firefox continuously core dumping

hruodr said:
PMc, sidetone mentioned a bug that was fixed on 2018-08-17, but I have a FreeBSD 11.2 from 2018-06-22 (r335510). I think, the solution is to wait a little.
Click to expand...

Aye, that would appear in 11.3. I put those recommended buffer space adjustments in. Lets see...
 
For my laptop with 11.2-RELEASE-p8 I have no problem with FF. But since last few months I had same core dump for www/palemoon & FF on my desktop with port system. Have build FF wih same setting of FF package but without improvement. I noticed that after opening of 2/3 tab the next tab crashed in FF. Then that sysctl setting started appearing from update of FF without any result. After disabling the swap partition I had no problem so far. Can not say any reason.
 
I'm stuck doing some development work on Chromium right now. I left FF open and tried to make myself do as much there as I could last night--about five hours--and had no issues but some of you say you leave FF on all the time. When I develop, I do the same but it depends on the browser I'm testing in or start with.
 
Been running my "test" build of www/firefox all morning with between 13 and 21 tabs open to various sites, mainly the junk on FF's front page when you first open it. I am randomly closing the browser and letting the warning about "you have 1 billion tabs open" come up. No logins, no configuration, nothing, just the build I mentioned in a prev post.

No core dumps so far. Memory is 4-5gb used but I have www/chromium up as well.
 
For me Firefox always seems to come up clean with all tabs intact and no problems. However, it seems like it's some script on some website that causes it to lock up and I have to kill it. I don't know what it is. It has never crashed by itself.

There may indeed be a significant difference between the situation with people who close FF sometimes and those of us who leave it running indefinitely. I have a dozen or more windows with FF, and many hundred tabs in those, so I know I'm pushing it.
 
I can browse this forum all day without closing Firefox. When I look at my statistics from "top" they are consistent and don't seem to show any obvious memory leaks. The same is essentially true of freshports.org, wikpedia.org, wictionary.org, google.com, my own localhost and local area network php development sites, and most of the other sites I use frequently. But if I go to some of the other less well-behaved sites I use every now and then, like facebook.com in particular, then when I check "top" I can see that my memory usage will remain obscenely bloated, even after I've closed the tab, and it won't completely recover until I've closed down Firefox completely and restarted it.
 
For what it's worth, I removed the config lines from /etc/sysctl.conf that I mentioned earlier in this thread and have actually not had it crash at all. So it does, at least for me, appear that this isn't an issue on 12.0 like it was on 11.x For the others that still have an issue, it could be a dodgy extension or something? I personally only have ublock origin. That's the only extension I have installed.
 
xtaz said:
For what it's worth, I removed the config lines from /etc/sysctl.conf that I mentioned earlier in this thread and have actually not had it crash at all. So it does, at least for me, appear that this isn't an issue on 12.0 like it was on 11.x
Click to expand...

That sounds correct. As far as I glanced over the commitlogs, the fix was introduced after 11.2, and well before 12.0..
 
I can now confirm that the above mentioned remedy (increasing net.local.stream.send/recvspace) does NOT solve my coredump issue: they continue to appear - and they appear long after firefox has been terminated, probably at system shutdown.

What happens is: when firefox is ended, after closing all windows one firefox process does stay and sleep - it does nothing. Then wenn trying to start firefox anew, this process writes a coredump. (Qualified as: the usual crap, not worth debugging.)
 
Same here. Firefox hangs as soon as i open heavy websites (i.e. Youtube and Outlook). I use uBlock Origin and HttpsEverywhere. Desktop Computer with AMD A4-5300 APU and 8 GB of RAM, FreeBSD 12.0 x64 as operating system.
PS: after using truss to track system calls it seems to be a d-bus related issue. Firefox fails to read the machine id which is supposed to be stored in /etc/machine-id or /var/lib/dbus/machine-id.
Possible fix
Gain root privileges:
su
then type
dbus-uuidgen >> /var/lib/dbus/machine-id
to generate (and store) a valid machine id.
Close Firefox and launch it again, now the problematic websites should work fine. Now Youtube and Outlook works without any issue, even when open at the same time.
 
I just updated www/firefox to the latest version in ports.

The first thing I did was opened a window here, then opened another to youtube and my tab crashed and core dumped. I tried again as I write and it crashed 6 times in a row with this salient message:

6106

I updated to the latest version in hopes it would solve this problem but seems to have made it worse.

In addition, it has developed the habit of taking a shot every time you bookmark a page, making it, to me, potentially:

6107

Maybe I don't want a shot taken of every site I visit, "Godzilla".

I get better performance out of www/firefox-esr and moving to that on all my machines. It doesn't crash and knows its role.
 
I've not been following this thread but Trihexagonal images caught my eye.

On my workstation at home, I have absolutely no problems with Firefox but, yesterday, I was at a friend's house who, for fun, tried to install Clear Linux. I believe he got that exact same error message after installing Firefox. (He can't get Linux to work properly either.)
 
It appears that dbus is not the real cause of Firefox core dumps. In my case setting up dbus properly solved the issue, but this fix cannot apply for PMc who doesn't run dbus. It is weird! Can someone else start firefox with truss -o filename_of_your_choice firefox, make it crash and share the result of tail -n 100 filename_of_your_choice or tail -n 100 filename_of_your_choice | grep -i dbus?
 
Code: 
$ truss -o /usr/home/jitte/Documents/funkfox01.txt firefox
truss: Unable to enable LWP events for pid 4818: No such process
$

funkfox.txt was created but blank once I got it to crash.
 
MathematicsStudent said:
It appears that dbus is not the real cause of Firefox core dumps. In my case setting up dbus properly solved the issue, but this fix cannot apply for PMc who doesn't run dbus. It is weird!
Click to expand...

In my case it is most likely a different issue, as it practically never happened while running, only when terminating: it seemed to disappear orderly (I stop it explicitely before shutting down the machine), but next day the core file was there.
And this issue seems to have become significantly better after I
  1. added the above mentioned buffer adjustments,
  2. added 4G ram.
I suppose the reason is in the latter, because now the beast lives in RAM, while before it was partially paged out and had to do some paging work when terminating. Coding errors manifesting only with unusual delays at such a place are not unlikely, and difficult to pinpoint.
 
I'm using my T400 Fluxbox desktop with:

FreeBSD 11.2-RELEASE-p9
Intel Core2 Duo P8600 @ 2.40GHz
8GB RAM
RV620/M82 [Mobility Radeon HD 3450/3470]
Swap: 3979M Total, 3979M Free

It's the machine I've been using the most for some time now and was logged in 8 hours at a site last night using www/palemoon without a tab crashing and at least 6 open at all times.
 
I am on 12.0-RELEASE-p3, pkg only, plasma5-plasma and some (kde5) apps. Firefox version 65.0.1_1,1 runs fine here on HP Microserver n54l with 4GB of memory and 8GB of swap space.

I have this in /boot/loader.conf:

#
kern.ipc.shmall=32768
kern.ipc.shmseg=1024
kern.ipc.shmmni=1024
kern.maxproc=100000
#

And this in /etc/sysctl.conf:

# Enhance shared memory X11 interface
kern.ipc.shmmax=67108864
kern.ipc.shmall=32768

kern.sched.preempt_thresh=224

#kern.maxfiles=100000
hw.syscons.bell=0

kern.ipc.shm_allow_removed=1

net.local.stream.recvspace=65536
net.local.stream.sendspace=65536
#net.local.stream.recvspace=16384

Those were collected from various sources in and outside this forum.

I use some some add-ons like ublock-origin and clear cookies and cache on a regular basis.
 
You must log in or register to reply here.
Back
Top