Bug 7189 – XSLT result parsing stops prematurely

• Component: XML
• Status: VERIFIED FIXED

Summary: WebKit stops tokenizing and parsing an HTML document created by the XSLT processor prematurely, resulting in only the first part of the document showing.

This is sometimes reproducible with PLT results such as in the URL (depending on network speed and caching). It seems to be easier to reproduce with a local copy from disk. When the problem occurs, the results are truncated at some point and the legend (normally on the right) doesn't appear.

The problem is that XSLTProcessorImpl::createDocumentFromSource calls setParsing(false) on the resulting document right after write(), when it might have not parsed everything yet. Normally it's the frame's that calls setParsing(false) from its signal handler, but for XSLT-generated documents it doesn't.

I have a patch that removes the setParsing(false) call from XSLTProcessorImpl and keep the frame wired to the document so that it will call setParsing(false) correctly instead.

Created attachment 6410 [details]
Testcase

If your computer isn't too fast, the square will remain red because parsing will stop before reaching the script that turns the square green.

Created attachment 6411 [details]
Proposed patch

I don't know how to make a test for this. I think I need a way to preempt the tokenizer that's not timer-based (and hence CPU-speed-dependent). No ChangeLog entry yet either.

[Darin Adler]

Comment on attachment 6411 [details]
Proposed patch

Looks great, r=me. I see the difficulty in making a reliable test for this, and I suggest we land it even if we can't figure out a way to make a more-reliable test.

Verified in r12770 nightly