It's very interesting. I've used gdb -core with one of the
core files on the server. It happens in rcs.c:935, there are
sanity checks that the pointer is valid. I can't find the
cvs-1.9.27 sources on hub.org, so I grabbed my private copy
local.
It happend while cvs parses the head of psql.c,v (the rcsbuf
contains just the first portion of that file). The file looks
O.K. for me, and an rlog is happy with it, so I don't think
the repository file is corrupt.
But it's really funny. The char *ptr is set from rcsbuf->ptr
2 lines above. The values in rcsbuf look O.K. (gdb print
*rcsbuf), but ptr definitely is wrong. I cannot imagine how
this happens.
Marc, could you take a look at it and eventually upgrade to
cvs-1.10 on hub.org?
BTW: I removed most of the read lock files in pgsql/bin/psql
(after being sure they are dead ones). But the problem
remains.
Jan
--
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#======================================== jwieck@debis.com (Jan Wieck) #