43648 - Browser very slow when loading _very_ large page (was a crash at one point in time)

Status: VERIFIED WORKSFORME

(Core :: DOM: HTML Parser, defect, P2)

Description

[Dave]

Overview Description: 

The above page takes at least a minute to load under NS 4.7 and never finishes
loading under Mozilla because the browser crashes. The html on it appears to be
fairly simple so I'm assuming it's the sheer volume of it that's causing the
problem.

This bug also sounds a little like bug 42138 but I don't think the behaviour is
quite the same... my browser didn't render the page at all with the test case
for 42138.


Steps to Reproduce: 
1) Go to the above URL
2) Wait for the browser to crash (may take quite a while).

Reproducibility: 

always

Build Date & Platform Bug Found: 

2000062220

Additional Builds and Platforms Tested On: 

works on NS 4.7

Comment 1

[R.K.Aa.]

*** This bug has been marked as a duplicate of 21637 ***

Comment 2

[Andreas Franke (gone)]

Well, it seems to me that this page crashes due a to stack overflow (tested on
PC/Linux, build 2000062220). Don't know if that is covered by bug 21637.
Ok, for the user it doesn't matter much whether its a deadlock/hang, an infinite
loop, or a crash... 

Comment 3

[R.K.Aa.]

Reason i tagged this dup is i would have reported the same site as here a bug,
hadn't it been for finding 21637 with what i assume is the same problem.

In bug 25509 it's suggested using bug 21637 as track bug for issues with text
layout performance. But you're right..no crash mentioned there.

Testing some of the sites tagged as dups of 21637 i find they DO load now. Not
quick, but eventually they come around.

Testing the url here again with Linux 2000-062321 it indeed crashes. Reopening.

Comment 4

[R.K.Aa.]

Confirming & raising severity a notch.

Comment 5

[Blake Ross]

Adding crash keyword to all open crashers.

Comment 6

[rickg]

Off to buster due to stack overflow

Comment 7

[Marc Attinasi]

I cannot get to the URL to check if it is still crashing. Is there an updated 
URL?

Comment 8

[Dave]

I don't know of any. I just retested a couple days ago though and the error
looks temporary... the page should be up again before too long.

Comment 9

[Dave]

The error you just got connecting to bugs.gnome.org looks temporary, that is :)

Comment 10

[buster]

doesn't crash for me on NT, will test on Linux.  does have poor performance.

Comment 11

[buster]

just tested on our linux box, no crash there either.

Comment 12

[buster]

unless someone tells me that this is crashing with a very recent build, I'm 
going to remove the crash keyword, reduce the severity, and treat it as a 
performance bug, as waterson suggests in bug 29641.  Chris checked in a fix-ish 
kinda thing on 8/30, so please report your results with builds after 9/1 only.

We could really use a quantify log for this page.  Rick, you got the new 
quantify up and running yet?  I'll try tomorrow if no one else gets there first.

Comment 13

[Chris Waterson]

If I had to guess, I'd guess it would be something to do with list renumbering 
(even though it's a <ul>). I don't see any unclosed inline tags...

Comment 14

[rickg]

I'm just now getting a build on win2k. It should be done in 30 minutes, and then 
I can try to run the new quantify. I'll keep you posted.

Comment 15

[buster]

I ran the new WinNT Quantify on this page, and after 2 hours the page still 
hadn't finished loading.  I gave up.  I'm going to build a subset of the 
testcase, and try quantifying that.

If your machine isn't needed for the rest of the day, you could still try it on 
the full test case on Win2k.

Thanks

Comment 16

[buster]

I'm going to mark this WorksForMe.  On my windows NT box (which showed the same
performance problems as linux, at least a few weeks ago), the preformance on
this page is on par with Nav4, whether loaded from net or from local file.  I
have not seen a crash.  If you disagree, please re-open and be very specific
about what you see and what kind of hardware and net connection you are testing
with.  Thanks!

Comment 17

[bsharma]

updated qa contact.

Comment 18

[bsharma]

Verified on:
build: 2001-04-16-06-Mtrunk
platform: WinNT

The above url cannot be reached.

Comment 19

[Chris Waterson]

So you re-open the bug? There is no useful information on how to reproduce it. 
Closing again.

Comment 20

[Christopher Hoess (gone)]

bugs.gnome.org seems to be gone.  Verifying.