On Mon, 28 Oct 2002, Udi Mottelo wrote:
On Sun, 27 Oct 2002, Bart Schaefer wrote:
Actually what happens is that one character gets lopped off each time a
filter trips the bug. So depending on how many filters you have, it's
In my case it was not exactly, The number of the missing chars
was not the number of the filters in the rcfile, _BUT_, I noted
that it _DOES_ match the number of the filters that are included
in the last nested recipe in the rcfile.
That may be coincidence. Filters don't -always- trip the bug, or it'd be
easier to track down. Without checking after each filter, you don't know
which ones are the culprits.
BTW - the bug had appeared in Solaris7 and RedHat7.1. After
upgrading to procmail v3.22 the bug disappeared from the two
systems! I wander because I here that the bug still with us(?)
It's a strange bug. It's been reported to disappear after recompiling,
only to appear again after another recompile. For a while I thought it
was related to some builds using mmap(), but that's been ruled out. Now I
think it's related to file descriptors shared between the parent procmail
(the one that rescues the message when the filter exits nonzero) and the
child forked off to run the filter.
In fact, a good thing to look for might be: Does this occur only with
filters that have to run a shell to interpret pipes or metacharacters?
Or perhaps only with filters that _don't_ run a shell?
_______________________________________________
procmail mailing list
procmail(_at_)lists(_dot_)RWTH-Aachen(_dot_)DE
http://MailMan.RWTH-Aachen.DE/mailman/listinfo/procmail