Closed Bug 123034 Opened 24 years ago Closed 14 years ago

Xfree86 hogs CPU when certain characters encountered

Categories

(Core :: Layout: Text and Fonts, defect)

x86
Linux
defect
Not set
major

Tracking

()

RESOLVED INCOMPLETE

People

(Reporter: mal0rd, Unassigned)

References

()

Details

(Keywords: perf)

Attachments

(1 file)

From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.7) Gecko/20020112 BuildID: 20020112 I don't know why yet, but Apache is sending a bunch of wierd (>128) characters and they crash mozilla. I used a trial and error process to reduce the the resultant to file to as small as possible. I don't trust cutting and pasting the characters (wouldn't want to crash other's when trying to view the bug) so here are the ASCII values I got from glimmer: -38,-70,-39,-115 And here are the hex from khexedit: da, ba, d9, 8b, 0a They need to come in order, but in any part of the file it makes the XFree86 process hog all the CPU and very very nearly make the computer unusable until mozilla is killed (with -KILL). I realize these are most definitally illegal characters, but it shouldn't make mozilla crash. Reproducible: Always Steps to Reproduce: Open page with those characters Actual Results: Dead in the water. Expected Results: Mozilla should have said: "Excuse me sir, but I don't think that page is valid. Please fix it." Mozilla gives no error message, nor does X. I will now proceed to fix the source of the problem.
If you want to see if this crashes your build, then have an xterm availible with "killall -9 mozilla-bin" already typed in. After mozilla freezes, you'll have about 1 second every 20 to do something. Use it to press ENTER when the xterm has focus.
Testcase loads fine for me, XFree86 4.0.1, using xfs, mozilla build 2002-02-01 morning on Linux.
Worksforme with 1-31-2002 build and also 1-14-2002 build. XFree86-4.1.0-15 with xfs under RH linux.
With 2002012814 I am experiencing a similar behavior. Instead of freezing indefinitally it is only for 49 seconds. Then the page is displayed correctly. Subsequent reloading of effected page in same browser session yields expected behavior. Under debian, XFree 4.1.0-9 with internal font engine. GTK 1.2.10-9.
w2k's notepad says those characters include bidi direction markers. so this is either a bidi problem, or an intl detection problem. [my bet is the latter, but i'm going to ask the former to do the analysis]
Assignee: asa → mkaply
Component: Browser-General → BiDi Hebrew & Arabic
QA Contact: doronr → zach
Notepad seems to be interpreting the data as UTF-8, which means that these four bytes are displayed as two Arabic characters: 06BA;ARABIC LETTER NOON GHUNNA and 064D;ARABIC KASRATAN. Unless you have the default encoding set to UTF-8 in Mozilla, this isn't a Bidi problem.
I had the defaut character encoding set to UTF-8. But I've tested it on both 0.9.7 and 2002012814 with the Western character set and am experiencing the same behavior as when they were set to UTF-8.
Does this still happen with a recent nightly? I've tried setting my Character Coding to UTF8 but the attachment loads instantly without a freeze.
I am still experienceing buggy behavior but it is not nearly as bad. The browser only takes about 5 seconds of 100% cpu utilization before displaying the attachment. Then I see a wierd double forward very small slash and a very small curved inward u. build 20020412.
sounds confirmed with enough info to be set 'new'
Status: UNCONFIRMED → NEW
Ever confirmed: true
Sounds like more "when we see a weird char we have to look for a font for it and X is butt slow to do that" stuff.....
Depends on: 116645
Component: Layout: BiDi Hebrew & Arabic → Layout: Text
QA Contact: zach → layout.fonts-and-text
Assignee: mozilla → nobody
Does the problem still occur in a Nightly build? http://nightly.mozilla.org/
Whiteboard: [closeme 2011-11-10]
Severity: critical → major
Keywords: perf
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → INCOMPLETE
Whiteboard: [closeme 2011-11-10]
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: