Bug 11627 – REGRESSION: Reproducible crash at IMDb

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

[Daniele Metilli]

Follow these steps:

1 - Open WebKit TOT (r17825)
2 - Go to this page: http://imdb.com/name/nm0000217/filmorate
3 - Before the page finishes loading, click on the "by type" link on the left
4 - WebKit crashes

This only works if step 3 is done quickly, i.e. before the page is fully loaded.

[Daniele Metilli]

Created attachment 11549 [details]
Crash report

[Alexey Proskuryakov]

I have tried several times, but I couldn't reproduce - FWIW.

[Mark Rowe (bdash)]

I tried for several minutes and was unable to reproduce this with r17993.

[Daniele Metilli]

I can still reproduce it with r17993. Maybe this bug only works with a slow connection.

Another strange thing is that the progress indicator in Safari's address bar goes to the end and then restarts from the beginning, as if it was loading a second page.

[Daniele Metilli]

Important notice: this bug only happens if you have an account at IMDb and are logged in. I discovered it today.

[Joost de Valk (AlthA)]

Reproduced this, i have created an account, logged in and if you click the link in step 3 fast enough, WebKit crashes...

[Mark Rowe (bdash)]

<rdar://problem/4960250>

[Anders Carlsson]

I can't reproduce this with r19407. Daniele, can you try a newer nightly build?

[Daniele Metilli]

I can still reproduce this bug in r19418.

Remember that you have to be registered and logged in.

[Roland Corrigal]

I registered and logged into IMDb with build r19494 and tried several times to reproduce this. Didn't happen. I move that we close this one.

[Daniele Metilli]

But I can still reproduce it in r19494...

[Roland Corrigal]

So, we've gotta figure what's different between mine and yours. If it makes any difference in the world I'm using a PowerPC Mac.

[Roland Corrigal]

Nope not that. I tried to reproduce it on my MacBook Pro at home and still can't. Tried click before and after the page has finished loading. hmm.

[Roland Corrigal]

I couldn't reproduce it with r17825 either.

[Mark Rowe (bdash)]

It's likely that there is a timing component to reproduce this.  For example, a slower network connection may provide more time in which to click the link while the state is such that WebKit will crash.  Listing revisions you cannot reproduce it on isn't all that useful unless you were able to reproduce it initially.  All it indicates is that you couldn't reproduce it (whether due to getting timing wrong, or some other reason), not that the issue has been resolved.

[Roland Corrigal]

Sorry, about the excessive posts, just trying to help. So, I tried a few more times on a dial up account, modem and DSL connection. Still nothing. All I wanted to do was reproduce it and create a reduction. :(

Thanks Mark, I'll take what you said into consideration in any future postings.

[Mark Rowe (bdash)]

Roland, your help is appreciated.  I was just noting that if Daniele can still reproduce the issue across a range of revisions while you are unable to, it makes it unlikely that the version is a relevant factor.  Isolating exactly what the factor is that makes it so hard to reproduce for those of us besides Daniele would still be a very useful thing to do, as it would make fixing this substantially easier.

[Anders Carlsson]

I was finally able to reproduce this - and hopefully fix it!

Committed revision 19931.