Bug 67006 – [chromium] Add MockWebGraphicsContext3D, for compositor unit testing

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[Iain Merrick]

[chromium] Add MockWebGraphicsContext3D, for compositor unit testing

[Iain Merrick]

Created attachment 105288 [details]
Patch

[Nat Duca]

Comment on attachment 105288 [details]
Patch

I have a high level question for @kbr, though, which is whether this is the right layer to hook at. I just can't make my mind up, personally.

Clearly, its the only place we can hook right now because everything higher is non-virtual.

However, this means that all our testing becomes specific to the Chromium port.

Is this OK? I guess some of this ties back to me not understanding WebKits' overal position on unit tests...

[Kenneth Russell]

(In reply to comment #2)
> (From update of attachment 105288 [details])
> I have a high level question for @kbr, though, which is whether this is the right layer to hook at. I just can't make my mind up, personally.
> 
> Clearly, its the only place we can hook right now because everything higher is non-virtual.
> 
> However, this means that all our testing becomes specific to the Chromium port.
> 
> Is this OK? I guess some of this ties back to me not understanding WebKits' overal position on unit tests...

I think this is fine for the time being. There's been discussion in recent months about settling on a unit testing framework for the overall WebKit project -- see https://lists.webkit.org/pipermail/webkit-dev/2011-April/016510.html -- but so far everyone's been too busy to push it forward. There was some consensus at the last WebKit meeting in April that the project would try to standardize on gtest, which is full featured and well maintained, so adding a gtest-compatible mock and tests sounds good even if they're currently Chromium specific. They could be made port-independent in the future.

[Kenneth Russell]

Comment on attachment 105288 [details]
Patch

This generally looks nice. Hopefully this will allow us to eventually catch some errors earlier. I am concerned though that adding new methods to WebGraphicsContext3D (which happens reasonably frequently) will cause the WebKit unit tests to break and developers won't catch that until fairly late. Do the Chromium EWS bots build and run the webkit unit tests?

[Nat Duca]

Hmm, the builders do, but I'm not sure how they relate to EWS... +abarth :)

[Nat Duca]

@kbr, can you do a formal review of this? I'm going to pick up Iain's patch and try to land it before I land 66807.

[Kenneth Russell]

Comment on attachment 105288 [details]
Patch

OK. If we find that we're breaking things frequently on the canaries we can change what gets compiled in which targets. r=me

[WebKit Review Bot]

Comment on attachment 105288 [details]
Patch

Clearing flags on attachment: 105288

Committed r93928: <http://trac.webkit.org/changeset/93928>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.