On Wed, 27 Oct 2004, Jorgen Lundman wrote:

> I find mozilla being unstable if you compile in unmodified.
>
> I posted about the DNS settings 28/01/04, for some reason our getaddrinfo() 
> do not appear to be thread safe, or at least it can (easily) stall forever. I 
> found if I take out "_PR_HAVE_GETADDRINFO" in 
> ./nsprpub/pr/include/md/_netbsd.h it
> run _a lot_ better. Even with NetBSD 2.0 RC4.

I guess that's because you're taking out IPV6 support. The "broken DNS" 
I wrote of below involves blackholing requests for AAAA records.  (Top 
posting, huh.)

> Another issue is it frequently runs out of shmem. Not sure if it is leaking, 
> or just never releasing it. This is noticable when you view photoshop 
> competitions and so on. Increasing shmem in your kernel config (or gdb your 
> kernel) improves this. "ipcs -m" shows you used memory IIRC. This becomes a 
> bigger burden if you use X11vnc/x0rfbserver as well.
>
> Apart from those two, 1.7.3 has been very stable.
>
> Lund

That's interesting. I've already bumped shmem to 32Mb on all my X log 
hosts, so maybe that accounts for my better luck with the random crashes?

Frederick

> Frederick Bruckman wrote:
>> In article <417ED51E.8050607@mukappabeta.de>,
>> 	mkb@mukappabeta.de (Matthias Buelow) writes:
>> 
>>> Hello folks,
>>> 
>>> anyone else seeing this?
>>> 
>>> Mozilla 1.7.3 seems to be _extremely_ unstable on 2.0rc4 (i386). Here, it 
>>> crashes every couple minutes, sometimes more often, without any 
>>> recognizable pattern.  Even when it's doing trivial things, or just 
>>> sitting there and idling.  I mean, it's a long learned lesson that Mozilla 
>>> may crash every now and then, although that has become quite rare with 
>>> newer Mozilla releases, but I've never seen a situation when it is as bad 
>>> as it was now.  Not even in the heyday Netscape 4 days. 
>> 
>> 
>> I'm running 1.7.3 on a couple of different 2.0rc4 hosts without such
>> problems. The worst issue I see, is that hangs up forever on links to
>> sites with broken DNS, particularly "doubleclick.net".
>> 
>> 
>>> For comparison, the same version on FC2 Linux and WinXP runs very stable. 
>>> Unfortunately, the core file isn't really helpful, the trace always looks 
>>> more or less like the following:
>>> 
>>> #0  0x486a4feb in kill () from /usr/lib/libc.so.12
>>> #1  0x486c31a8 in raise () from /usr/lib/libc.so.12
>>> #2  0x4bd021e9 in nsProfileLock::FatalSignalHandler(int) ()
>>>    from /usr/pkg/lib/mozilla-gtk2/components/libprofile.so
>>> #3  0x4867fc73 in pthread__signal_deferred () from 
>>> /usr/lib/libpthread.so.0
>>> 
>>> Do others also experience this?  Or are my .mozilla profiles broken in 
>>> some way that cause it to misbehave so grossly?  Could it be some badness 
>>> with the new scheduler activations stuff and the pthread lib?
>> 
>> 
>> You may have a corrupted profile. I seem to remember running into
>> something like that before. Have you ever run Netscape 7 with the
>> same profile you now use for mozilla?
>> 
>> For what it's worth, since purging Netscape 7 from all my systems,
>> I haven't had any problems sharing profiles between disparate hosts
>> (Windows and NetBSD). At some point in the past, I copied the whole
>> profile directory from one to the other, and I only push around the
>> files "bookmarks.html cookies.txt *.s *.w" to keep them in sync.
>> 
>> 
>> Frederick