Site lock-ups with Firefox 3.5.x ?

[Edward J. Shornock-3]

Does anyone else have a problem with Firefox locking on artist pages
that have a crap load of releases, such as the WA Mozart
artist page
<http://musicbrainz.org/artist/b972f589-fb0e-474e-b64a-803b0364fa75.html>,
or even J.S. Bach
<http://musicbrainz.org/artist/24f1766e-9635-4d58-a4d4-9413f9f98a4c.html>?

I've been able to successfully reproduce this problem with
- Debian's Iceweasel 3.5.x x64
- Firefox 3.5RC3 x86
- A self-compiled 3.5.0 x64
- Firefox 3.5.1 running under Wine
- Firefox 3.5.1 running inside of WinXP in VMWare.

What I experience is that the page(s) load and the browser is
responsive, for about all of 10 seconds. Afterwards, the browser
locks--hard. After waiting for an average of 50 seconds, the browser
will respond normally again. The lock-ups also occur when using the
parameter "-safe-mode".

I cannot reproduce the lock-ups with the official Linux Firefox (x86)
3.0.12,
nor the Debian (x64) Iceweasel derivitive. I also cannot reproduce the
lock-ups when I'm not logged in. Disabling JavaScript does not help, nor
does creating a new Firefox/Iceweasel profile.

At the time of lock-up, strace shows (being a non-programmer, this means
nothing to me):
"madvise(0x7f79ad7f0000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79ad7ed000, 8192, MADV_DONTNEED) = 0
madvise(0x7f79ad7eb000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79ad7e7000, 8192, MADV_DONTNEED) = 0
madvise(0x7f79ad7e5000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79ad7e3000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79ad7df000, 12288, MADV_DONTNEED) = 0
madvise(0x7f79b0476000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b0473000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b0471000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b03f1000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b03ee000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b03eb000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b03e9000, 4096, MADV_DONTNEED) = 0"

It seems like a regression in the new Firefox, and I wanted to confirm
it here before creating a bugzilla account to report it.


Ed / Crazee_Canuck


_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

[Brian Schweitzer]

This sounds familiar to what I've heard for other sites (see http://support.mozilla.com/tiki-view_forum_thread.php?locale=no&forumId=1&comments_parentId=383802 for an example).  Given the number of links on any page, esp a page like the Mozart listing where there's at least several hundred links, the link prefetch may very well be the problem.  Have you tried disabling it to see if the same behaviour remains?

Brian

On Sat, Jul 25, 2009 at 4:49 PM, Edward J. Shornock <ed.shornock@...> wrote:

Does anyone else have a problem with Firefox locking on artist pages
that have a crap load of releases, such as the WA Mozart
artist page
<http://musicbrainz.org/artist/b972f589-fb0e-474e-b64a-803b0364fa75.html>,
or even J.S. Bach
<http://musicbrainz.org/artist/24f1766e-9635-4d58-a4d4-9413f9f98a4c.html>?

I've been able to successfully reproduce this problem with
- Debian's Iceweasel 3.5.x x64
- Firefox 3.5RC3 x86
- A self-compiled 3.5.0 x64
- Firefox 3.5.1 running under Wine
- Firefox 3.5.1 running inside of WinXP in VMWare.

What I experience is that the page(s) load and the browser is
responsive, for about all of 10 seconds. Afterwards, the browser
locks--hard. After waiting for an average of 50 seconds, the browser
will respond normally again. The lock-ups also occur when using the
parameter "-safe-mode".

I cannot reproduce the lock-ups with the official Linux Firefox (x86)
3.0.12,
nor the Debian (x64) Iceweasel derivitive. I also cannot reproduce the
lock-ups when I'm not logged in. Disabling JavaScript does not help, nor
does creating a new Firefox/Iceweasel profile.

At the time of lock-up, strace shows (being a non-programmer, this means
nothing to me):
"madvise(0x7f79ad7f0000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79ad7ed000, 8192, MADV_DONTNEED) = 0
madvise(0x7f79ad7eb000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79ad7e7000, 8192, MADV_DONTNEED) = 0
madvise(0x7f79ad7e5000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79ad7e3000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79ad7df000, 12288, MADV_DONTNEED) = 0
madvise(0x7f79b0476000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b0473000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b0471000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b03f1000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b03ee000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b03eb000, 4096, MADV_DONTNEED) = 0
madvise(0x7f79b03e9000, 4096, MADV_DONTNEED) = 0"

It seems like a regression in the new Firefox, and I wanted to confirm
it here before creating a bugzilla account to report it.


Ed / Crazee_Canuck

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEYEARECAAYFAkprb80ACgkQYy325GUFAIxk0wCbBFAQwhZCBgjCwKoTZLJ2NxFM
VioAoKs4I3V5ulr/ddLnGgQ94ZeR8Ry+
=9vGB
-----END PGP SIGNATURE-----

_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

[Age Bosma-2]

Edward J. Shornock wrote:
>
> It seems like a regression in the new Firefox, and I wanted to confirm
> it here before creating a bugzilla account to report it.
>

Confirmed on Windows XP SP3 (native) with Firefox 3.5.1

Yours,

Age

_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

[Chad Wilson-2]

Edward J. Shornock wrote:

> Does anyone else have a problem with Firefox locking on artist pages
> that have a crap load of releases, such as the WA Mozart
> artist page
> <http://musicbrainz.org/artist/b972f589-fb0e-474e-b64a-803b0364fa75.html>,
> or even J.S. Bach
> <http://musicbrainz.org/artist/24f1766e-9635-4d58-a4d4-9413f9f98a4c.html>?
>
> I've been able to successfully reproduce this problem with
> - Debian's Iceweasel 3.5.x x64
> - Firefox 3.5RC3 x86
> - A self-compiled 3.5.0 x64
> - Firefox 3.5.1 running under Wine
> - Firefox 3.5.1 running inside of WinXP in VMWare.
>
> What I experience is that the page(s) load and the browser is
> responsive, for about all of 10 seconds. Afterwards, the browser
> locks--hard. After waiting for an average of 50 seconds, the browser
> will respond normally again. The lock-ups also occur when using the
> parameter "-safe-mode".
>
> I cannot reproduce the lock-ups with the official Linux Firefox (x86)
> 3.0.12,
> nor the Debian (x64) Iceweasel derivitive. I also cannot reproduce the
> lock-ups when I'm not logged in. Disabling JavaScript does not help, nor
> does creating a new Firefox/Iceweasel profile.
>
>  

Confirmed on Windows 7 32-bit, FFX 3.5.1.

Firefox becomes completely unresponsive for 20-30 seconds or so, rather
than 50. Processor/OS/network speed difference with your env I suppose?
Bach is 5 or 6 seconds slower than Mozart :p

Also cannot reproduce on FFX 3.0.

Disabling link prefetching as suggested by Brian (using about:config
network.prefetch-next=false) did not fix the problem unfortunately.

Chad / voice

_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

[Edward J. Shornock-3]

* Chad Wilson <chad.wilson@...> [26-07-2009 13:30 EEST]:
> Edward J. Shornock wrote:
> > Does anyone else have a problem with Firefox locking on artist pages
> > that have a crap load of releases, such as the WA Mozart
> > artist page
> > <http://musicbrainz.org/artist/b972f589-fb0e-474e-b64a-803b0364fa75.html>,
> > or even J.S. Bach
> > <http://musicbrainz.org/artist/24f1766e-9635-4d58-a4d4-9413f9f98a4c.html>?

> Confirmed on Windows 7 32-bit, FFX 3.5.1.

I've also since gotten access to a WinXP, SP3 box. I successfully
reproduced it there as well.

> Firefox becomes completely unresponsive for 20-30 seconds or so, rather
> than 50. Processor/OS/network speed difference with your env I suppose?
> Bach is 5 or 6 seconds slower than Mozart :p
>
> Also cannot reproduce on FFX 3.0.
>
> Disabling link prefetching as suggested by Brian (using about:config
> network.prefetch-next=false) did not fix the problem unfortunately.

I just found that it also happens if one saves the page as an HTML file
and attempts to load that, even without a network connection (I've
disconnected the CAT5 cable, set the browser to "Work Offline", and
tried using a proxy at "0.0.0.0" to confirm). The hanging duration is
fairly consistent, lasting for about 20-30 seconds or so, as Chad said.

At the time of the posting last night, my box was pegged by software
compilation which I'm fairly certain was responsible for the longer
durations that I experienced.

Thank you (everyone!) for your confirming that it's not just me.

Ed / crazee_Canuck


_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

[Edward J. Shornock-3]

Now I've also reproduced the freezing with
Build identifier: Mozilla/5.0 (X11; U; Linux x86_64; en-US;
rv:1.9.1.1pre) Gecko/20090701 SeaMonkey/2.0b1pre

...and here's the (badly filed?) bug report:
https://bugzilla.mozilla.org/show_bug.cgi?id=506553


_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

[Edward J. Shornock-3]

* Edward J. Shornock <ed.shornock@...> [26-07-2009 19:37 EEST]:
> Now I've also reproduced the freezing with
> Build identifier: Mozilla/5.0 (X11; U; Linux x86_64; en-US;
> rv:1.9.1.1pre) Gecko/20090701 SeaMonkey/2.0b1pre
>
> ...and here's the (badly filed?) bug report:
> https://bugzilla.mozilla.org/show_bug.cgi?id=506553

...and it was closed as a dup.
https://bugzilla.mozilla.org/show_bug.cgi?id=477564

Both the about:config workaround, and the patch to the source code,
fixed the problem for me.


_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

[Chad Wilson-2]

Edward J. Shornock wrote:

> * Edward J. Shornock <ed.shornock@...> [26-07-2009 19:37 EEST]:
>  
>> Now I've also reproduced the freezing with
>> Build identifier: Mozilla/5.0 (X11; U; Linux x86_64; en-US;
>> rv:1.9.1.1pre) Gecko/20090701 SeaMonkey/2.0b1pre
>>
>> ...and here's the (badly filed?) bug report:
>> https://bugzilla.mozilla.org/show_bug.cgi?id=506553
>>    
>
> ...and it was closed as a dup.
> https://bugzilla.mozilla.org/show_bug.cgi?id=477564
>
> Both the about:config workaround, and the patch to the source code,
> fixed the problem for me

Hmm, radio buttons appear to be a major problem too, but not
specifically mentioned on that bug. I imagine the same cause though.
Does
http://musicbrainz.org/edit/track/change.html?trackid=1406535&releaseid=143908&tindex=4&solo=1 
work correctly on your patched/workarounded version?

It actually gets stuck twice for me... frees up after 30 seconds for a
few sconds and then locks up again.

Chad

_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

[Edward J. Shornock-3]

* Chad Wilson <chad.wilson@...> [28-07-2009 20:19 EEST]:
> Edward J. Shornock wrote:
> Hmm, radio buttons appear to be a major problem too, but not
> specifically mentioned on that bug. I imagine the same cause though.
> Does
> http://musicbrainz.org/edit/track/change.html?trackid=1406535&releaseid=143908&tindex=4&solo=1 
> work correctly on your patched/workarounded version?

It works fine with the patched version. It freezes for me with the
unpatched 3.5.x. The workaround posted in the bug
(browser.sessionstore.privacy_level to 2) solves the freezing as well.





_______________________________________________
MusicBrainz-users mailing list
MusicBrainz-users@...
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users