Re: ask-2.5.0-1 IOerror

In message <[email protected]>, Marco Paganini writes:
> Erez,
> 
> Very strange. You're not even delivering to a pipe. Can you reproduce the
> problem? Can you send yourself a single message and send me the ask.log
> file with loglevel=10 ?
> 
> Regards,
> Paga

I cannot easily reproduce it, but I believe this had something to do with
one of the pipes ASK is using being closed unexpectedly.  Recall, I run ASK
in my .forward as follows:

	"| spamc | ask.py ... --procmail | mh-slocal"

mh-slocal is sometimes unhappy when fed a particularly misconfigured sender
address, and it'll core dump, closing it's pipe unexpectedly.  During these
core dumps I saw some of the IOerrors I reported.

I lost my workstation disks during the power outage last week.  In the days
that followed, during the recovery and restore procedure of my email, I had
to reinitialize my .ask/ and .spamassassin/ directories. So I fed into them
all my accumulated mail in the past year(!)  Sometimes things were taking
too long and I got impatient, so I hit ^C to abort the script that fed all
my email, one message at a time, into "spamc|ask|slocal".  It was during
this ^C aborts that I also got some of the IOerror messages.

So you might want to check how ask.py behaves when it gets SIGINTR or
SIGPIPE, or an error reading from stdin, or an error writing to stdout.
This isn't an easy task b/c you have to consider that these errors or
signals could arrive at any time in ask.py's run.  Maybe you might setup
some sighandlers just to catch these situations, and later on figure out how
to gracefully recover from these situations; or block certain signals during
critical times.

Thanks,
Erez.


-------------------------------------------------------
This SF.net email is sponsored by: VM Ware
With VMware you can run multiple operating systems on a single machine.
WITHOUT REBOOTING! Mix Linux / Windows / Novell virtual machines
at the same time. Free trial click here:http://www.vmware.com/wl/offer/358/0