263672 - Load of UTF-8 encoded file takes forever

Status: RESOLVED WORKSFORME

(Core Graveyard :: GFX: Gtk, defect)

Description

[Peter Kovář]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; sk-SK; rv:1.8a5) Gecko/20041008
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; sk-SK; rv:1.8a5) Gecko/20041008

Open the following test file and have a lot of patience.

Reproducible: Always
Steps to Reproduce:
1. Download test.utf8.bz2
2. bunzip2 test.utf8.bz2
3. File -> Open test.utf8

Actual Results:  
Nothing, just neverending wait. The same with Firefox.



Opera crashed X server immediately!

Comment 1

[Peter Kovář]

Attachment: bzip2 compressed test.utf8

bunzip2 test.utf8.bz2

Comment 2

[Peter Kovář]

Attachment: UTF-8 degenerator

cc UTF-8.c -o UTF-8

./UTF-8 0x20 0xFFFF > test.utf8

Comment 3

[Boris Zbarsky [:bzbarsky]]

I see us spending forever in nsFontMetricsGTK::GetTextDimensions and
nsFontMetricsGTK::FindFont on this testcase, presumably because this file covers
all of UTF-8 and hence we have to look for all chars, and most fonts don't have
most chars, which means that the time taken to render the page is O(# of chars
times # of fonts installed on the machine).

Per reporter's comments in other bugs, he is using a GTK2/XFT build; I have no
idea what performance looks like for those (that is, where the time is spent).

Comment 4

[Mats Palmgren (inactive)]

WFM, Firefox trunk on Ubuntu-feisty.

Can anyone else reproduce this bug in a recent trunk build?
http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-trunk/

Comment 5

[[:Aleksej]]

Debian 4.0 with all the free Unicode fonts installed and some third-party fonts like MathML ones.

Confirmed with Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a6pre) Gecko/20070630 Minefield/3.0a6pre hung with 100% CPU usage, and was still hung after 4 minutes.

With Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.5pre) Gecko/20070630 BonEcho/2.0.0.5pre, the loading is complete after something like 17 seconds.

Comment 6

[[:Aleksej]]

(In reply to comment #5)
PS: according to fc-list, that's 462 fonts.

Comment 7

[Zack Weinberg (:zwol)]

Loads in a reasonable amount of time for such a large file (less than five seconds) with Firefox 8.

Comment 8

[[:Aleksej]]

Can’t reproduce (takes 10-20 seconds) with Firefox 8.0 or 3.6.14. Note that, when testing, you have to make sure the file is opened as UTF-8, not e.g. ISO-8859-1. Of course, the computer and Debian version are different now.