[Asterisk-Dev] SEGV in CVS version of Asterisk
chrisalbertson90278 at yahoo.com
Wed Dec 10 09:50:36 MST 2003
I've been writting software with UNIX now for almost 20 years.
ANY time a program seg faults. (Plenty of mine have.) it is
100% absolutly for sure the fault of the programmer. He simply
missed something. Using a pointer without first checking is
is just not the "best practice" No excuses. That said. Just
try writting a 50,000 line program without even a mis-placed
comma. Programs are written by humans
One thing about Asterisk is that there IS NO RELEASE SCHEDULE.
With other software there is a freature freeze followed
by a beta test period and then some "release cadidates" and
finally a "stable release" put out some place as a .tgz file.
Asterisk ONLY has a leading edge "current CVS" that has under
gone very little testing or QC. This is not bad. IT simply means
WE have to do the final testing and verioning controls.
If the current CVS segfault, fix it and send in a patch
One other open source software I depend on is "PostgreSQL"
those people test the heck out of each release and write a
detailed document listing all the changed from the last
release along with instructions on how to upgrade. Major
releases come out infrequently. Of course they also have
a CVS but claim that the current CVS may not even compile.
It would be neat if someone could do this with Asterisk
but, you know we all have other things to do. So instead of
one person doing it we all do it ourselves
--- Jonathan Tew <jonathan at ultracart.com> wrote:
> I would have to disagree with the advice that you guys are giving
> person. If the module compiles in to Asterisk by default then it
> be coded to handle the null pointer gracefully. Other non-developers
> could experience the same crashing of Asterisk and have no idea what
> going on or how to diagnose it. All they end up doing is either stop
> using the product because it seems buggy or harass people over and
> again on the users mailing list.
> Matteo Brancaleoni wrote:
> >or better, just rm -fR all modules you don't need
> >in /usr/lib/asterisk/modules
> >so no need to bother 'bout modules.conf
> >but be aware that there're some deps between them
> >(ie. can't have chan_modem_blah without chan_modem.so)
> >Il mer, 2003-12-10 alle 05:48, Steven Critchfield ha scritto:
> >>in /etc/asterisk/modules.conf
> >>On Tue, 2003-12-09 at 21:53, Ron OHara wrote:
> >>>Maybe someone can suggest how to bypass this problem.
> >>>I took a copy of asterisk from CVS and did the 'make clean ; make
> >>>install ; make samples' and tried starting it (with all the
> >>>unchanged) just to see if it would start.
> >>>Nope - instant Segmentation Fault.
> >>>So - after a bit of tracking around with 'ddd' I found that the
> >>>is in the 'monitor' thread of the 'chan_vpb' driver.
> >>>The SEGV occurs on line 142 of 'vpb-driver-2.2.34/src/fifo.cpp' -
> >>>surprising since it is handed a NULL pointer as it's queue to read
> >>>This was called from line 2185 of
> >>>and is where the 'vpb_get_event_sync' routine tries to do an
> >>>read from a queue pointer (variable name is APIQ) ... and this
> >>>is also NULL.
> >>>Since I dont have a Voicetronix Voice Processing Board (VPB) in
> >>>system, it's not surprising that all the setup stuff is null or
> >>>My question is: How do I tell the configs that I dont have a VPB
> >>>installed? - as a secondary issue, somewhere in the startup, the
> >>>initialisation of the VPB card should have failed and not let this
> >>>situation move forward to where it SEGV's .... if there is no
> >>>file change I can make, can someone point me to (roughly) where
> >>>initialisation happens so that I can make it tolerate the lack of
> >>>Ron O'Hara
> >>>Asterisk-Dev mailing list
> >>>Asterisk-Dev at lists.digium.com
> Asterisk-Dev mailing list
> Asterisk-Dev at lists.digium.com
Home: 310-376-1029 chrisalbertson90278 at yahoo.com
Office: 310-336-5189 Christopher.J.Albertson at aero.org
Do you Yahoo!?
New Yahoo! Photos - easier uploading and sharing.
More information about the asterisk-dev