Bug 38127 – MediaTokenizer::writeRawData() shouldn't call finish()

• Status: RESOLVED
• Resolution: INVALID
• Component: Media

[Yong Li]

MediaTokenizer::writeRawData() shouldn't call finish(). FrameLoader will call finish() when loading finishes, and before that, ASSERT(document()->isParsing()) will be hit.

writeRawData() shouldn't call finish() anyway, because more data can come next.

The similar problem exists in PluginTokenizer::writeRawData()

[Yong Li]

Created attachment 54305 [details]
the patch

[Darin Adler]

Comment on attachment 54305 [details]
the patch

Which regression test triggers the assertion failure? If there is no regression test that triggers it, then we should write one. Or we need to explain why it's not possible to write on in the change log.

We do not accept patches to fix bugs without regression tests. If there is some reason a regression test cannot be created, then feel free to add a comment explaining that and put it back up for review.

[Yong Li]

(In reply to comment #2)
> (From update of attachment 54305 [details])
> Which regression test triggers the assertion failure? If there is no regression
> test that triggers it, then we should write one. Or we need to explain why it's
> not possible to write on in the change log.
> 
> We do not accept patches to fix bugs without regression tests. If there is some
> reason a regression test cannot be created, then feel free to add a comment
> explaining that and put it back up for review.

Yeah, sorry for that. just an initial patch. Actually I'm not sure it's a good way to fix it. I hope someone who is familiar with this code can help. I'll create a test case.

[Yong Li]

I'm trying to run webkit in safari with debug config. It always complains "Your  copy of Safari is missing important software resources. Please reinstall Safari."

Darin, any idea?