The RCS file is essentially of the same size as is the working copy, since
there is only one revision. One particular file is 360 MB in size, which
is approximately the resident memory size of the server CVS process
immediately before it fails.
The behavior of the server CVS process makes sense to me. I am more
interested in why it leads to a memory allocation failure and whether
there are ways to configure CVS to require a smaller memory footprint or
the OS to allow a larger memory footprint.
On a side note, I actually sent my question before the amazingly similar
one by Kevin Bulgrien--not entirely certain why it took so long to
deliver. Also, sorry about the duplicate message; I sent it in hope that
the first message simply got dropped somewhere along the way.
Alex
On Thu, 13 Nov 2008, Paul Sander wrote:
We just had a similar question yesterday. When CVS reads RCS files, it mmaps
them into virtual memory. When CVS fails, how large is the RCS file it's
processing at the time, in comparison to the metrics you indicate below?
