1247762 - Dump debug logs from GLContexts on destruction

Status: NEW

(Core :: Graphics: CanvasWebGL, task, P3)

Description

[Jeff Muizelaar [:jrmuizel]]

Mesa seems to leak these messages so while we should fix mesa, printing them is probably more helpful.

Comment 1

[Jeff Muizelaar [:jrmuizel]]

Attachment: Dump debug logs from GLContexts on destruction

Comment 2

[Kelsey Gilbert [:jgilbert]]

Comment on attachment 8718624 [details] [diff] [review]
Dump debug logs from GLContexts on destruction

Review of attachment 8718624 [details] [diff] [review]:
-----------------------------------------------------------------

::: gfx/gl/GLContext.cpp
@@ +2359,5 @@
> +            GLsizei length;
> +
> +            GLuint numFound;
> +            do {
> +                numFound = fGetDebugMessageLog(1, maxMsgLen, &source, &type, &id, &severity, &length, &msgData[0]);

Please leave a comment that includes a link to this bug, to help explain why we're bothering with this. It seems silly otherwise.

Comment 3

[Jeff Muizelaar [:jrmuizel]]

This is not fixing the leak...

Comment 4

[Kelsey Gilbert [:jgilbert]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #3)
> This is not fixing the leak...

Can we patch Mesa? Would that be workable?
Thomas Daede in my office already contributes to Mesa, so I may ask him about this.

Comment 5

[Jeff Muizelaar [:jrmuizel]]

(In reply to Jeff Gilbert [:jgilbert] from comment #4)
> (In reply to Jeff Muizelaar [:jrmuizel] from comment #3)
> > This is not fixing the leak...
> 
> Can we patch Mesa? Would that be workable?
> Thomas Daede in my office already contributes to Mesa, so I may ask him
> about this.

We can but we need to find the leak which I've thus far failed to do. The alternative is to add an address-sanitizer supression.

Comment 6

[Kelsey Gilbert [:jgilbert]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #5)
> (In reply to Jeff Gilbert [:jgilbert] from comment #4)
> > (In reply to Jeff Muizelaar [:jrmuizel] from comment #3)
> > > This is not fixing the leak...
> > 
> > Can we patch Mesa? Would that be workable?
> > Thomas Daede in my office already contributes to Mesa, so I may ask him
> > about this.
> 
> We can but we need to find the leak which I've thus far failed to do. The
> alternative is to add an address-sanitizer supression.

We should do whatever's easiest. We don't really care about the driver leak per se, only that we can run tests against the driver.

Comment 7

[Andrey Bienkowski]

https://bugzilla.mozilla.org/show_bug.cgi?id=1548706