Re: Could not map memory to RCS archive

[Laurent Duperval]

Garyl Erickson wrote:
> There seems to be some confusion here between process memory size
> limitations and disk file size limitations. 32-bit Intel processors can
> address 4 GB of memory (RAM), and, if I remember correctly, the O/S
> keeps half that for itself, hence the 2 GB limit for user processes. But
> the file systems running on those same processors support much bigger
> files than 2 GB (for example, see ext2,3,4 and ReiserFS at
> http://en.wikipedia.org/wiki/Comparison_of_file_systems), so unless your
> entire file has to fit in real memory at one time, for some reason, you
> shouldn't have a problem using or creating it. Normally, the O/S streams
> the file in much smaller chunks when reading or writing it.
> 

Hi,

Well, I am not sure at what point it starts happening. After scouring
the Net I found a few threads about this. I ended up having to use a new
filename in order to continue working. It is definitely not kosher but
at this point I couldn't wait for a solution.

I am not sure why the CVS process on the server would refuse to feed the
file if it got close to or over 1GB. Maybe it tries to allocate memory
to store the entire file in memory, and then make a copy? I don't know.

What I saw is that after increasing the memory limit for all processes,
I was able to update the file on the client, but I could not check in a
new revision because the problem reappeared.

L