Extract Chromium-specific parts out of third_party/khronos/GLES2 .

Extract Chromium-specific parts out of third_party/khronos/GLES2 .

BUG=161389
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=174699

[greggman, 2012-12-27 03:31:30]

thanks for doing this. I was on my todo list

Is auto-generating this stuff is really the right thing to do?

1) Some of these extensions may get submitted to khronos
as official extensions at which point they'll end up being 
copied back into gl2ext.h. It seems like it would be easier 
to copy from one file and paste into another and not also have
to edit this code.

2) It seems like auto-generating really only makes sense if 
the data is used to generate stuff in multiple places. This 
is only generating stuff in one place.

How about renaming the current gl2chromium.h to gl2chromium_autogen.h and then
copying and pasting all the stuff deleted from gl2ext.h into gl2chromium.h and
have that include gl2chromium_autogen.h?

[Paweł Hajdan Jr., 2012-12-27 05:47:59]

On 2012/12/27 03:31:30, greggman wrote:
> Is auto-generating this stuff is really the right thing to do?

I think it at least makes sense, but I'd be fine with changing that if
necessary.

> 1) Some of these extensions may get submitted to khronos
> as official extensions at which point they'll end up being 
> copied back into gl2ext.h. It seems like it would be easier 
> to copy from one file and paste into another and not also have
> to edit this code.

Well, the checked-in header is still there. The code generation ensures uniform
generation (i.e. all #ifndef guards around things and other that in fact were
not always present in the previous version of the code).

The snippets to copy can always be taken from the generated file.

> 2) It seems like auto-generating really only makes sense if 
> the data is used to generate stuff in multiple places. This 
> is only generating stuff in one place.

Note that the signatures of functions are re-used (from the .txt file that
build_gles2_cmd_buffer.py reads). This is some advantage...

> How about renaming the current gl2chromium.h to gl2chromium_autogen.h and then
> copying and pasting all the stuff deleted from gl2ext.h into gl2chromium.h and
> have that include gl2chromium_autogen.h?

I'd be fine with that, just please consider above. I think it's better to
auto-generate, to avoid possible mistakes and duplication of function
signatures.

[greggman, 2012-12-27 06:02:55]

On 2012/12/27 05:47:59, Paweł Hajdan Jr. wrote:
> On 2012/12/27 03:31:30, greggman wrote:
> > Is auto-generating this stuff is really the right thing to do?
> 
> I think it at least makes sense, but I'd be fine with changing that if
> necessary.
> 
> > 1) Some of these extensions may get submitted to khronos
> > as official extensions at which point they'll end up being 
> > copied back into gl2ext.h. It seems like it would be easier 
> > to copy from one file and paste into another and not also have
> > to edit this code.
> 
> Well, the checked-in header is still there. The code generation ensures
uniform
> generation (i.e. all #ifndef guards around things and other that in fact were
> not always present in the previous version of the code).
> 
> The snippets to copy can always be taken from the generated file.
> 
> > 2) It seems like auto-generating really only makes sense if 
> > the data is used to generate stuff in multiple places. This 
> > is only generating stuff in one place.
> 
> Note that the signatures of functions are re-used (from the .txt file that
> build_gles2_cmd_buffer.py reads). This is some advantage...
> 
> > How about renaming the current gl2chromium.h to gl2chromium_autogen.h and
then
> > copying and pasting all the stuff deleted from gl2ext.h into gl2chromium.h
and
> > have that include gl2chromium_autogen.h?
> 
> I'd be fine with that, just please consider above. I think it's better to
> auto-generate, to avoid possible mistakes and duplication of function
> signatures.

I don't disagree so much but the same could be said of all C++.

class %(name)s {
 public:
   $(name);
   ~$(name);

 private:
   DISALLOW_COPY_AND_ASSIGN($(name)s);
};

etc..

This seems along the same lines as in not enough replication to justify code
generating?

I'll leave it up to you. 

lgtm