Re: flexfax: mail to fax gateway: mailer died with signal 13

[Phil Watkinson <pkw@elgro.co.uk>]

At 19:16 06/04/1998 GMT, "Eloy A. Paris" wrote:
>
>So, to me it looks like faxmail is dumping core because of a
>segmentation fault (13 in octal is 11 in decimal and it corresponds to
>SIGSEGV).
>

Ahhh.. I see. However, you said that that the fax was sent correctly, so
if faxmail is then seriously falling over, it must be a BIG hint as to
where the problem is.

A _quick_ look at the source ( and I can't read C++ ) indicates that the
last thing faxmail does is the direct delivery, and if there is a problem
fxFatal should put an error message in the log and exit with -1.

>
>Indeed, procmail is my local delivery agent but I don't understand
>what it has to do in this case. Sendmail uses the mailer called "fax"
>to do the final delivery, not the "local" mailer (which happens to be
>procmail).
>

Yes, you're right. A bit of muddled thinking on my part, but my mindset
was still on notification messages and broken pipes.

>: For notify, you should be able to see invocation of the script in your
>: system logs  -  this extract is from my /var/log/messages :
>
>Well, my log looks like this:
>
>sendmail[1671]: OAA01671: from=root, size=52, class=0, pri =30052, nrcpts=1,
>msgid=<199804061854.OAA01671@polux.ven.ra.rockwell.com>, relay=root@localhost
>HylaFAX[1675]: Filesystem has SysV-style file creation semantics.
>FaxQueuer[112]: SUBMIT JOB 58
>sendmail[1673]: OAA01671: SYSERR(root): mailerfax died with signal 13
>sendmail[1673]: OAA01671: to=eparis@9431645.fax, ctladdr=root (0/0), 
>       delay=00:00:02, xdelay=00:00:02, mailer=fax, relay=9431645,
>       stat=Deferred
>FaxSend[1682]: MODEM BOCA RESEARCH INC V.34 External/V1.510-V34_DS
>FaxSend[1682]: SEND FAX: JOB 58 DEST 9431645 COMMID 000000 53
>

Wow. See what you mean. The second sendmail entry is, I suppose, a requeue
of the original submission. Looks like you will get another copy of the
fax every time there is a 'sendmail -q'.

>
>I agree. I think I'll have to fill faxmail with printf's and see where
>it is crashing.
>

Well, just wonder if its worth trying to extract more info out of
sendmail and bump up the debugging to, say, '-d10.1' ( I'm guessing again,
this is the level I use to check DNS lookups ).

You could actually try this on the command line, if your test fax is
still in the mailq - something like :

# sendmail -q -d10.1 > debug.log

would run the submission through sendmail again, producing lots of lovely
info. Saves having to restart sendmail and whathaveyou.

Just a thought.

Cheers,
Phil Watkinson,
Boston, UK.