Closed
Bug 71857
Opened 24 years ago
Closed 23 years ago
Massive multiple refetch of images on auto-refresh
Categories
(Core :: Graphics: ImageLib, defect, P2)
Tracking
()
RESOLVED
FIXED
mozilla0.9.8
People
(Reporter: jesup, Assigned: pavlov)
References
()
Details
(Keywords: perf, testcase)
Attachments
(2 files)
FreeBSD 4.1 20010312xx
When displaying the attached testcase, it takes a really Long Time to display or
resize. Note that each cell is:
"<td BGCOLOR="#xxxxxx"><img src="Blank.gif"></td>"
in a "<table width="100%" height="100%" CELLSPACING=0 BORDER=0>"
I suspect a seriously large number of reflows since the images are not
constrained in size, even thought they're all the same image. This is related
to bug 54542, and may in fact be the same, except that this has blank images and
thus may tickle the reflow code differently.
Also there are semi-random gaps when the table is displayed (in a large window),
but that's another bug probably.
Reporter | ||
Comment 1•24 years ago
|
||
Reporter | ||
Comment 2•24 years ago
|
||
Reporter | ||
Updated•24 years ago
|
Reporter | ||
Comment 3•24 years ago
|
||
Adding original URL - please don't slashdot his site. Note that the colors
generated each time are different. Also note we don't support <table
height="100%">
Comment 4•24 years ago
|
||
Marc, this may be a dup of another bug. Moving to m0.9.
Assignee: karnaze → attinasi
Target Milestone: --- → mozilla0.9
Comment 5•24 years ago
|
||
In serious need of Quantify, but it sure looks like bug 54542
Status: NEW → ASSIGNED
Updated•24 years ago
|
Status: ASSIGNED → RESOLVED
Closed: 24 years ago
Resolution: --- → DUPLICATE
Comment 6•24 years ago
|
||
This is a dup of bug 54542 - the same problem but on a smaller scale. Once the
overwhelming impact of the style sharing degeneration is fixed, if it can be,
then we may see some differences due to the transparent images in the cells
(this case will likely be worse, proportionally). But at this time, the problem
is the same: too many style contexts slowing down style sharing.
*** This bug has been marked as a duplicate of 54542 ***
Comment 7•24 years ago
|
||
Verifications. Tests (if necessary) were done with 2001052504 on Windows 2000.
Please forgive the spam.
Status: RESOLVED → VERIFIED
Comment 8•24 years ago
|
||
Reopening. This loads fine on the initial load (in fact, it's not that slow
even before the pending style system changes). But there is a 'Refresh: 60'
header and when that reload occurs it becomes very slow. This something
to do with HTTP and/or cache. I'm looking into this some more and will post
further (and get this out of attinasi's hands, since this isn't style, as
far as I can see).
Status: VERIFIED → REOPENED
Resolution: DUPLICATE → ---
Comment 9•24 years ago
|
||
So, the document is a 25x40 cell table, with each TD being something
like this '<TD BGCOLOR="#7f986c"><IMG SRC="Blank.gif"></TD>'.
When the document is initially loaded, the top-level document and one copy
of 'Blank.gif' are fetched from the server. The page is layed out fairly
quickly, about 2-3s on a 500MHz/128MB system (not real fast, but not
pathologically slow).
But, when the 'Refresh: 60' triggers the reload, we do one unconditional
GET for the top-level document, and follow that up with (oops!) 1000 additional
unconditional GETs of the same 1px GIF that is contained in every table
cell. That's not good. (I expect the "random gaps" that appear while the
table is layed out are just reflecting the asynch fetch of each redundant
copy).
Clearing the mozilla0.9 milestone that was set (since that has passed), and
adding mozilla0.9.2 kw, and after flipping a coin, giving to pavlov.
It would be good to at least determine the scope of this problem (i.e., is it
only for some narrow set of conditions, or is something more fundamentally
wrong). Note: I tried this in a number of old builds on win2k, and I could see
this "load 1000" in builds back to 4/18, but not in a build of 4/11 or 4/13.
[I'm probably guilty of morphing this bug from it's original intent. Apologies]
Assignee: attinasi → pavlov
Status: REOPENED → NEW
Component: Layout → Networking: HTTP
Keywords: mozilla0.9 → mozilla0.9.2
QA Contact: petersen → benc
Target Milestone: mozilla0.9 → ---
Assignee | ||
Comment 10•23 years ago
|
||
this is prolly my validate always thing
Assignee | ||
Updated•23 years ago
|
Target Milestone: --- → mozilla0.9.3
Comment 11•23 years ago
|
||
If this is an imagelib thing, I'd like to move it out.
Reporter | ||
Comment 12•23 years ago
|
||
Probably not an imaglib thing - note that it's assigned to http.
Comment 13•23 years ago
|
||
->imagelib, since pav has it.
Comment 14•23 years ago
|
||
Doesn't look like this is getting fixed before the freeze tomorrow night.
Pushing out a milestone. Please correct if I'm mistaken.
Target Milestone: mozilla0.9.3 → mozilla0.9.4
Assignee | ||
Updated•23 years ago
|
Target Milestone: mozilla0.9.4 → mozilla0.9.5
Reporter | ||
Comment 15•23 years ago
|
||
Reverified with mozilla 2001083103 (I'll check with a more recent build soon).
We're still massively refetching images on automatic reloads.
Reporter | ||
Comment 16•23 years ago
|
||
Bug still there in 2001092703 - we should give this some attention - it will
affect news sites that auto-reload like CNN.com
Reporter | ||
Comment 17•23 years ago
|
||
Changed subject to match jrgm's change of the focus of this bug. On
auto-refresh, we're fetching the same image (Blank.gif) thousands of times; on
the initial load we fetch it once.
Keywords: mozilla0.9.2
Summary: VERY slow formatting of table of transparent gifs w/ background colors → Massive multiple refetch of images on auto-refresh
Comment 19•23 years ago
|
||
It seems like the target milestone keeps moving...
Assignee | ||
Updated•23 years ago
|
Status: NEW → ASSIGNED
Priority: -- → P2
Reporter | ||
Comment 20•23 years ago
|
||
Pav: Please, please please try to address this bug. It eats bandwidth.... Lots
of news sites auto-refresh.
Assignee | ||
Comment 21•23 years ago
|
||
fix checked in
Status: ASSIGNED → RESOLVED
Closed: 24 years ago → 23 years ago
Resolution: --- → FIXED
You need to log in
before you can comment on or make changes to this bug.
Description
•