The above site when scrolled (unless done *very* slowly) is extremely jerky in redrawing the page. This is true in both Moz 0.9.6+ and NS 6.2. Investigation showed that the background image, tile.gif, is 6 x 1500 (wot the heck was this designer doing????) causing the image to be redrawn roughly a schmillion times for a full page redraw. I edited the page and cloned the background image to make one maybe 600 or so pixels tall and using this image locally the page scrolls nicely. No problem scrolling in IE 5.5.
Forgot to add I'm on a P4 1.7, 98SE, ATI all-in-wonder Radeon 32M.
Is it just as bad in current trunk builds? There have been bugs filed on similar issues before and at least some are fixed...
Sorry, but I haven't used ay of the overnights so I can't comment on their performace in this matter.
OK - my curiosity of the nightly builds got the better of me so I tried 2001122908, which didn't fix the problem, still jerky.
Very strange... Scrolling is fine for me, Win98SE, 2001122703 (and GeForce2MX, if that matters...)
Perhaps your video performance is enough to overcome the jerkiness although the ATI AIW Radeon is no slouch either.
Assignee: asa → attinasi
Status: UNCONFIRMED → NEW
Component: Browser-General → Layout
Ever confirmed: true
QA Contact: doronr → petersen
*** Bug 120054 has been marked as a duplicate of this bug. ***
Reassigning to Don.
Assignee: attinasi → dcone
Target Milestone: --- → mozilla1.0
This is most likely caused by a regression which resulted in no longer calling patblt on WIN32. It also is not using the cross-platform code which quickly replicates the background image into a larger tile.
I ran this page on both explorer 6.0 and Mozilla. I found the following: 1.) There is an animated GIF that seems to make both browsers render chunky.. but if you scroll down to where the animated GIF is not seen... both pages scroll fast. 2.) The two browsers render the page differently.. its probably a layout bug.. but its definetly different.. and it could be one of the causes of the chunky scrolling. 3.) The speed may or may not be identical to explorer..but it seems to me -- very simular.. the only way I could think to increase this speed much more is to cache the tiles we create to increase the blitting speed. I dont think that the differences are enough to warrent that. In conclusion.. I dont think this is a bug for Windows 98. Windows 2k, NT and XP.. OS's will get faster when I fix the background bugs by using the PatBlt call... which 98 does not support. I am marking this wont fix.
Status: NEW → RESOLVED
Last Resolved: 16 years ago
Resolution: --- → WONTFIX
Target Milestone: mozilla1.0 → mozilla1.2
dcone: is there a bug # for the fix you are planning to do?
From bug 120054 : Changed URL to http://www.lindows.com/lindows_products.php This page has a tiny background image repeated many times. Scrolling is extremely slow. Re-opening. If I am out of line, feel free to change it back. But I feel that this is definately an issue. IE and NS 4.x have no slowness whatsoever scrolling the very same page.
This very same PC when running NT4 SP6 *does not* exhibit the scrolling problem. Have to check which video driver but if I remember it is the same as the 98SE one. In terms of animated gifs and page locations, the original posted espn page still shows the jerkiness no matter what part of the page is being viewed when scrolled. The jerkiness is really bad, sometimes a second or so for the redraw, 1024x768. The lindows link given suprisingly gives very little trouble for me, almost, but not quite, being smooth.
I am working on NT4 SP6 with a PII 400 / 380 MB RAM and the lindows link is really slow scrolling on my box... On the same box, with IE the page scrolls fine with minimal CPU usage On linux, with a PIII 733 / 128 MB RAM everything is fine too.
Let me get this straight. For Jeff Givens.. Windows 98 is slow. Windows NT is fine. This makes sense since NT uses a PatBlt for the background tiling of small images. for Andrea Aime.. Windows NT is extremely slow. Could be a video card incompatibily. I dont understand this at all. What pull are you using. WD -- what OS and pull are you using? Describe what slow is for you.. are you seeing 1 second delays, or a chunkiness? For me.. 98 (400 pII) is a tad chuncky.. but no noticible delays. NT(400 PII) .. smooth as silk. W2K(1500 athlon) smooth as silk. I just had a few other run 98 and they did not notice any chunkiness with 98 400 PII.. but I think the chunkiness I saw was just under there radar. Before I disposition this bug.. I will run on a few more machines.. NT 4.0 and up and see what I can see.
Ok, some more info: my box has a Matrox G100 video card driving a 19'' Monitor (sounds a bit strange, I now, but that's the way it goes, I cannot change it... :-( Now, it's supposed to be a low end one, but let me repeat that IE runs on the same hardware and it is smooth as silk, while Mozilla is slow, slow, slow (only on this page, usually I cannot see any speed difference between IE5 and Mozilla). They do run on the same machine, so the hardware is for sure not the only problem... BTW, I'm running Moz 2002022703...
For our tiling we use a PatBlt if the machine is NT (NT 4.0, W2K, XP). I pulled the PatBlt for windows 98 because people were having video driver incompatibilities (machines with bad drivers would crash,those with the correct drivers would blaze, this was a documented Windows Bug).. so there was some sacrifice in speed. One theory here might be that your having a video driver problem on your NT box. Your comparing apples and oranges when you say Explorer works fast and we don't. They could and probably do have a competely different architecture as far as what they have in memory. They could have an entire page buffered offscreen, etc. We re-render every single time you scroll to an offscreen buffer. Mozilla depends on very fast blitting and rendering of tiles. For small background images we blit a trillion times..thats just the way it is. For now I want to find out why your having difficulties compared to other mozilla users.. not compared to other software.
Created attachment 72676 [details] [diff] [review] Patch to get NT to PatBlt no matter what tile size.
Comment on attachment 72676 [details] [diff] [review] Patch to get NT to PatBlt no matter what tile size. firstname.lastname@example.org
Attachment #72676 - Flags: review+
Comment on attachment 72676 [details] [diff] [review] Patch to get NT to PatBlt no matter what tile size. sr=attinasi, but please update the comment to reflect all of the cases in that complex 'if' statement
Attachment #72676 - Flags: superreview+
Comment on attachment 72676 [details] [diff] [review] Patch to get NT to PatBlt no matter what tile size. a=asa (on behalf of drivers) for checkin to the 1.0 trunk
Attachment #72676 - Flags: approval+
The patch was checked in.
Status: REOPENED → RESOLVED
Last Resolved: 16 years ago → 16 years ago
Resolution: --- → FIXED
On Win2k, http://www.sysinternals.com/ntw2k/source/filemon.shtml scrolls *extremely* slow. Moz is nearly unusable on my Athlon 1800+ / Geforce 256 on that page. Build 2002032103 No DHTML or anything goofy. Just a small image repeated as a background: http://www.sysinternals.com/images/background.gif re-opening
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
WFM: Using cvs release build on 3/23 it scrolls smoothly on 750Mhz athlon.
http://www.sysinternals.com/ntw2k/source/filemon.shtml works almost ok for me on the offending win98 box. A tiny bit jerky but nothing like the espn.com example. Speaking of which, something of interest: The other day on the win98 box that all this was reported upon, the espn page scrolled nice and smooth, no problems at all. I thought maybe the changed their page with a different background image. I did file, edit page, and Moz crashed. Upon resarting the jerkiness is back and remains. Odd.
Page scrolls fine(little jerkiness) for me on my 400 NT 4.0 machine. I think maybee somekind of driver issue.. with some offending machines. My Athlone 1500 with a GForce II scrolls very smoothly.
I am marking this works for me.. and then I opened another bug which I think is the only right way to solve this bug... 133261. I dont think I can go much further with PatBlt and the cacheing in 133261 is the only way to go as far as I can tell.
Status: REOPENED → RESOLVED
Last Resolved: 16 years ago → 16 years ago
Resolution: --- → WORKSFORME
Marking verified wfm with the April 23nd build (2002-04-23-06) under Windows ME.
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.