Bug 23992 – REGRESSION: crash on windows loading http://www.stickam.com/liveStreams.do

• Status: RESOLVED
• Resolution: FIXED
• Component: Page Loading

[Ojan Vafai]

I only have a debug build of Chromium. So you can see the stack here: http://crbug.com/7682

But a *Windows* webkit nightly seems to crash in the same place. Mac nightlies load fine.

[Ojan Vafai]

Looks like it started with darin's change here: http://trac.webkit.org/changeset/40824

Specifically removing the null check for doc in Frame::contentRenderer() (Frame.cpp:1130)

[Eric Seidel (no email)]

marking p1 since this is a regression.

[Ojan Vafai]

I'm happy to just add the null-check back in. Don't know if that's the right fix though. Will send a patch with the null-check if I don't hear otherwise.

[Ojan Vafai]

Created attachment 27738 [details]
Fix crash on page load of http://www.stickam.com/liveStreams.do. Just readds a null-check that was removed in r40824.

 WebCore/page/Frame.cpp |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

[Alexey Proskuryakov]

Comment on attachment 27738 [details]
Fix crash on page load of http://www.stickam.com/liveStreams.do. Just readds a null-check that was removed in r40824.

r=me

The code change looks fine, but I'm somewhat confused whether this needs a layout test. Did this problem cause crashes in existing tests on Windows? If it did not, can a new test be created for this fix?

[Ojan Vafai]

IRC discussion:
ojan: ap: regarding https://bugs.webkit.org/show_bug.cgi?id=23992...i haven't really been able to come up with a reduced case for the crash. it does crash everytime loading the page in question, but i haven' been able to figure out where exactly. saving a static version of the page and loading it doesn't cause a crash. i can dig further into the stack trace to try and understand when we're hitting this, but i'm not sure how much fruit that will yield.
ap: ojan: ok

I did dig further. Looks like it's crashing in the actual load event. I think (not sure) that an iframe is being appended to the page onload of the top-level window. The contentRenderer is grabbed before the iframe has a document.

[Darin Fisher (:fishd, Google)]

Landed as http://trac.webkit.org/changeset/41066