Closed Bug 310854 Opened 19 years ago Closed 19 years ago

Browser generates CPU load while idle

Categories

(SeaMonkey :: General, defect)

Product:
SeaMonkey
Component:
General
Platform:
x86
All
Type:
defect
Priority:
Not set
Severity:
major

Tracking

(Not tracked)

Status:
VERIFIED DUPLICATE of bug 311223

People

(Reporter: sgautherie, Unassigned)

References

Details

(Keywords: regression, Whiteboard: [Regression TimeFrame: 0929-0930 !])

[Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.9a1) Gecko/20051002 SeaMonkey/1.1a] (nightly) (W98SE) 1. Start the Browser, and let it idle on start page (I use <about:>). (MailNews part may have +/- the same bug.) R. SeaMonkey uses 20+ % of my (K6-200) CPU. [Whereas there should be "none".] I think MailNews is +/- affected too. I wonder if it could be related to (FireFox) bug 309956 somehow ?? [Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.9a1) Gecko/20051001 SeaMonkey/1.1a] (nightly) (W98SE) [Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.9a1) Gecko/20050930 SeaMonkey/1.1a] (nightly) (W98SE) +/- same bug, as I recall. [Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.9a1) Gecko/20050929 SeaMonkey/1.1a] (nightly) (W98SE) This seemed to work fine !? [Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.8b4) Gecko/20050910 SeaMonkey/1.0a] (release) (W98SE) This one is fine !
I checked again with this test case, and confirm the regression timeframe: 0929-0930.
Whiteboard: [Regression TimeFrame: 0929-0930 !]
*** Bug 310932 has been marked as a duplicate of this bug. ***
Setting OS to ALL, as the dupe was on Linux. I've seen the bug last two days on Win98SE and can confirm the time range for regression. I'm back to BuildID 2005092906 because the bug was unbearable, too difficult to start a download. Starting the browser and looking at the 'bugs filed today' (from QA in the menu bar) was sufficient to bring the 512MB Athlon XP1600+ to 98% CPU. When I then tried to download a file with right-click, Save Link Target as... the Filebox opened, but the context menu didn't close.
OS: Windows 98 → All
Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.9a1) Gecko/20051004 SeaMonkey/1.1a On buggy builds CPU usage is per window, measured with Systems Information from Process Explorer: http://www.sysinternals.com/Utilities/ProcessExplorer.html Tests done online, empty windows, Win98SE, Athlon XP1600+ BuildID 2005092906 uses 1% CPU, doesn't matter if 1 or 5 windows. BuildID 2005100405 uses 8% CPU one window, 12% for 2 windows, 24% for 5 windows. Tests done offline, empty windows, Win98, Celeron 333: BuildID 2005092906 uses about 5% CPU, small difference between 1 and 3 windows. BuildID 2005093006 uses 33% CPU one window, 50% for 2 windows, 60% for 3 windows. The CPU usage is slowly oszillating plus/minus 5%.
intense CPU load now also reported in bug 311010, bug 311247, bug 311223
Hermann, I'm working on your request in bug #311058, haven't installed that window monitor yet, but see that the 10-4 build's zip file has something new, a read-only file named NSSCKBI.DLL. It appears to contain security certificate info.
Herman, I'm running Process Explorer and the 2005100405 Win32 build and am getting Seamonkey processor usage of 88-95% typically. Had to kill some prgs & mail window. Bad. Mouse interrupt handling isn't right on scroll bar, not continuous scroll on long click. 256mb and short of resources. malloc problem? no swapping.
With Win32 build 2005100305, Seamonkey's using mostly 80-95% of cpu time with the mail window and one browser window open, with no activity. Clicking and holding the mouse on the scroll bar only causes one scroll movement, not continuous. But it's not hanging at this point. Haven't had to kill any programs in the tray. So the 10-04 build seems a bit worse. System specs: 256mb, 900mhz Duron, Win98SE with most or all upgrades.
Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.9a1) Gecko/20051006 SeaMonkey/1.1a Celeron 333, Win98, 128MB: Load measured after peak went down to a stable level, which was held for 20 seconds. 1 / 2 / 3 / 4 empty windows: 24% / 45% / 64% / 85% after idling 20 minutes on one empty window: 20% Measurements taken after browserstart, online, no pages loaded. BuildID 2005092906 trunk and Mozilla 1.7.12 no difference seen testing startpage only and testing startpage plus one empty page. BuildID 2005092906 took a stable 3%, Mozilla 1.7.12 was toggling between 3 and 4%, timing was updated each 5 seconds. Seamonkey 1Window/2Tabs (gmail + this bug) + Firefox about: = 33% Same as above plus one more empty page in Firefox: same percentage, 33% Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.9a1) Gecko/20051002 Firefox/1.6a1 So current Firefox doesn't increase CPU load visibly with the number of open windows. John, what do you see if you open the browser only, one empty page only, homepage or whatever? Does the load visibly increase if you open a second empty window?
Blocks: 311010
Depends on: 311223
Keywords: perf
Serge, an increase in CPU load from 0.1% per window to 20% per window isn't a perf problem, it is a bug. What I don't know is, if that is a problem on Win98/Win98SE only, or also on XP or Linux, MacOSX. Lots of bugs are filed for nightlies after this date because load jumps randomly to 100%, but nobody looks at the initial load. If comparing nightlies on a 333 MHz CPU I see a jump from 5% to 33% cpu in an empty window, right after start of the browser, or 1 to 8% on a 1600 MHz CPU, that can't be perf, that is a bug. If I see a jump from 5% to 33% CPU in an empty window right after start, I don't have to test how the browser behaves on a non-empty page, it only could get worse.
Under Linux, Seamonkey build 2005100202, I do *NOT* see an initial load for a page display - in fact I can open a very LARGE number of tabs with no significant CPU load whatsoever. I only see problems when I compose a mail message. Given comment #11 I thought this datum would be of use.
Loading per page: 1 window 7-15% Moz home. 2 15-25% 3 25-35% 4 35-55% 5 40-50% These are "about Seamonkey" pages, no processing. Almost died during editing of this, killed proc browser, etc. again no swap file usage, just "out of resources"
Ahhh, back using 1.7.12. The 09-29 build worked well, but the 10-03 build showed some cpu loading, though not as severe as with the 10-04 build. If the 10-02 build worked ok, that pins down the time the problem developed. What was checked in for those builds? Besides the loading, continuous mouse clicks aren't working right on the scrollbars. They cause only a one-time movement, not scrolling. Could this show that there's some interrupt handling connected to the cpu loading? Maybe interrupts that keep allocating more memory and never releasing it, so the system runs out of resources (RAM)? Any way to test that idea?
(In reply to comment #11) > Serge, an increase in CPU load from 0.1% per window to 20% per window isn't a > perf problem, it is a bug. Then, removing this keyword: I added it because it was set on the two d.o/b. bugs...
Keywords: perf
Tested again with todays build, XP Pro, SP2 + all security updates: After startup on a blank page the apps CPU load varied between 2-6%. Then loaded query for "bugs I commented in, modified the last day" in bugzilla (a couple of pages long). CPU usage jumped to 97%, then stayed between 93-97% for minutes after the page was loaded. Exited before it fried anything. Severity "major" looks like an understatement. Any testing is out of the question in this state. This bug is a total blocker as far as I'm concerned. I suggest upping severity unless some workaround exists. The CPU is an Intel Pentium M, 1.6 MHz.
Seems Bug 301411 fits the bill, see bug 311223 comment 10 http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=SeaMonkeyAll&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2005-09-29+19%3A29&maxdate=2005-09-29+19%3A29&cvsroot=%2Fcvsroot Don't use nsRect's operator== for comparisons where different sized empty rects should be treated as different. But 301411, r+sr=roc As I&#180;m seeing a huge increase in CPU load for the first (empty) window, and also a huge increase for every other window, I suspect it is this bug which regressed the numbers in comment 4, comment 10, comment 13
Depends on: 301411
So is this basically a dup of bug 311223?
Installed browser only - same bug. Perhaps bug 311223 is basically a dup of this?
No idea. Please retest once that patch lands.
Thank you, Boris. Just tested a 2005100720 Seamonkey Creature build with your patch for bug 311223 in place, and everything seems to be working perfectly: - CPU usage normal for browser and mailnews, - toolbars loading/painting normally, and - scrolling normal for both mousewheel and manual click'n'drag (problems reported in bug 310916).
No longer blocks: 311010
*** Bug 311010 has been marked as a duplicate of this bug. ***
wfm Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.9a1) Gecko/20051007 SeaMonkey/1.1a Tinderbox build from CREATURE, Build ID: 2005100722 from comment 4: Tests done offline, empty windows, Win98, Celeron 333: BuildID 2005092906 uses about 5% CPU, small difference between 1 and 3 windows. BuildID 2005093006 uses 33% CPU one window, 50% for 2 windows, 60% for 3 windows. BuildID 2005100722 uses 5% CPU, from 1 up to 5 windows. Seems bug was fixed by checkin: Wallpaper for bug 311223 pending a fix of the underlying tree bug. http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=SeaMonkeyAll&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2005-10-07+19%3A32&maxdate=2005-10-07+19%3A32&cvsroot=%2Fcvsroot
*** This bug has been marked as a duplicate of 311223 ***
Status: NEW → RESOLVED
Closed: 19 years ago
Resolution: --- → DUPLICATE
No longer depends on: 311223
[Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.9a1) Gecko/20051008 SeaMonkey/1.1a] (nightly) (W98SE) V.Duplicate, as bug reporter.
Status: RESOLVED → VERIFIED
No longer depends on: 301411
confirming at BeOS - 2005-10-04-trunk - CPU is busy until first page loaded. about 50% at Dual PIII-550. But with 2005-10-08-trunk it seems gone
You need to log in before you can comment on or make changes to this bug.