[comp.mail.sendmail] sendmail 5.61 on ISC 3.2 unix i386

james@raid.dell.com (James Van Artsdalen) (01/07/90)

I am trying to compile sendmail 5.61, straight from ucbarpa.berkeley.edu,
for ISC SysV 3.2 i386.  I have it pretty much working with TCP/IP and
daemon and so forth, except that I am seeing some core dumps.  It appears
that sendmail never dumps core for the uux mailer, only SMTP.  Over
time, the mqueue directory builds up mail that was locked when sendmail
dumped core during delivery.  When sendmail dumps core, it is usually
in malloc(3c) or _filbuf, with a stray pointer.

Has anyone successfully ported sendmail to ISC unix with SMTP working?
--
James R. Van Artsdalen          james@raid.dell.com       "Live Free or Die"
Dell Computer Corporation  9505 Arboretum Blvd Austin TX 78759  512-338-8789

peter@orfeo.radig.de (Peter Radig) (01/09/90)

james@raid.dell.com (James Van Artsdalen) writes:

>that sendmail never dumps core for the uux mailer, only SMTP.  Over
>time, the mqueue directory builds up mail that was locked when sendmail
>dumped core during delivery.  When sendmail dumps core, it is usually
>in malloc(3c) or _filbuf, with a stray pointer.

As far as I know some problems may arise if the `bcopy' does not copy
overlapping structures quite right. I had always problems when replacing
the calls to `bcopy' with `#define bcopy(a,b,c) memcpy(b,a,c)'.

Peter
-- 
Peter Radig
Voice: +49 69 746972
USENET: peter@radig.de  or:  uunet!unido!radig!peter

netnews@uafcveg.uark.edu (Netnews account) (01/10/90)

In article <886@uudell.dell.com>, james@raid.dell.com (James Van Artsdalen) writes:
> I am trying to compile sendmail 5.61, straight from ucbarpa.berkeley.edu,
> for ISC SysV 3.2 i386.  I have it pretty much working with TCP/IP and
> daemon and so forth, except that I am seeing some core dumps.  It appears
> that sendmail never dumps core for the uux mailer, only SMTP.  Over
> time, the mqueue directory builds up mail that was locked when sendmail
> dumped core during delivery.  When sendmail dumps core, it is usually
> in malloc(3c) or _filbuf, with a stray pointer.
> --
> James R. Van Artsdalen          james@raid.dell.com       "Live Free or Die"
> Dell Computer Corporation  9505 Arboretum Blvd Austin TX 78759  512-338-8789

You might want to check the 'fdopen' routine on your system to make sure that
they provide a buffer.  Some systems don't create a buffer when the 'fdopen'
routine is called.  This created Mega-problems for me when I tried to port it
to HCX-UX 3.0.  I finally just called 'setbuf' on a 'malloc'ed buffer after
EVERY 'fdopen' routine call, just to make sure.

- David Summers

james@raid.dell.com (James Van Artsdalen) (01/11/90)

In <3633@uafcveg.uark.edu>, netnews@uafcveg.uark.edu (Netnews account) wrote:

> In article <886@uudell.dell.com>, james@raid.dell.com (me) writes:
| I am trying to compile sendmail 5.61, straight from ucbarpa.berkeley.edu,
| for ISC SysV 3.2 i386.  [...] I am seeing some core dumps.

> You might want to check the 'fdopen' routine on your system to make
> sure that they provide a buffer.

Indeed, this was part of the problem.  Several people sent mail to the
effect that "fdopen(3c) doesn't allocate a buffer like fopen(3c)".
This is only partly true: SysVr3.2 doesn't allocate a buffer in
fopen(3c), but apparently ISC's streams/stdio interaction can require
it.

Another apparent problem is that sendmail fdopen(3c)'s a file
descriptor more than once.  I'm a little amazed this hasn't been
caught before, but I guess BSD stdio doesn't notice the problem as
long as the file descriptor isn't a regular file.  Simply dup(2)'ing
the descriptor to fdopen(3c) it helped a lot.

Thanks to Doug McCallum for providing most of the clues.
--
James R. Van Artsdalen          james@raid.dell.com       "Live Free or Die"
Dell Computer Corporation  9505 Arboretum Blvd Austin TX 78759  512-338-8789