460323 - [FIX]"ASSERTION: Bad loading table" involving get.live.com

Status: RESOLVED FIXED

(Core :: CSS Parsing and Computation, defect)

Description

[Jesse Ruderman]

Following up on my promise from bug 438236 ;)

Steps to reproduce:
1. Load http://get.live.com/writer/overview
2. Paste this bookmarklet into the address bar and press enter:

javascript:
function newSS()
{
  var s = document.createElement("link");
  s.setAttributeNS(null, "href", "http://get.live.com/css/WL/common.css");
  s.setAttributeNS(null, "rel", "stylesheet");
  s.setAttributeNS(null, "type", "text/css");
  return s;
}
var head = document.getElementsByTagName("head")[0];
for (var i = 0; i < 10; ++i) void head.appendChild(newSS());

Result:
###!!! ASSERTION: Bad loading table: 'mLoadingDatas.Get(&key, &loadingData) && loadingData == aLoadData', file layout/style/nsCSSLoader.cpp, line 1605

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: This seems to be the testcase

Comment 2

[Boris Zbarsky [:bzbarsky]]

Attachment: Fix

Comment 3

[Boris Zbarsky [:bzbarsky]]

Attachment: Same as diff -w

If the already-loaded sheet is modified, we need to fall back on looking at loading sheets, not just creating a new sheet.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Pushed changeset 457619abbaa5.  Jesse, any idea how to sanely turn this into a crashtest?  The hard part is knowing when to stop the test.... :(

Comment 5

[Jesse Ruderman]

If you use setTimeout at the end of doIt, are you guaranteed that the stylesheets all load before the timeout fires, thanks to using data: URLs?  Or can you use the DOM styleSheet API to tell whether the stylesheets have all loaded, and even test that the second and third stylesheets weren't modified by the call to change the first one?

Comment 6

[Boris Zbarsky [:bzbarsky]]

No guarantees on data:.  And DOM stylesheet doesn't provide "has this sheet loaded?" info, really.

I suspect the test will have to wait on bug 185236.

Comment 7

[Jesse Ruderman]

You can use the DOM Stylesheet API to see whether all the rules are there, but I guess that's not quite the same thing as knowing whether it has finished loading.

Comment 8

[Boris Zbarsky [:bzbarsky]]

http://hg.mozilla.org/integration/mozilla-inbound/rev/b045aa0216fb

Comment 9

[Michael Wu [:mwu]]

https://hg.mozilla.org/mozilla-central/rev/b045aa0216fb

Comment 10

[Boris Zbarsky [:bzbarsky]]

Nah, this was fixed for like Firefox 3.5 or something.  That last checkin was just a test.