logs archiveIRC Archive / Freenode / #firefox / 2010 / June / 29 / 1
kbrosnan
e01: not even on the radar, nokia was doing some qt work for their mameo moblile browser but I don't know how much of that/if any translates into firefox
Renegade001: example of what you are doing?
Renegade001
Sorry, I deleted my code but whenever I tried to add an inset box-shadow and another type of box-shadow, nothing appeared.
so say
.class { -moz-box-shadow: inset 0px 0px 3px #fff, 1px 1px 1px #000; }
alkisg
Hi, I'm having a problem where firefox makes xorg to temporarily use an additional 200MB of memory while loading a jpg image. That's crashing my thin clients as they don't have much RAM (firefox runs on the server, xorg on the clients).
This doesn't happen with other browsers, just with firefox. And it doesn't even happen with all images with firefox even if they have the same dimensions.
Here's a test case: http://people.ubuntu.com/~alkisg/tmp/
All images have the same dimensions (6480x4320). One of them uses no xorg memory, the .png one uses 100mb, and the jpg one 200 mb. That's only while loading
bholley
alkisg: just saw your issue. I'm one of the image guys - let me take a look
alkisg
bholley: thank you - I'll be around for 30' if you need direct feedback
kbrosnan
well there is a big difference between the images
the no usage one is a solid color
alkisg
True. But it results in the same RGBA raw data...
...as it has the same dimensions
rockband
anyone here
i got an issue
alkisg
The problem is while loading though, so I'd guess that RAM is used by the jpg/png decoders?
         

bholley
alkisg: the memory usage goes away after loading?
alkisg
bholley: yes
rockband
when i view source...it shows the source code as unicode and i have to try a few time until i get the readable version
like
<link rel="alternate"
bholley
which version of firefox is it?
rockband
anyone know whats the issue
newest
just updated
alkisg
bholley: 3.6.3+nobinonly-0ubuntu4
kbrosnan
bholley: partially, the swap stays around 100mb on a 256mb vm
bholley
so, this image is about 30 million pixels
kbrosnan
yeah 28m to be a little closer :P
rockband
must be from nasa's new telescope
lol
bholley
which means that an RGB representation should be somewhere ~100 megs
alkisg
Heh... It's sad but images like that are actually used on the web :-/ I got that one from a web page that crashed my clients.
kbrosnan
rockband: square it is only about 4k by 3k
bholley
so the 200 meg thing is curious
alkisg
bholley: it's also curious why the solid color doesn't use the 100 megs...
(nor the 200)
bholley
alkisg: there are some paletting optimizations, which might be it
kbrosnan
see similar results using Mozilla's build
alkisg
And why the same image is different with png/jpg
bholley
I'd have to investigate it
alkisg
png==100 megs, jpg==200 megs
         

bholley
can you tell me whether the 200 meg footprint goes away right after load? or does it take something like 15 seconds?
alkisg
They both go away after loading
(png/jpg)
So I can only see them for 1 or 2 seconds at most
awircser
Hi
I've unchecked images and scripts on FF, can it still load flash websites?
or video sites that uses flash plugin like youtube?
kbrosnan
tools > addons > plugins > disable flash
bholley
alkisg: load the image, then tab away from it and wait 15 seconds. Next, click back on the tab with the image. It should hang for a little bit before displaying. Confirm that, then see if you see the same memory spike
awircser
Thanks kbrosnan
bholley
(firefox discards decoded frames after 15 seconds of inactivity)
alkisg
bholley: I'm seeing a delay and some cpu usage, but I'm not seeing any xorg memory usage
Watching `top` isn't a very good way to monitor changes like that though, so if there was a memory usage for half a second, I might not see it even after 3-4 tries.
bholley: Nope, I've told `top` to refresh every 0.1 secs, so I should be seeing the changes, if any. No, alt+tab after 15 seconds doesn't change the xorg memory usage.
bholley
alkisg: ok, so my sense is that this is X-specific. When I run this on my mac, I do get a spike in firefox memory usage 50 megs higher than expected during first load, which is almost certainly the result of bug 517119, which I fixed on trunk a few days ago. I also get firefox memory usage jumping up and down 100 megs after 15s inactivity, as the frames discard/redecode. But this is all expected, and what you're talking about sounds specif
the window manager
let me see if there's anyone on IRC who knows about X (I don't)
alkisg
(Action) tries to see if the memory used is exactly 2x(raw image size)...
bholley
(Action) found an X owner - asking now
alkisg: So apparently what's going on is that the whole image needs to be sent over the wire, and then it's copied to the X buffer
so for a moment there are 2 copies of the image on X's side
alkisg
bholley: but that also happens locally
bholley
alkisg: ?
alkisg
If I try it on a regular pc, I get the same results
I just have more RAM there so it's no crashing...
bholley
alkisg: I'm talking about the pipe between firefox and X
alkisg
bholley: hmmm yeah I just realized that
Sorry I thought you were talking specifically about the thin client architectury
bholley
alkisg: this could be solved with shared memory buffers. but on most systems those are limited to about 33 megs
alkisg
bholley: why isn't the "no-xorg-memory.jpg" having that problem?
(btw, chromium doesn't have that problem either, on any of the test case images)
bholley
hm that is interesting
alkisg: either way though, if you were to load 2 such images, you'd still crash right?
alkisg
(Action) tries...
bholley
(it's possible that chromium is converting large images into tiles or something)
alkisg
I wonder if it's doing any off-screen rendering, so it's only using x memory for the final window (1024x768)
rebooting the client...
bholley: I loaded all 4 of the test case images in 4 separate chromium windows concurrently, no crash.
powerpc
hi! I just wonder why firefox (and especially on sites with lots of javascript) works tens times slower than opera?
dolske
Opera was slower first.
powerpc
I'm using 10.60 beta and firefox 3.6.4
bholley
alkisg: this is all viewing the images full-size right? so most of the image is scrolled offscreen?
alkisg
bholley: I tried both methods, no xorg memory difference and no crashing in either case (with chromium)
I don't know enough about X to be sure, but it doesn't seem logical to me for the whole image to travel over the Xorg socket
I think only the part that actually gets drawed should go over the wire...
Resizing happens on the client only when XV acceleration is used (while watching video). For normal images, resizing happens on the server.
bholley
alkisg: I think it's currently just one big x surface that we manipulated by strings and pulleys from the firefox side
some layering work that's coming up soon might save the day, but I'm not sure
alkisg
bholley: that wouldn't explain why the "xorg-no-memory.jpg" doesn't have the problem, though... so I still don't think that that's the reason for the temporarily increased xorg memory
dolske
I hope these are just test images, and you're not actually using those for a website. :)
alkisg
And, after the loading, ALL that memory goes away
dolske
there's a graphics optimization that gets triggered when images are one solid color.
alkisg
dolske: I got the photo one from a website that crashed my clients :)
Ah, that would explain the difference
powerpc
is there any way to speed js in firefox at all? maybe some forks or patches or such
speed up*
dolske
powerpc: try a nightly build? http://nightly.mozilla.org
though if you're using a PowerPC, I wouldn't be surprised if that's slower because we don't do PPC JITting.
alkisg
bholley: after the image gets loaded though, I don't have 100MB of xorg usage. If this one-big-surface was true, shouldn't I be having 100MB xorg RAM usage?
sdwilsh
and we may even drop support for it soonish
djustice
mmkay. ive a weird problem. im making a bundle system. to execute firefox from ~/.firefox/usr/lib, but the included xulrunner wants to look at /etc/gre.d instead of ~/.firefox/etc/gre.d .. how can i force it? ideas?
alkisg
bholley: I'm getting from 88MB xorg => image loading => 288 MB xorg => image done loading => 88MB xorg. Shouldn't the last one be 188MB if it was one big surface?
djustice
or. better question. when xulrunner says: Could not find compatible GRE between version 1.9.2.3 and 1.9.2.3. what ,exactly, does it mean?
dolske
how are you measuring memory?
alkisg
dolske: `top` and `htop`
...and actual client crashing ;)
powerpc
"we don't do PPC JITting": is it so hard or why? what does it mean?
dolske
it's time consuming, low-level work (assembly), and there are vanishingly small number of PPC users.
patches welcome!
powerpc
js handling requires assembly-level coding? o_0
dolske
yes. it generates native code for certain JS patterns, so that it runs as fast as native compiled code instead of being intrepreted.
alkisg
bholley, dolske: guys it's almost 3 am here so I'll be heading to sleep. Should I file a bug? And, where? In the mozilla tracker or in ubuntu (launchpad)?
powerpc
I downloaded nightly build and it is called "Minefield.app"; is it firefox?
dolske
alkisg: sure, feel free to file a bug.
alkisg
Thank you. Good night :)
dolske
powerpc: yes, nightly builds are called "Minefield".
powerpc
terrific title
dolske
they're actually impressively stable these days, though, in large part due to our huge test suite.
powerpc
okay, launching; wish me some luck
Mozilla/5.0 (Macintosh; U; PPC Mac OS X 10.5; en-US; rv:1.9.3a6pre) Gecko/20100628 Minefield/3.7a6pre ;)
sunspider results for this build of firefox: http://pastebin.org/364538
the same for opera osx ppc 10.54: http://pastebin.org/364587 ... hmm, firefox is 2 times faster now
xira
hello, can anyone explain to me why my back button is screwed up like in this picture? http://images.4chan.org/g/src/1277769329150.png (SFW)
it's fine in small mode
fumanchu182
Anyone having issues on ubuntu with flash support?
i.e. fullscreen crashes the browser
kbrosnan
make sure you are using the latest version of firefox and flash
get a crash report by using a build from http://www.mozilla.com
or if you know gdb https://wiki.ubuntu.com/MozillaTeam/Bugs#Obtain%20a%20backtrace%20from%20an%20apport%20crash%20report%20%28using%20gdb%29
fumanchu182
yeah my release version doesn't have debug symbols
already tried -g and --sync
kbrosnan
the build from ubuntu would need their dbg packages
fumanchu182
yeah and launchpad timed out twice already so i am getting quite frustrated
I have gdb running, i have called "run"
I try to go to a video in youtube and cause it to crash
when i type "backtrace" it says i have none.
is there a way to search launchpad for similar issues?
kbrosnan
i've never really been able to get good results from launchpad
fumanchu182
Well here is the deal.
For two weeks I haven't been able to watch flash on my Linux desktop.
So I am either A.) Going back to Windows. or B.) Not using Linux anymore. Either way I will get my flash working agian.
err again.
I sat here for an hour and a half trying to submit a stupid but to launchpad and the damn thing fails to not time out.
I have learned more about gdb then I care to know but yet still have no backtrace nor the ability to watch videos in full screen with flash.
kbrosnan
did you try the http://www.mozilla.com build it can catch most crashes and provide a crash report
i suspect #launchpad would be a better place to express the launchpad isssues you are having
i rarely use it
gpathy
when we do mousedown on a page and move the mouse, underlying text will be selected. Is this possible to achieve the same, with javascript program? i want to write a js program, where if i pass two points p1(x1,y1) and p2(x2,y2), the underlying text should be selected. Is there any API for that?
Cork
gpathy: there is code for it, but its not available to content javascript
gpathy
Cork,not available to content javascript ? i want to select the text,which be displayed on webpage/html page when i open with firefox.
Cork
gpathy: there is content javascript (things served on a website), and chrome javascript
chrome javascript is used in extensions, and the core functions of firefox
outside of the chrome, you can'r do anything with selection
*can't
killerchicken
When I launch Firefox from OS X in a new location, it starts up, flashes some window for a second, and immediately quits again. Then it relaunches itself and works normally. What is going on there and why is Firefox relaunching itself?
Is there any way at all to prevent the relaunching?
Thanks for your consideration!
gpathy
Cork, oh ok thanks...Actually i am using a custom app -which uses gtkmozembed- for rendering the html pages. I am trying to emulate "selection" by dispatching ( ele.dispatchEvent() ) 'mousedown' , mousemove' and 'mouseup' signals, on <body> element. Do you think, it can be achieved in this way?
Cork
no
gpathy
Cork, :( that means,simply, it can not be done at all(?)
Cork
let me see if i can find the code
gpathy
ok
Cork
interesting, looks like they have added it to the dom https://developer.mozilla.org/en/DOM/document.createRange
give it a try, it might work
gpathy
Cork, createRange() works differently, it needs us to pass xpath(startNode,startoffset, endNode and endoffset) of start position and end position, but in this case i can pass only points, p1(x1,y1) and p2(x2,y2).
Cork
you mean graphical points?
gpathy
Cork, yes, graphical points
Cork
gpathy: pm
alkisg
bholley, dolske: I filed a bug about the Xorg memory problem that you were helping me yesterday. Thank you. https://bugzilla.mozilla.org/show_bug.cgi?id=575523
dolske: I filed a bug about the Xorg memory problem that you were helping me yesterday. Thank you. https://bugzilla.mozilla.org/show_bug.cgi?id=575523
kraftb
hello !
i have a question regarding the new thunderbid version. is this the correct place to ask?
there is a channel named "thunderbird" but noone there ...
my question: I just noticed the new "attachment reminder" featuer of thunderbid. When you type the word "attachment" or "attached" like in the sentence "I have attached the solution for the problem ... blabla" then it pops up a nice box reminding you to add the file
but sadly I write messages in german AND english. So I can configure it to either use the english dictionary and recognize words like "attachment" oder "attached" or configure it to use the german dictionary and recognize words like "anhang" or similar
is there a solution to react on both dictionaries ?
ok. I just tried installing a german dictionary but it still just recognizes "attachment"
ok. Just found it. Remember: altavista longer before you ask dumb questions on an IRC channel
Edit > Preferences > Composition > General > Mark checkbox: Check for missing attachments
lint
how do i get rid of an empty bookmark bar?
Thomas`-
hey guys which ajax library does firefox use ?
Anomie21
Is there no decent spam extension for thunderbird?
You should be able to right click -> block sender like in Outlook :/
Saturn2888
Where is the thunderbird channel?
Cork
#thunderbird @ irc.mozilla.org
Saturn2888
found it
different server
laeg
the search bar not sorting alphabetically is ridiculous, and has been nagging at me a long time, is there a fix? the addon i found has not been maintained.
joeym
I just updated to firefox 3.6.4 on Freebsd 8.0. The locations where plugins are searched for often seems to change with different freebsd ports of firefox. Is there a way to tell what directories firefox is searching for plugins?
kbrosnan
joeym: plugins are always checked when they are in ~/.mozilla/plugins and /path/to/firefox/plugins , other locations such as /usr/lib/mozilla/plugins can be added by the package maintainer
joeym
kbrosnan: thanks
AivarasKivilius
Then firefox 3.6.6 was relised?
kbrosnan
yes
AivarasKivilius
No, I mean how many hours ago?
Mentifisto
3.6.6 is the latest, stable version? would it possibly fix the places.sqlite problem?
luckz
is there a about:config option or other command to force 4.0b2pre to load "incompatible" extensions?
« prev 1 2 next »