Remove headers for CL, EGL, GLES2, KHR, VG

Remove headers for CL, EGL, GLES2, KHR, VG

In chromium we should never use these headers from mesa. We use the
OpenGL ES 2.0 and KHR headers from khronos. Having duplicates in the
tree is confusing and a recipe for disaster if the two ever diverge and
different versions are used by different libraries, especially since they
all use the same include guard so which version is seen by a particular
target depends on the include path order.

[jamesr, 2014-08-15 22:34:20]

I don't know how to actually land this patch but it appears to compile cleanly
for me.

[Ken Russell (switch to Gerrit), 2014-08-15 22:44:45]

LGTM if this compiles and runs tests. Did you build the osmesa target and make
sure that layout tests continue to work with --use-gl=osmesa?

The best way I know of to land this patch:
 - Check out a fresh (svn) workspace of /trunk/deps/third_party/mesa
 - Make the changes there and commit them
 - Roll the version of Mesa in DEPS

Let me know if you have any trouble.

[piman, 2014-08-15 22:46:49]

LGTM, worth adding mention of it in README.chromium with the DEPS roll.

[jamesr, 2014-08-15 22:52:35]

On 2014/08/15 22:44:45, Ken Russell wrote:
> LGTM if this compiles and runs tests. Did you build the osmesa target and make
> sure that layout tests continue to work with --use-gl=osmesa?

I did a compile of 'all' on linux and it worked.  I'll try to figure out how to
throw this at the trybots.  I think the layout tests use osmesa by default,
right?  I'll confirm those continue to pass.

> 
> The best way I know of to land this patch:
>  - Check out a fresh (svn) workspace of /trunk/deps/third_party/mesa
>  - Make the changes there and commit them

commit via svn?

>  - Roll the version of Mesa in DEPS

OK.  I'll include a README.chromium update 

> 
> Let me know if you have any trouble.

[Ken Russell (switch to Gerrit), 2014-08-15 23:04:13]

On 2014/08/15 22:52:35, jamesr wrote:
> On 2014/08/15 22:44:45, Ken Russell wrote:
> > LGTM if this compiles and runs tests. Did you build the osmesa target and
make
> > sure that layout tests continue to work with --use-gl=osmesa?
> 
> I did a compile of 'all' on linux and it worked.  I'll try to figure out how
to
> throw this at the trybots.  I think the layout tests use osmesa by default,
> right?  I'll confirm those continue to pass.

It'll be quite difficult, I think, without actually committing this. If you've
built "osmesa" explicitly and verify locally that something -- content_shell,
anything -- works as well as it did before with --use-gl=osmesa, just go ahead.
(I just tried running my locally built Chromium on Mac with --use-gl=osmesa and
it didn't have any GPU support, so I think the layout tests are the best way to
verify this change works.)

> > 
> > The best way I know of to land this patch:
> >  - Check out a fresh (svn) workspace of /trunk/deps/third_party/mesa
> >  - Make the changes there and commit them
> 
> commit via svn?

Yes. I don't remember how code reviews work with the deps/third_party/mesa
workspace. I think gcl will auto-upload them to codereview.appspot.com. I don't
know how to set up git-svn for that workspace, but if you figure out how to do
so, feel free to git cl dcommit.

[jamesr, 2014-08-15 23:10:29]

On 2014/08/15 23:04:13, Ken Russell wrote:
> On 2014/08/15 22:52:35, jamesr wrote:
> > On 2014/08/15 22:44:45, Ken Russell wrote:
> > > LGTM if this compiles and runs tests. Did you build the osmesa target and
> make
> > > sure that layout tests continue to work with --use-gl=osmesa?
> > 
> > I did a compile of 'all' on linux and it worked.  I'll try to figure out how
> to
> > throw this at the trybots.  I think the layout tests use osmesa by default,
> > right?  I'll confirm those continue to pass.
> 
> It'll be quite difficult, I think, without actually committing this. If you've
> built "osmesa" explicitly and verify locally that something -- content_shell,
> anything -- works as well as it did before with --use-gl=osmesa, just go
ahead.
> (I just tried running my locally built Chromium on Mac with --use-gl=osmesa
and
> it didn't have any GPU support, so I think the layout tests are the best way
to
> verify this change works.)

The poster circle renders for me (choppily) with --use-gl=osmesa in
content_shell.

[Ken Russell (switch to Gerrit), 2014-08-15 23:13:26]

On 2014/08/15 23:10:29, jamesr wrote:
> On 2014/08/15 23:04:13, Ken Russell wrote:
> > On 2014/08/15 22:52:35, jamesr wrote:
> > > On 2014/08/15 22:44:45, Ken Russell wrote:
> > > > LGTM if this compiles and runs tests. Did you build the osmesa target
and
> > make
> > > > sure that layout tests continue to work with --use-gl=osmesa?
> > > 
> > > I did a compile of 'all' on linux and it worked.  I'll try to figure out
how
> > to
> > > throw this at the trybots.  I think the layout tests use osmesa by
default,
> > > right?  I'll confirm those continue to pass.
> > 
> > It'll be quite difficult, I think, without actually committing this. If
you've
> > built "osmesa" explicitly and verify locally that something --
content_shell,
> > anything -- works as well as it did before with --use-gl=osmesa, just go
> ahead.
> > (I just tried running my locally built Chromium on Mac with --use-gl=osmesa
> and
> > it didn't have any GPU support, so I think the layout tests are the best way
> to
> > verify this change works.)
> 
> The poster circle renders for me (choppily) with --use-gl=osmesa in
> content_shell.

Great. Then just commit the change to the Mesa workspace directly, but please
send the DEPS roll through the CQ.