85631 - nsStyleLinkElement::UpdateStylesheet should be aware of active previous loads

Status: RESOLVED WONTFIX

(Core :: DOM: CSS Object Model, defect, P3)

Description

[Peter Van der Beken [:peterv]]

Spin-off from bug 38761. Currently, UpdateStylesheet only removes a previous
stylesheet when it has fully loaded. It should cancel the load of a previous
stylesheet that hasn't. The hard part is knowing if the stylesheet is still
loading, using the StyleSheetLoaded callback doesn't work because the stylesheet
load might have failed. I have a somewhat working patch but it involves some
changes to the CSS loader that I'm not entirely happy about. I'll atach it when
I've cleaned it up.

Comment 1

[Hixie (not reading bugmail)]

If I had a document with two stylesheets, the first being a simple preferred 
stylesheet and the second being an alternate stylesheet that, being generated 
by a CGI script, took 30 seconds to load, and after loading this document I 
switched from the preferred stylesheet to the alternate stylesheet and then 
immediately back to the preferred stylesheet, so that the alternate stylesheet
didn't have time to download before needing to be cancelled, would that trigger
this bug?

If so, I'll go ahead and make a testcase at some point soon.

Comment 2

[Peter Van der Beken [:peterv]]

Hixie: no, this is for example when changing the link element's src attribute
(through the DOM) twice in a row. It'll start loading the stylesheet pointed to
by the first new value of src, and then start loading the stylesheet pointed to
by the second new value of src without stopping the first load. If the first
stylesheet has loaded, then it just replaces and all is good.

styleElement.src = "http://foo/A.css";
styleElement.src = "http://foo/B.css"; -> won't stop the loading of A.

Comment 3

[Hixie (not reading bugmail)]

Ah, ok. Thanks.

Comment 4

[Esben Mose Hansen]

I've created an attachment with a testcase that reprocues the bug (for me at leat)
Build: 2001092803
Mozilla/5.0 (Windows; U; Win 9x 4.90; en-US; rv:0.9.4+) Gecko/20010928The
javascript is encapsuled in a javafunction that is called as a onMouseover on
the body element. Any comments are very welcome, this is my very first bug testcase.

Comment 5

[Esben Mose Hansen]

Attachment: A testcase, three files. The bug is activated by onMousover.

Comment 6

[Heikki Toivonen (remove -bugzilla when emailing directly)]

Removing unneeded qawanted keyword; if you disagree please add it back and state
what you want from QA.

Comment 7

[Boris Zbarsky [:bzbarsky]]

I'm sorta working on this stuff...

Comment 8

[Boris Zbarsky [:bzbarsky]]

I've been thinking about this.  At this point, the bug does not really appear,
since the sheet is hooked up to the element at load start time, not load end
time.  So sheets can no longer race -- the last sheet to start loading is the
one hooked up to the element, period.

Now, I can certainly go and add some code to cancel the load of the previous
sheet.  But is that desirable?  Wouldn't we want to just keep loading it in the
background and have it cached for next time someone wants it?

Comment 9

[Hixie (not reading bugmail)]

Just let it load. Odds are you'll need it again soon anyway.

Comment 10

[Boris Zbarsky [:bzbarsky]]

So in other words, I think the changes I made in bug 107567 fix this bug.

Comment 11

[Boris Zbarsky [:bzbarsky]]

Marking wontfix, per discussion.