568470 - [HTML5] A page with a sleep() in the script generating it shows no data until after the sleep ends

Status: RESOLVED FIXED

(Core :: DOM: HTML Parser, defect, P2)

Description

[Boris Zbarsky [:bzbarsky]]

Attachment: Testcase-ish

I'm pointing to a url that shows the bug, but since that might go away, also attaching the page in question.  See the "sleep() call goes here" comment in the attachment for where the sleep() should be in the script on the server.

In this case we should render the content before the sleep() ends.

Comment 1

[Henri Sivonen (:hsivonen) (temporarily not reading bugmail)]

Attachment: Flush tree ops more eagerly when stopping speculation

Writing the test case for this was pretty annoying... Still need to check that the timing don't go orange on the tinderbox.

Comment 2

[Ben Newman (:bnewman) (:benjamn)]

Attachment: Suggestion for avoiding setInterval.

These changes make the iframe call parent.parsed() directly instead of requiring the parent to poll the iframe's DOM using setInterval.

If this seems like a reasonable change, you could play with the timings to make the test shorter, but IMO it's short enough already.

r=me with the changes I suggested, or an explanation of why they're invalid.

Comment 3

[Ben Newman (:bnewman) (:benjamn)]

Comment on attachment 447946 [details] [diff] [review]
Flush tree ops more eagerly when stopping speculation

r=me given what I said in comment 2.

Comment 4

[Henri Sivonen (:hsivonen) (temporarily not reading bugmail)]

Thanks. The suggested changes to the test invalidate the test, because the added </script> tag flushes the tree ops unconditionally and the test is trying to test if the ops get flushed without a </script> or EOF forcing a flush.

Comment 5

[Henri Sivonen (:hsivonen) (temporarily not reading bugmail)]

http://hg.mozilla.org/mozilla-central/rev/6060488c2971

Comment 6

[Henri Sivonen (:hsivonen) (temporarily not reading bugmail)]

The test case when perma-orange on Linux tinderboxen, so I first tried to adjust the timings http://hg.mozilla.org/mozilla-central/rev/a9501fcd9c25 and then disabled the test http://hg.mozilla.org/mozilla-central/rev/0e0e2516f04e

It seems this one is hard to test reliably. :-(