Closed
Bug 150801
Opened 24 years ago
Closed 17 years ago
Gecko freezes MFCEmbed (and all embedded apps)
Categories
(Core :: Layout, defect)
Tracking
()
RESOLVED
INCOMPLETE
Future
People
(Reporter: lucxaw, Unassigned)
References
()
Details
(Keywords: perf)
MFCEmbed’s UI will temporarily freeze every time it renders a page. Once
rendering is complete, MCDEmbed will become responsive again. This is only
noticeable on slow computers (ie 200MHz).
Specs: Windows 98, 128 MB RAM, Boradband connection
How to reproduce:
This behavior is probably only reproducible on slower processors.
1) Start MFCEmbed and open two windows.
2) In window #1, open a page that takes a long time to render (ie. a Slashdot
page with nested comments at -1 threshold)
3) Open the file menu and move the mouse back and forth opening all the menus,
when Gecko begins to render the page, the menus will stop responding to your
mouse movements. You cannot click on window #2. You cannot move either window
via its title bar nor interact with either window in any way.
4) When the page finishes rendering and displays, both windows become responsive
again.
I’m guessing that with a broadband connection, Gecko must render a large amount
of data in a short amount of time, thus starving host app of any processor time.
However, I have seen and heard of this behavior on modem connections as well.
This happens every time Gecko renders a page. If it takes 0.1 seconds to render,
the host app will freeze for 0.1 seconds.
Affected applications include MFCEmbed, K-meleon, and applications using the
Active X Control such as Winamp 3 Mini Browser, Topstyle, and Scope. If the
program has more than one window, all windows become unresponsive for the
duration of time Gecko takes to renders.
Mozilla, however, does not freeze completely during page load.
This may be related to Bug 91643 - “All Mozilla windows frozen during some
stages of loading a page” except 91643 describes shorter spasms of UI freeze
while this bug describes the complete UI freeze for the duration of page load.
This problem has been around for a very long time, probably as long as MFCEmbed
has existed, thus this may be a duplicate of another bug. K-meleon has suffered
from this behavior ever since it was created.
Now that 1.0 has shipped, I’m hoping this problem receives the attention it
deserves.
Gecko freezes because everything is happening on the same thread as the UI. If
you have a complicated page then it might momentarily freeze during rendering in
order to calculate the layout. If this is not a problem in Mozilla then it may
suggest embedding may have a problem with timers or the default layout settings.
We don't change the default layout pref settings, but it might be worth tuning
them from embedding.
You might like to see if adding and playing with these prefs in
defaults\pref\all.js affects the results you're seeing.
layout.reflow.timeslice - Upper bound on time spent reflowing (default is -1)
layout.reflow.async.duringDocLoad - async reflow (default PR_FALSE)
In particular, capping the timeslice to some value might improve responsiveness
at the cost of longer layouts.
Status: UNCONFIRMED → NEW
Ever confirmed: true
I tried both prefs with timeslice values of 1, 100, 1000, 5000, 100000, but did
not see any improvement.
Timeslice did not have any effect on Mozilla either.
It probably goes without saying, but Internet Explorer embedded does suffer
similarly, but not nearly to the extent of Gecko and only under extremely high
loads.
Over to layout for ideas
Assignee: adamlock → other
Component: Embedding: APIs → Layout
QA Contact: mdunn → ian
Comment 4•24 years ago
|
||
This seemss like a pretty big deal.
Some of the other timers (the content sink timers, which have to prefs,
"content.notify.backoffcount" (a small integer or -1 to disable, -1 being the
default), "content.notify.interval" (default 120000),
"content.interrupt.parsing" (default true), "content.max.tokenizing.time"
(default is content.notify.interval times 3), "content.switch.threshold"
(default 750000) may also have an effect. (We have too many knobs.)
There are other related bugs, such as bug 91643 (see bug 91643 comment 16).
I wonder if this:
http://bugzilla.mozilla.org/show_bug.cgi?id=165039
helped address this problem? We definitely have had fewer reports in K-Meleon of
this problem and we have users on very slow systems.
Updated•17 years ago
|
Assignee: layout → nobody
QA Contact: ian → layout
Updated•17 years ago
|
Status: NEW → RESOLVED
Closed: 17 years ago
Resolution: --- → INCOMPLETE
You need to log in
before you can comment on or make changes to this bug.
Description
•