I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter
...and choosing one of the two newsletters, linked there.
I'm writing for a couple reasons.
I'm curious to see if this is only my problem, or if other have experienced it or can replicate it.
And I was also wondering how I can remove Firefox 3.5 and its associated support files and get back to the regular repository.
Firefox 3.5, for the most part, is very stable and is snappier than 3.0.11 -- only two sites have pages have crashed for me, but when it crashes it takes down my whole system and that's not good.
And, if someone could mention this on the forum (there is a thread about Firefox 3.5) I would appreciate it. I can't log on to the forums. I am supposed to ask for another password, but I can't remember what my original email address was (2 or 3 years ago).
Thanks for any pointers.
At Sat, 11 Jul 2009 18:23:18 -0500 CentOS mailing list centos@centos.org wrote:
I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter
...and choosing one of the two newsletters, linked there.
I'm writing for a couple reasons.
I'm curious to see if this is only my problem, or if other have experienced it or can replicate it.
I have heard about various problems people over in MS-Windows land have had with Firefox 3.5. It appears that FF 3.5 is not quite ready for production systems.
And I was also wondering how I can remove Firefox 3.5 and its associated support files and get back to the regular repository.
rpm -hUv --oldpackage firefox-3.0.10-<mumble>.i386.rpm
Were there other dependcies? If so, you may have roll them back too.
Firefox 3.5, for the most part, is very stable and is snappier than 3.0.11 -- only two sites have pages have crashed for me, but when it crashes it takes down my whole system and that's not good.
And, if someone could mention this on the forum (there is a thread about Firefox 3.5) I would appreciate it. I can't log on to the forums. I am supposed to ask for another password, but I can't remember what my original email address was (2 or 3 years ago).
Thanks for any pointers.
Ron Blizzard wrote:
I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter
...and choosing one of the two newsletters, linked there.
I'm writing for a couple reasons.
I'm curious to see if this is only my problem, or if other have experienced it or can replicate it.
Confirmed - also i386 mharris packaging.
Tring x86_64 (rebuild of his src.rpm) momentarily ...
Michael A. Peters wrote:
I'm writing for a couple reasons.
I'm curious to see if this is only my problem, or if other have experienced it or can replicate it.
Confirmed - also i386 mharris packaging.
update - it didn't actually bring the OS down, it brought X11 down - though it looked like CentOS was down. Pressing the power button (thinkpad) resulted in X11 restarting followed by a clean shutdown.
Michael A. Peters wrote:
Michael A. Peters wrote:
I'm writing for a couple reasons.
I'm curious to see if this is only my problem, or if other have experienced it or can replicate it.
Confirmed - also i386 mharris packaging.
update - it didn't actually bring the OS down, it brought X11 down - though it looked like CentOS was down. Pressing the power button (thinkpad) resulted in X11 restarting followed by a clean shutdown.
My own i386 build of firefox also does not crash (built when I did x86_64 build but I never installed it until now).
Only difference in spec file was I added pcre-devel to BuildRequires before building it - but the my build also has all the language packs.
Not sure if that's the difference or not.
Michael A. Peters wrote:
update - it didn't actually bring the OS down, it brought X11 down -
What happens if you add the following to Section "Device" in /etc/X11/xorg.conf?
Section "Device" Option "AccelMethod" "XAA" Option "XAANoOffscreenPixmaps" "true" EndSection
I am using firefox 3.5 from www.mozilla.com without any problems. But without the code above the Xorg process uses 100% cpu on some www sites. In my opinion it is not a firefox but an xorg bug.
regards Olaf
On Sun, Jul 12, 2009 at 12:33 AM, Michael A. Petersmpeters@mac.com wrote:
Michael A. Peters wrote:
I'm writing for a couple reasons.
I'm curious to see if this is only my problem, or if other have experienced it or can replicate it.
Confirmed - also i386 mharris packaging.
update - it didn't actually bring the OS down, it brought X11 down - though it looked like CentOS was down. Pressing the power button (thinkpad) resulted in X11 restarting followed by a clean shutdown.
Yeah, the first time it did this it dropped to my login screen -- and, I think, once I was able to ALT-CTRL-Backspace out of it. But every other time it just locked and it seemed like I lost my keyboard and everything. Though it didn't reboot on its own, so I guess something was still working.
Michael A. Peters schrieb:
Michael A. Peters wrote:
I'm writing for a couple reasons.
I'm curious to see if this is only my problem, or if other have experienced it or can replicate it.
Confirmed - also i386 mharris packaging.
update - it didn't actually bring the OS down, it brought X11 down - though it looked like CentOS was down. Pressing the power button (thinkpad) resulted in X11 restarting followed by a clean shutdown.
If you can bring down X11 with a userspace program then this is not an issue of your application but of X11. It means there is a bug in X11. I guess you were doing this as a normal user and not as root?
Rainer
On Sun, Jul 12, 2009 at 12:19 AM, Michael A. Petersmpeters@mac.com wrote:
Ron Blizzard wrote:
I'm curious to see if this is only my problem, or if other have experienced it or can replicate it.
Confirmed - also i386 mharris packaging.
Tring x86_64 (rebuild of his src.rpm) momentarily ...
Thanks.
Ron Blizzard wrote:
I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter
...and choosing one of the two newsletters, linked there.
I'm writing for a couple reasons.
I'm curious to see if this is only my problem, or if other have experienced it or can replicate it.
My x86_64 build does not crash on the newsletters, interestingly enough.
Ron Blizzard wrote:
I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter
...and choosing one of the two newsletters, linked there.
FWIW (about $0.0002), Firefox 3.5 on Windows XP (32bit) doesn't crash on either newsletter.
John R Pierce wrote:
Ron Blizzard wrote:
I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter
...and choosing one of the two newsletters, linked there.
FWIW (about $0.0002), Firefox 3.5 on Windows XP (32bit) doesn't crash on either newsletter.
Nor does Firefox 3.5 running on Fedora 11. No problem at all.
+0.0002 ;-)
Robert Nichols wrote:
John R Pierce wrote:
Ron Blizzard wrote:
I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter
...and choosing one of the two newsletters, linked there.
FWIW (about $0.0002), Firefox 3.5 on Windows XP (32bit) doesn't crash on either newsletter.
Nor does Firefox 3.5 running on Fedora 11. No problem at all.
+0.0002 ;-)
My suspicion is that it has to do with the language packs, since the same src.rpm built w/o disabling language packs doesn't crash in CentOS 5.3 on either arch.
<body lang="en" dir="ltr">
I wonder if that is what triggers it - IE a crash when FireFox can't find the language pack for "en".
Maybe try fetching the page with wget, removing that, and see what happens?
Also of interest - the mh build gave a Fedora start page, the build I did gave a CentOS start page. Maybe his mock config didn't define the rhel macro, so the build did some thing the Fedora way.
On Sun, Jul 12, 2009 at 5:46 AM, Michael A. Petersmpeters@mac.com wrote:
Robert Nichols wrote:
John R Pierce wrote:
Ron Blizzard wrote:
How about using GNU's IceCat? I use it alongside Firefox, but none of the settings intermix.
Scott
Michael A. Peters wrote:
My suspicion is that it has to do with the language packs, since the same src.rpm built w/o disabling language packs doesn't crash in CentOS 5.3 on either arch.
I feel like such a dope.
I forgot to run createrepo on my private i386 repo so when I removed the mh built 3.5 and yum installed firefox to get what I built, it grabbed 3.0.11 and I didn't even notice.
I realized this when I looked at the spec file and saw there was no way in hot firey place that the default start page could be anything other than the fedora start page.
Since I don't feel like crashing my laptop, I won't check it now - but I *don't* know that my i386 ff35 build works with the crash page, I only know the x86_64 works - and it probably isn't language packs, the language packs aren't in my x86_64 or i386 builds.
I thought they were because I saw a bunch of language packs after the i386 yum install that grabbed 3.0.11 that I though grabbed 3.5.
Arrgggh.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Michael A. Peters wrote:
Michael A. Peters wrote:
My suspicion is that it has to do with the language packs, since the same src.rpm built w/o disabling language packs doesn't crash in CentOS 5.3 on either arch.
I doubt that is the case. The language packs are optional plugins essentially, which the Fedora builds include in the rpm directly, but which upstream does not include by default. They are completely non-essential and the majority of people using Firefox do not have them installed - except those using distributions that include them forcibly such as Fedora/RHEL/CentOS. I prefer to not have the language packs by default and let people who need them get them themselves from the mozilla site.
What is more likely the case, is that the act of rebuilding it changed something, and I would bet money that if you forcibly disable the language packs and rebuild it, that you'll end up with something that still works, because something else is changing during rebuild.
I build for the i686 target for example, which has to be manually specified. If someone does a straight rebuild they'll get i386, and that could potentially change things for them. There are lots of other variables too.
Ultimately though, the most sensible thing to do is to get it built using mock in standardized clean buildroots, with all dependencies properly specified and vetted. That will come in the future.
I feel like such a dope.
I forgot to run createrepo on my private i386 repo so when I removed the mh built 3.5 and yum installed firefox to get what I built, it grabbed 3.0.11 and I didn't even notice.
hehe
I realized this when I looked at the spec file and saw there was no way in hot firey place that the default start page could be anything other than the fedora start page.
I should probably update the package to point to CentOS defaults in the future too. I was just in a hurry for firefox 3.5 chocolatey goodness that I didn't want to muck around with it excessively right away to try and win any packager of the year awards or anything like that. hehe
Since I don't feel like crashing my laptop, I won't check it now - but I *don't* know that my i386 ff35 build works with the crash page, I only know the x86_64 works - and it probably isn't language packs, the language packs aren't in my x86_64 or i386 builds.
Hard to say... I haven't had firefox crash on any systems I've installed my builds on yet. That doesn't mean there isn't a problem with it though. Until I get some cleaner builds done in mock on a sterile build environment, I consider these builds somewhat "experimental, use at own risk" material. ;o) I've got other things on the go right now though so I haven't looked into it yet as my firefox is working good. When 3.5.1 comes out I'll probably take a stab at it again, and perhaps get it through Karan's buildsystem.
I thought they were because I saw a bunch of language packs after the i386 yum install that grabbed 3.0.11 that I though grabbed 3.5.
Naw, as mentioned above, the language packs are optional components that are normally not supplied with firefox, and you have to go to the mozilla website to download them. Mozilla.org distributes them like that so that you only get the language datasets for the languages you require, without bloating the downloads. To the best of my knowledge, all of the mozilla.org builds of firefox for all OS platforms exclude the langpacks by default for this reason. Fedora/RH just include them by default to be ready to use out of the box for the widest number of people out there regardless of what languages they use, and without requiring them to configure anything. The tradeoff is that it is "simpler", but wastes more disk space, network bandwidth, etc.
In the future, if I can figure out an easy way to supply the language packs in a separate src.rpm package, I'll probably provide an rpm for that too, but the majority of English-only users out there don't need them, and it shouldn't have any effect on runtime stability or any dependencies.
Hope this helps.
- -- Mike A. Harris http://mharris.ca | https://twitter.com/mikeaharris
On Sun, Jul 12, 2009 at 12:53 AM, John R Piercepierce@hogranch.com wrote:
FWIW (about $0.0002), Firefox 3.5 on Windows XP (32bit) doesn't crash on either newsletter.
I tried it in Windows 2000 in VirtualBox and it worked fine -- I should have mentioned that.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Ron Blizzard wrote:
(followup on original post from previous reply)
I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
That is definitely not a firefox bug.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter ...and choosing one of the two newsletters, linked there.
I've just visited that page and clicked on both links and it works fine. I've even disabled all of my security/privacy extensions on that page and it still works fine. If the page is causing your system to crash, I'd start examining /var/log/messages and your X server log for kernel panics or Oops et al.
- -- Mike A. Harris http://mharris.ca | https://twitter.com/mikeaharris
On Thu, Jul 16, 2009 at 7:09 PM, Mike A. Harrismharris@mharris.ca wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Ron Blizzard wrote:
(followup on original post from previous reply)
I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
That is definitely not a firefox bug.
Okay. But there is some kind of issue between Firefox 3.5 and Xorg. I reverted back to Firefox 3.0.11 and it works fine.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter ...and choosing one of the two newsletters, linked there.
I've just visited that page and clicked on both links and it works fine. I've even disabled all of my security/privacy extensions on that page and it still works fine. If the page is causing your system to crash, I'd start examining /var/log/messages and your X server log for kernel panics or Oops et al.
Interesting. I'll see what I can find in /var/log/messages and share anything that might point to the problem. Thanks for writing.
On Fri, Jul 17, 2009 at 12:56 AM, Ron Blizzardrb4centos@gmail.com wrote:
Interesting. I'll see what I can find in /var/log/messages and share anything that might point to the problem. Thanks for writing.
Here's what the logs show...
Jul 17 00:31:51 localhost last message repeated 3 times Jul 17 00:32:01 localhost gconfd (ronb-2823): starting (version 2.14.0), pid 2823 user 'ronb' Jul 17 00:32:01 localhost gconfd (ronb-2823): Resolved address "xml:readonly:/etc/gconf/gconf.xml.mandatory" to a read-only configuration source at position 0 Jul 17 00:32:01 localhost gconfd (ronb-2823): Resolved address "xml:readwrite:/home/ronb/.gconf" to a writable configuration source at position 1 Jul 17 00:32:01 localhost gconfd (ronb-2823): Resolved address "xml:readonly:/etc/gconf/gconf.xml.defaults" to a read-only configuration source at position 2 Jul 17 00:32:04 localhost gconfd (ronb-2823): Resolved address "xml:readwrite:/home/ronb/.gconf" to a writable configuration source at position 0 Jul 17 00:32:05 localhost pcscd: winscard.c:304:SCardConnect() Reader E-Gate 0 0 Not Found Jul 17 00:32:05 localhost last message repeated 4 times Jul 17 00:32:06 localhost nm-system-settings: Loaded plugin ifcfg-rh: (c) 2007 - 2008 Red Hat, Inc. To report bugs please use the NetworkManager mailing list. Jul 17 00:32:06 localhost nm-system-settings: ifcfg-rh: parsing /etc/sysconfig/network-scripts/ifcfg-eth0 ... Jul 17 00:32:06 localhost nm-system-settings: ifcfg-rh: read connection 'System eth0' Jul 17 00:32:06 localhost nm-system-settings: ifcfg-rh: parsing /etc/sysconfig/network-scripts/ifcfg-lo ... Jul 17 00:32:06 localhost nm-system-settings: ifcfg-rh: error: Ignoring loopback device config. Jul 17 00:32:56 localhost kernel: e1000: eth0: e1000_watchdog_task: NIC Link is Up 100 Mbps Full Duplex, Flow Control: None Jul 17 00:47:20 localhost gconfd (root-3405): starting (version 2.14.0), pid 3405 user 'root' Jul 17 00:47:20 localhost gconfd (root-3405): Resolved address "xml:readonly:/etc/gconf/gconf.xml.mandatory" to a read-only configuration source at position 0 Jul 17 00:47:20 localhost gconfd (root-3405): Resolved address "xml:readwrite:/root/.gconf" to a writable configuration source at position 1 Jul 17 00:47:20 localhost gconfd (root-3405): Resolved address "xml:readonly:/etc/gconf/gconf.xml.defaults" to a read-only configuration source at position 2 Jul 17 00:48:21 localhost gconfd (root-3405): GConf server is not in use, shutting down. Jul 17 00:48:21 localhost gconfd (root-3405): Exiting Jul 17 01:31:11 localhost gconfd (root-3889): starting (version 2.14.0), pid 3889 user 'root' Jul 17 01:31:11 localhost gconfd (root-3889): Resolved address "xml:readonly:/etc/gconf/gconf.xml.mandatory" to a read-only configuration source at position 0 Jul 17 01:31:11 localhost gconfd (root-3889): Resolved address "xml:readwrite:/root/.gconf" to a writable configuration source at position 1 Jul 17 01:31:11 localhost gconfd (root-3889): Resolved address "xml:readonly:/etc/gconf/gconf.xml.defaults" to a read-only configuration source at position 2
In this time period Xorg crashed twice. Earlier in the log (on July 15th and 14th) I have similar records (which correlate with the other crashes) but there is no "Exiting" line. These were times where I had to reboot to get the machine back into a working state.
It looks like it has something to do with an XML lock?
Mike A. Harris wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Ron Blizzard wrote:
(followup on original post from previous reply)
I downloaded Firefox 3.5 from the M. Harris site and, for the most part, have had good luck with it. But I have also had hard crashes that take down CentOS, not just Firefox.
That is definitely not a firefox bug.
It happened to me twice on eBay (on the same page) -- and now I can replicate it as many times as I want by going to...
http://wiki.centos.org/Newsletter ...and choosing one of the two newsletters, linked there.
I've just visited that page and clicked on both links and it works fine. I've even disabled all of my security/privacy extensions on that page and it still works fine. If the page is causing your system to crash, I'd start examining /var/log/messages and your X server log for kernel panics or Oops et al.
Where it happens with me is the i386 build on a Thinkpad T20. I'll have to check the driver.
It does not happen on my x86_64 running proprietary nvidia driver.
-
John R Pierce -
Michael A. Peters -
Mike A. Harris -
Olaf Mueller -
Rainer Traut -
Robert Heller -
Robert Nichols -
Ron Blizzard -
Scott Ehrlich