)
 * Interface changes - ~10 patches
 * Bugfixes - ~3 patches
 * New "features" - 1 patch (already present with the dri2 backend)
2. Over the last two years I've not seen any bug reports from people
using either st/egl or st/vega. Must admit I was not looking too closely.
3. Afaict the VMWare or other commercial products do not use it.

So based on those my naive question was "Is there anyone actually using
those state-trackers, rather than just building them" - i.e. it was
meant as a question, rather than a message of hate wrt the code-base :-)

I must admit I cannot predict the future (i.e. what VMWare, Khronos
and/or others have in plan) but based on the lack of testers,
maintainers and new improvements imho it make sense to remove the stale
code. As soon as any of that changes we can always bring it back.

So I would not call it short-sighted, but imho it does not make sense to
cling onto something in the hopes that one day someone may use it.

Cheers,
Emil