update WebGL to initial release 1.0.0

- updated tarball with newest test suite
- created new patch
- transfer more detailed information about failures from browser to autotest
- waive failures that are not yet fixed in Chrome
- transfer text summary from test

BUG=chromium-os:14112
TEST=

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=d92193e

[ilja, 2011-04-14 02:55:13]

Inviting first feedback.

[zbehan, 2011-04-14 23:46:53]

First piece of feedback: I'd generally welcome if we were commiting less binary
tarballs into the git repo. The autotest has been setup to download it off a gs
bucket, I've also been pondering a solution where we could reuse gentoo
DISTFILES to properly mirror, store, and hash the tarballs, rather than commit
them.

The issue of pushing them into git is obvious, everyone who does a clean clone
of the git repo will be pulling all past tarballs of everything. This may be
tolerable with tiny tarballs or rarely updated tarballs, but eventually always
gets cumbersome due to O(n^2) space requirements.

On 2011/04/14 02:55:13, ihf wrote:
> Inviting first feedback.

[ihf, 2011-04-15 00:16:10]

That is a fair observation. Now I did ask you yesterday how to put my new
tarball so it is accessible at
http://commondatastorage.googleapis.com/chromeos-localmirror/distfiles/ +
tarball,

I have been unable to place it there myself. It seems the script is trying
to do this, but I don't understand
which arguments it expects. If anybody can give me instructions I will of
course put it there.


Ilja.

On Thu, Apr 14, 2011 at 4:46 PM, <zbehan@chromium.org> wrote:

> First piece of feedback: I'd generally welcome if we were commiting less
> binary
> tarballs into the git repo. The autotest has been setup to download it off
> a gs
> bucket, I've also been pondering a solution where we could reuse gentoo
> DISTFILES to properly mirror, store, and hash the tarballs, rather than
> commit
> them.
>
> The issue of pushing them into git is obvious, everyone who does a clean
> clone
> of the git repo will be pulling all past tarballs of everything. This may
> be
> tolerable with tiny tarballs or rarely updated tarballs, but eventually
> always
> gets cumbersome due to O(n^2) space requirements.
>
> On 2011/04/14 02:55:13, ihf wrote:
>
>> Inviting first feedback.
>>
>
>
>
> http://codereview.chromium.org/6849012/
>

[ilja, 2011-04-17 22:42:09]

Moved tarball as suggested and adjusted the code. Anything else?

On 2011/04/15 00:16:10, ihf_google.com wrote:
> That is a fair observation. Now I did ask you yesterday how to put my new
> tarball so it is accessible at
> http://commondatastorage.googleapis.com/chromeos-localmirror/distfiles/ +
> tarball,
> 
> I have been unable to place it there myself. It seems the script is trying
> to do this, but I don't understand
> which arguments it expects. If anybody can give me instructions I will of
> course put it there.
> 
> 
> Ilja.
> 
> On Thu, Apr 14, 2011 at 4:46 PM, <mailto:zbehan@chromium.org> wrote:
> 
> > First piece of feedback: I'd generally welcome if we were commiting less
> > binary
> > tarballs into the git repo. The autotest has been setup to download it off
> > a gs
> > bucket, I've also been pondering a solution where we could reuse gentoo
> > DISTFILES to properly mirror, store, and hash the tarballs, rather than
> > commit
> > them.
> >
> > The issue of pushing them into git is obvious, everyone who does a clean
> > clone
> > of the git repo will be pulling all past tarballs of everything. This may
> > be
> > tolerable with tiny tarballs or rarely updated tarballs, but eventually
> > always
> > gets cumbersome due to O(n^2) space requirements.
> >
> > On 2011/04/14 02:55:13, ihf wrote:
> >
> >> Inviting first feedback.
> >>
> >
> >
> >
> > http://codereview.chromium.org/6849012/
> >

[zbehan, 2011-04-17 22:57:17]

Much better from my point of view: LGTM , though since this really looks
more like a new test, rather than improved old test, and no piece of code
remained untouched, the original author should perhaps have a look and say.
(+zmo)

On Mon, Apr 18, 2011 at 12:42 AM, <ihf@chromium.org> wrote:

> Moved tarball as suggested and adjusted the code. Anything else?
>
>
> On 2011/04/15 00:16:10, ihf_google.com wrote:
>
>> That is a fair observation. Now I did ask you yesterday how to put my new
>> tarball so it is accessible at
>> http://commondatastorage.googleapis.com/chromeos-localmirror/distfiles/ +
>> tarball,
>>
>
>  I have been unable to place it there myself. It seems the script is trying
>> to do this, but I don't understand
>> which arguments it expects. If anybody can give me instructions I will of
>> course put it there.
>>
>
>
>  Ilja.
>>
>
>  On Thu, Apr 14, 2011 at 4:46 PM, <mailto:zbehan@chromium.org> wrote:
>>
>
>  > First piece of feedback: I'd generally welcome if we were commiting less
>> > binary
>> > tarballs into the git repo. The autotest has been setup to download it
>> off
>> > a gs
>> > bucket, I've also been pondering a solution where we could reuse gentoo
>> > DISTFILES to properly mirror, store, and hash the tarballs, rather than
>> > commit
>> > them.
>> >
>> > The issue of pushing them into git is obvious, everyone who does a clean
>> > clone
>> > of the git repo will be pulling all past tarballs of everything. This
>> may
>> > be
>> > tolerable with tiny tarballs or rarely updated tarballs, but eventually
>> > always
>> > gets cumbersome due to O(n^2) space requirements.
>> >
>> > On 2011/04/14 02:55:13, ihf wrote:
>> >
>> >> Inviting first feedback.
>> >>
>> >
>> >
>> >
>> > http://codereview.chromium.org/6849012/
>> >
>>
>
>
>
> http://codereview.chromium.org/6849012/
>

[zmo, 2011-04-18 16:48:06]

The change looks fine.  One thing I want to mention is that the current 1.0.0
suite still has some bugs (we fixed them on the main trunk), so you might want
to make another shot soon.

[ilja, 2011-04-18 18:44:36]

I will certainly update the suite once you let me know there is a new version.

That said there would IMHO a great benefit of running this and similar WebGL
tests already in Chromium before pulling the browser to the OS which happens
only once a day. Is there anything active/planned right now?

On 2011/04/18 16:48:06, zmo wrote:
> The change looks fine.  One thing I want to mention is that the current 1.0.0
> suite still has some bugs (we fixed them on the main trunk), so you might want
> to make another shot soon.

[zmo, 2011-04-18 19:07:18]

Right now these webgl conformance tests are running as LayoutTests, so
it's running with --in-process-webgl and on top of mesa library.  I am
planning to run them as Browser tests on top of NVIDIA gpu as part of
the gpu tests.

On Mon, Apr 18, 2011 at 11:44 AM,  <ihf@chromium.org> wrote:
> I will certainly update the suite once you let me know there is a new
> version.
>
> That said there would IMHO a great benefit of running this and similar WebGL
> tests already in Chromium before pulling the browser to the OS which happens
> only once a day. Is there anything active/planned right now?
>
> On 2011/04/18 16:48:06, zmo wrote:
>>
>> The change looks fine.  One thing I want to mention is that the current
>> 1.0.0
>> suite still has some bugs (we fixed them on the main trunk), so you might
>> want
>> to make another shot soon.
>
>
>
> http://codereview.chromium.org/6849012/
>