Test for issue 516298

Test for issue 516298

BUG=516298

[philipj_slow, 2015-08-13 12:49:15]

philipj@opera.com changed reviewers:
+ philipj@opera.com

[philipj_slow, 2015-08-13 12:49:16]

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/resources...
File LayoutTests/media/resources/track-dispose-inner.html (right):

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/resources...
LayoutTests/media/resources/track-dispose-inner.html:6: video =
document.getElementsByTagName('video')[0];
Is this deliberately to leave an HTMLCollection dangling? If so it can be made
more explicit, if not it can be querySelector().

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
File LayoutTests/media/track/track-dispose.html (right):

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
LayoutTests/media/track/track-dispose.html:15: }, 150);
What is it that happens on a timeout, is it some cleanup of the iframe, or a GC?
You can trigger async GC, but I'm guessing that's not the magic ingredient?

[fs, 2015-08-13 12:59:40]

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/resources...
File LayoutTests/media/resources/track-dispose-inner.html (right):

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/resources...
LayoutTests/media/resources/track-dispose-inner.html:6: video =
document.getElementsByTagName('video')[0];
On 2015/08/13 12:49:15, philipj wrote:
> Is this deliberately to leave an HTMLCollection dangling? If so it can be made
> more explicit, if not it can be querySelector().

No longer needs to be a HTMLCollection. Changed.

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
File LayoutTests/media/track/track-dispose.html (right):

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
LayoutTests/media/track/track-dispose.html:15: }, 150);
On 2015/08/13 12:49:15, philipj wrote:
> What is it that happens on a timeout, is it some cleanup of the iframe, or a
GC?
> You can trigger async GC, but I'm guessing that's not the magic ingredient?

Yeah, so this is the main gratuitous ugliness - it's "re-run the iframe for
150ms".

[philipj_slow, 2015-08-13 13:05:45]

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
File LayoutTests/media/track/track-dispose.html (right):

https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
LayoutTests/media/track/track-dispose.html:15: }, 150);
On 2015/08/13 12:59:40, fs wrote:
> On 2015/08/13 12:49:15, philipj wrote:
> > What is it that happens on a timeout, is it some cleanup of the iframe, or a
> GC?
> > You can trigger async GC, but I'm guessing that's not the magic ingredient?
> 
> Yeah, so this is the main gratuitous ugliness - it's "re-run the iframe for
> 150ms".

Oh, I see, it'll just reload it over and over again during 150 ms? Seems fine as
long as this in itself does not lead to flakiness, which happened in a somewhat
similar setup in the media-element-enqueue-event-crash.html test using
localStorage to run 5 times.

[fs, 2015-08-13 13:19:43]

On 2015/08/13 13:05:45, philipj wrote:
>
https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
> File LayoutTests/media/track/track-dispose.html (right):
> 
>
https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
> LayoutTests/media/track/track-dispose.html:15: }, 150);
> On 2015/08/13 12:59:40, fs wrote:
> > On 2015/08/13 12:49:15, philipj wrote:
> > > What is it that happens on a timeout, is it some cleanup of the iframe, or
a
> > GC?
> > > You can trigger async GC, but I'm guessing that's not the magic
ingredient?
> > 
> > Yeah, so this is the main gratuitous ugliness - it's "re-run the iframe for
> > 150ms".
> 
> Oh, I see, it'll just reload it over and over again during 150 ms? Seems fine
as
> long as this in itself does not lead to flakiness, which happened in a
somewhat
> similar setup in the media-element-enqueue-event-crash.html test using
> localStorage to run 5 times.

Yeah... Doesn't seem to crash as reliably on the bots as it did locally though,
so probably no good after all, ho-hum...

[philipj_slow, 2015-08-13 13:27:04]

On 2015/08/13 13:19:43, fs wrote:
> On 2015/08/13 13:05:45, philipj wrote:
> >
>
https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
> > File LayoutTests/media/track/track-dispose.html (right):
> > 
> >
>
https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
> > LayoutTests/media/track/track-dispose.html:15: }, 150);
> > On 2015/08/13 12:59:40, fs wrote:
> > > On 2015/08/13 12:49:15, philipj wrote:
> > > > What is it that happens on a timeout, is it some cleanup of the iframe,
or
> a
> > > GC?
> > > > You can trigger async GC, but I'm guessing that's not the magic
> ingredient?
> > > 
> > > Yeah, so this is the main gratuitous ugliness - it's "re-run the iframe
for
> > > 150ms".
> > 
> > Oh, I see, it'll just reload it over and over again during 150 ms? Seems
fine
> as
> > long as this in itself does not lead to flakiness, which happened in a
> somewhat
> > similar setup in the media-element-enqueue-event-crash.html test using
> > localStorage to run 5 times.
> 
> Yeah... Doesn't seem to crash as reliably on the bots as it did locally
though,
> so probably no good after all, ho-hum...

Oh well, just linking to this from the other CL is still something if someone
comes looking for repro steps in a few years.

[fs, 2015-08-13 13:31:37]

On 2015/08/13 13:27:04, philipj wrote:
> On 2015/08/13 13:19:43, fs wrote:
> > On 2015/08/13 13:05:45, philipj wrote:
> > >
> >
>
https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
> > > File LayoutTests/media/track/track-dispose.html (right):
> > > 
> > >
> >
>
https://codereview.chromium.org/1291063002/diff/1/LayoutTests/media/track/tra...
> > > LayoutTests/media/track/track-dispose.html:15: }, 150);
> > > On 2015/08/13 12:59:40, fs wrote:
> > > > On 2015/08/13 12:49:15, philipj wrote:
> > > > > What is it that happens on a timeout, is it some cleanup of the
iframe,
> or
> > a
> > > > GC?
> > > > > You can trigger async GC, but I'm guessing that's not the magic
> > ingredient?
> > > > 
> > > > Yeah, so this is the main gratuitous ugliness - it's "re-run the iframe
> for
> > > > 150ms".
> > > 
> > > Oh, I see, it'll just reload it over and over again during 150 ms? Seems
> fine
> > as
> > > long as this in itself does not lead to flakiness, which happened in a
> > somewhat
> > > similar setup in the media-element-enqueue-event-crash.html test using
> > > localStorage to run 5 times.
> > 
> > Yeah... Doesn't seem to crash as reliably on the bots as it did locally
> though,
> > so probably no good after all, ho-hum...
> 
> Oh well, just linking to this from the other CL is still something if someone
> comes looking for repro steps in a few years.

Yeah, let's hope so.