Bug 64612 – [SKIA] Must make skia's GL context current before drawing in GraphicsContext::clearRect

• Status: RESOLVED
• Resolution: FIXED
• Component: Platform

[Brian Salomon]

Must make skia's GL context current before drawing in GraphicsContext::clearRect

[Brian Salomon]

Created attachment 101017 [details]
Patch

[Brian Salomon]

Created attachment 101019 [details]
Patch

[Brian Salomon]

(In reply to comment #2)
> Created an attachment (id=101019) [details]
> Patch

Second patch fixes the Changelog.

[Stephen White]

Comment on attachment 101019 [details]
Patch

Looks good.

compositing/shadows/shadow-drawing.html looks like it was passing after James's change.  I'm guessing it's only exercised when run with --enable-accelerated-drawing?

[Brian Salomon]

(In reply to comment #4)
> (From update of attachment 101019 [details])
> Looks good.
> 
> compositing/shadows/shadow-drawing.html looks like it was passing after James's change.  I'm guessing it's only exercised when run with --enable-accelerated-drawing?

Oh, yes, that is true.

[WebKit Review Bot]

Comment on attachment 101019 [details]
Patch

Clearing flags on attachment: 101019

Committed r91093: <http://trac.webkit.org/changeset/91093>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Stephen White]

(In reply to comment #5)
> (In reply to comment #4)
> > (From update of attachment 101019 [details] [details])
> > Looks good.
> > 
> > compositing/shadows/shadow-drawing.html looks like it was passing after James's change.  I'm guessing it's only exercised when run with --enable-accelerated-drawing?
> 
> Oh, yes, that is true.

In the future, IWBN to have a test which exercises this case directly.

[Brian Salomon]

(In reply to comment #8)
> (In reply to comment #5)
> > (In reply to comment #4)
> > > (From update of attachment 101019 [details] [details] [details])
> > > Looks good.
> > > 
> > > compositing/shadows/shadow-drawing.html looks like it was passing after James's change.  I'm guessing it's only exercised when run with --enable-accelerated-drawing?
> > 
> > Oh, yes, that is true.
> 
> In the future, IWBN to have a test which exercises this case directly.

There are a bunch of clearRect tests in canvas/philip/tests. I guess they didn't provoke the bug because we lucked into being in the right GL context. Maybe a test having a popup would force us to have two SGC3Ds?

[Stephen White]

(In reply to comment #9)
> (In reply to comment #8)
> > (In reply to comment #5)
> > > (In reply to comment #4)
> > > > (From update of attachment 101019 [details] [details] [details] [details])
> > > > Looks good.
> > > > 
> > > > compositing/shadows/shadow-drawing.html looks like it was passing after James's change.  I'm guessing it's only exercised when run with --enable-accelerated-drawing?
> > > 
> > > Oh, yes, that is true.
> > 
> > In the future, IWBN to have a test which exercises this case directly.
> 
> There are a bunch of clearRect tests in canvas/philip/tests. I guess they didn't provoke the bug because we lucked into being in the right GL context. Maybe a test having a popup would force us to have two SGC3Ds?

Hmm, that could be tricky in DRT.  How about a test where the clear is the first call?  (And clears to something other than transparent black).  Would that leave the graphics context unset?

[Brian Salomon]

(In reply to comment #10)
> (In reply to comment #9)
> > (In reply to comment #8)
> > > (In reply to comment #5)
> > > > (In reply to comment #4)
> > > > > (From update of attachment 101019 [details] [details] [details] [details] [details])
> > > > > Looks good.
> > > > > 
> > > > > compositing/shadows/shadow-drawing.html looks like it was passing after James's change.  I'm guessing it's only exercised when run with --enable-accelerated-drawing?
> > > > 
> > > > Oh, yes, that is true.
> > > 
> > > In the future, IWBN to have a test which exercises this case directly.
> > 
> > There are a bunch of clearRect tests in canvas/philip/tests. I guess they didn't provoke the bug because we lucked into being in the right GL context. Maybe a test having a popup would force us to have two SGC3Ds?
> 
> Hmm, that could be tricky in DRT.  How about a test where the clear is the first call?  (And clears to something other than transparent black).  Would that leave the graphics context unset?

I don't think so because PlatformContextSkia::setSharedGraphicsContext() call makes it current. I'm not really sure how it is getting unset in these failures, though.