20894 - Mozilla does not startup with multiple kernel threads

Status: RESOLVED INCOMPLETE

(SeaMonkey :: Build Config, defect)

Description

[Cass Witkowski]

I tried starting up mozilla M11 on a 2-CPU AlphaServer 2100 running OpenVMS
V7.2-1.  After it seams to load all the files it becomes 100% CPU bound on both
CPUs.  It looks like it is going in and out of hibernation.  The SYSGEN
parameter MULTITHREAD was set to 2.  When I set MULTITHREAD to 1, mozilla
started up and I got a window.

Comment 1

[Colin Blake]

I don't know if this is an OpenVMS specific problem or not. Maybe someone
at Netscape can tell me if Mozilla (UNIX flavor) is running on multi-processor
systems yet?

In the meantime, I suggest you disable multiple kernel threads on the
mozilla-bin image:

$ set command sys$update:threadcp
$ rename mozilla.bin. m.exe
$ threadcp /disable==multiple
$ rename m.exe mozilla-bin.

Comment 2

[Colin Blake]

Mozilla on OpenVMS has not yet been qualified for multi-processor support. This
is on the list of things to do, but for now, its unsupported. Starting with M12,
the Mozilla image will be linked with multi-processor support disabled. This
will allow Mozilla to run (albeit on a single processor) on a multi-processor
system.

Comment 3

[Eli Goldberg]

Reassigning all of leger's unscreened Browser-General bugs to [e-mail]
for pre-screening and triage.

Comment 4

[Christian Mattar]

Changing Component to Threading...

Comment 5

[leger]

Sending over to Build folks to get comments

Comment 6

[Chris McAfee]

I have mozilla running fine on a dual 400 MHz Linux machine
running redhat 6.1.  I think colin should own this bug,
he seems to be the most knowledgable about VMS.

Comment 7

[Jon Granrose]

The only thing I know about VMS is restarting stuck printer queues...

reassigning to cls and cc'ing leaf in case they have any input.  otherwise, I 
agree, either colin or a unix threading guru should own this bug.

Comment 8

[cls]

Looks like it's unanimous.  Congratulations, Colin. :)

Comment 9

[Chris Yeh]

mass re-assign of all bugs where i was listed as the qa contact

Comment 10

[Colin Blake]

This is more than likely some problem in the OpenVMS pthread library. Mozilla 
isn't the only complex program that has problems when run in a multi-cpu 
environment. I've in communication with the pthread team at Compaq, but I don't 
think a solution is going to be immediate. Marking "future".

Comment 11

[Colin Blake]

I've been testing with multiple kernel threads enabled on a V7.3-1 system and
haven't run into any problems. Maybe this problem is fixed!

Comment 12

[Tin Pham]

I tried installing Mozilla 1.4 stable on my Pentium 4 2.4GHz with hyper-
threading(simulates 2 cpus) and got the same problem.

Went so far as to reimage my system to a fresh install with only XP.

Comment 13

[Mathew McBride]

Anyone been able to compile and start Mozilla on a SMP system? I'm commenting
here so I can avoid filing a possible dupe. Doing "ps aux" quickly after
executing run-mozilla.sh shows 8-10 threads, then they all disappear.

Ever since I switched from a single Duron 800@920 to 2 x Pentium II (Descutes)
350@400, I've never been able to run Mozilla on Linux. I'll take a CPU out
tomorrow and see if it works again. (I've been able to run a really old install
of Firebird though).

CC'ing myself.

Comment 14

[Chris McAfee]

I have a 2x400 rh9 system, I will pull/build the trunk and see what I get
later tonight.

Comment 15

[Mathew McBride]

Just run TestGtkEmbed. It appears, but the browser frame wasn't created. I'll
enable debug and see what I get

Comment 16

[Chris McAfee]

TestGtkEmbed and mozilla work fine for me, I built debug, trunk.
My kernel is 2.4.20-20.9smp, 2x400MHz

Comment 17

[Mathew McBride]

Heres more info:

Motherboard: MSI 6210 
CPU's: 2 x PII 350@400
RAM: 320MB unbuffered
Distribution: Debian/unstable
Kernel: 2.6.0-test6 (custom build)

Might be a gcc problem. Haven't 'apt dist-upgraded'ed since I switched froma 
single duron to dual pII. I haven't compiled it yet, so i'll come back later

Comment 18

[Travis Chase]

*** Bug 230043 has been marked as a duplicate of this bug. ***

Comment 19

[Serge Gautherie (:sgautherie)]

Colin,
Are you still working on this ?

Does anyone still has a problem, on any platform ?

Comment 20

[Colin Blake]

Sorry, I haven't been working on the OpenVMS port for several years now.