311223 - Mail or news composition (new or reply) grabs 85-95% system resources (processor / cpu)

Status: NEW

(Core :: XUL, defect)

Description

[Rinaldi J. Montessi]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a1) Gecko/20051005 SeaMonkey/1.1a
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a1) Gecko/20051005 SeaMonkey/1.1a

When composing news/email or replying to an existing news/email message the
seamonkey-bin process immediately jumps to about 90% cpu per "top".  

Reproducible: Always

Steps to Reproduce:
1.Start SeaMonkey
2.Open Mail/News
3.Compose new message or reply to existing

Actual Results:  
CPU monitor turns red and shows 100% usage.  Mail and news messages are
delivered with no problem and continued usage is unimpaired.  Resources never
return to normal.

Expected Results:  
Normal mail/news correspondance without system resources being continually consumed.

about:buildconfig

Build platform
target
i686-pc-linux-gnu

Build tools
Compiler 	Version 	Compiler flags
gcc 	gcc version 3.4.4 	-Wall -W -Wno-unused -Wpointer-arith -Wcast-align
-Wno-long-long -pedantic -pthread -pipe
c++ 	gcc version 3.4.4 	-fno-rtti -fno-exceptions -Wall -Wconversion
-Wpointer-arith -Wcast-align -Woverloaded-virtual -Wsynth -Wno-ctor-dtor-privacy
-Wno-non-virtual-dtor -Wno-long-long -pedantic -fshort-wchar -pthread -pipe
-I/usr/X11R6/include

Configure arguments
--enable-application=suite --enable-calendar --disable-tests
--enable-extensions=default,irc,tasks --without-system-nspr
--without-system-jpeg --without-system-zlib --without-system-png
--without-system-mng --disable-debug --enable-optimize=-Os --enable-crypto
--enable-default-toolkit=gtk2 --enable-xft --disable-freetype2 --disable-xprint
--with-pthreads --enable-canvas --enable-glitz

Thinking it might be something in my profile, I made a new profile from scratch.
 I rebuilt without pthreads, canvas, and glitz.

I ran the Mozilla 1.7.12 that came with the distro (slackware 10-2) and did not
see the same problem.

This has been happening for the past few days CVS builds; checkout finish: Wed
Oct 5 10:18:15 EDT 2005.  I deleted the entire mozilla tree prior to the last
build to be sure there were no artifacts messing things up.

I've been compiling with gcc-3.4.4 since 7/15/05. err 20050715 with no other
glitches except for an mplayer problem in ltd.h.

I did make a debug build but since it didn't crash there was nothing to bt.  I'm
sure there must be a way to isolate the threads and determine what's happening,
but I'd need help with that.

Comment 1

[R.K.Aa.]

recent intense CPU load also reported in bug 310854, bug 311010, bug 311247

Comment 2

[Rinaldi J. Montessi]

Possibly something broken in general, but here it shows only in mail/news
compose.  Browser and Mail/News have been open for several hours and show "0"
CPU at present.

Comment 3

[Rinaldi J. Montessi]

Attachment: strace -ff of runaway process (partial)

Here's a partial strace -ff from the runaway process.  It took about 15 seconds
to generate a 50MB file!

Comment 4

[Andrew Schultz]

confirmed with linux seamonkey trunk 2005100405.  This regressed between trunk
builds 2005092901 and 2005093005.

I'll attach some stacks.

Comment 5

[Andrew Schultz]

Attachment: stacktraces

Comment 6

[Andrew Schultz]

*** Bug 311247 has been marked as a duplicate of this bug. ***

Comment 7

[Mike Kaply [:mkaply]]

We are seeing a similar problem on OS/2.

Supposedly the patch that caused this went on the branch too - is this problem
happening on the branch for you?

Comment 8

[Andrew Schultz]

1.8 Branch is fine for me.  So perhaps some interaction with another fix that
didn't make the branch.

Comment 9

[Boris Zbarsky [:bzbarsky]]

What a mess.  We're ending up with a height 0 treebody. Its width ends up
oscillating, and so whether it has horizontal overflow ends up oscillating.  Now
whenever its overflow changes it dispatches an overflow or underflow event which
the tree binding catches...  and in said event handler the tree binding changes
the node in question (removes or adds a scrollbar).  Which changes whether the
node is overflowing, etc.  So we end up in a cycle handling alternating overflow
and underflow events.

Now the only reason bug 301411 is involved is that as I said the height is 0. 
Which means that before the patch to that bug we just ignored some resizes and
cut off this cycle.

Comment 10

[Boris Zbarsky [:bzbarsky]]

Attachment: One possible fix

If we can't figure out why trees are screwing us over here, we should probably
just back this part of bug 301411 out -- it's not strictly needed to fix that
bug.  :(

Comment 11

[Rinaldi J. Montessi]

Works for linux.  Thank you.

Comment 12

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 198724 [details] [diff] [review]
One possible fix

Checked this in on trunk.  Leaving the bug open to fix the real underlying
issue with trees here...

Comment 13

[Boris Zbarsky [:bzbarsky]]

*** Bug 310854 has been marked as a duplicate of this bug. ***

Comment 14

[Wayne Mery (:wsmwk)]

*** Bug 311663 has been marked as a duplicate of this bug. ***

Comment 15

[Wayne Mery (:wsmwk)]

better summary

Comment 16

[Adam Guthrie]

*** Bug 311369 has been marked as a duplicate of this bug. ***

Comment 17

[Scott MacGregor]

Mike can you elaborate on why you plussed this bug for RC1.

This bug report is listed as being a trunk bug. 

I am unable to reproduce this on the branch at all. 

Downgrading the plus to a ? so the delivery team can evaulate and decide if this
is a blocker for the branch. Would be helpful if someone has a test case for the
branch to see this on. I'm not able to. 

Comment 18

[Mike Kaply [:mkaply]]

Sorry, the reason I plussed it was because I was concerned that this problem
would have some sort of backlash on the branch even though we hadn't seen anything.

It bothers me a great deal that a patch has such a bad effect on the trunk and
does nothing on the branch...

Comment 19

[Boris Zbarsky [:bzbarsky]]

Frankly, we should probably just take the "one possible fix" patch on the branch...

Comment 20

[Boris Zbarsky [:bzbarsky]]

Although...  The reason this broke on trunk is that it's a bad interaction with
horizontal scrolling of trees.  Since that's not on the branch and doesn't plan
to land there, I think we're ok.

Comment 21

[James Teh [:Jamie]]

This is not reproducible for me in Thunderbird version 1.6a1 (20051010), but 
was reproducible in builds of a few days ago. Was anything changed? Can anyone 
confirm this?

Also see my comment (#64) in Bug 246909.

Comment 22

[Boris Zbarsky [:bzbarsky]]

See comment 12.  Sometimes I wonder why I even bother commenting on bugs... :(

Comment 23

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Boris, I read everything you write most carefully :-)

Comment 24

[James Teh [:Jamie]]

My apologies. I usually do pay attention to comments, but I missed this one 
somehow, probably due to only skimming this bug after Bug 311369 was marked as 
duplicate. Sorry for the needless spam. :)

Comment 25

[Brock Haywood]

Not sure if this is the exact same bug but i have had sunbird on several
occasions run at approximately 40% cpu while minimized and idle and 50% while
idle and maximized.  I have noticed this behaviour on both winxp and 2000.  Even
if i quit sunbird and restart sunbird, the problem persists, however, it does
not persist after a reboot and wont show up for possibly a couple weeks before i
see it again.  (i cant reproduce it reliably).

Comment 26

[Bernd]

re comment 19: before this makes it to branch there is some evidence that this patch has something to do with bug 322074.

Comment 27

[Wayne Mery (:wsmwk)]

bz, should attachment 198724 [details] [diff] [review] be marked obsolete? (after all, it did check in)

Comment 28

[BugBot [:suhaib / :marco/ :calixte]]

The severity field for this bug is relatively low, S3. However, the bug has 4 duplicates.
:enndeakin, could you consider increasing the bug severity?

For more information, please visit auto_nag documentation.

Comment 29

[BugBot (nomail) [:suhaib / :marco/ :calixte]]

The last needinfo from me was triggered in error by recent activity on the bug. I'm clearing the needinfo since this is a very old bug and I don't know if it's still relevant.