I found the mmap in CVS 1.11.23, and would not be at all surprised to
find it also in the unstable release that Alex is using. I'm certain
it's been in CVS since CVS absorbed RCS.
Once Alex has determined which part of CVS is crashing (client or
server), the first thing I would check is ulimit for the account
running the CVS binary. Also try running side-by-side something that
monitors the allocation of swap space and a tracing facility on the
CVS binary. This should give you some idea of what in CVS is
consuming the most space.
On Nov 13, 2008, at 1:01 PM, Arthur Barrett wrote:
Alex,
My server configuration is:
CVS 1.12.13 on Solaris 11.
My client configuration is:
CVS 1.12.13 on Debian 4.0
Just like with Kevin I think the first thing to do is be 100% sure
that
it is the client or the server failing. Try coping the file to the
server and doing a commit with :local: if that fails it'll be easier
to
debug too.
No per-process memory limits appear to be set. However, the
cvs process is
unable to memory-map the file:
All the memory map stuff is new in 1.12 I think - you could try going
back to 1.11...
If it is the server that is the problem then there are probably
quite a
lot of solaris parameters that could affect it - particularly if the
memory maps in 1.12 are using shared memory. I'm no Solaris expert
and
AFAICT Solaris 11 is not even a released product.
http://en.wikipedia.org/wiki/Solaris_(operating_system)#Versions
Regards,
Arthur
