Re: upgrading repository from 1.10.8+ to 1.11, solaris to linux RedHa

[Larry Jones]

Gilles-Eric Descamps writes:
>  
> cvs is running through inetd in pserver mode.
> files of the repository are accessed through NFS.

Why are you using NFS to access the repository if you're running
pserver?  It would be a lot more efficient to run pserver on the machine
where the repository is.

> Sometimes, I get the error message:
> edit: too many open files (when we try to do a cvs edit on a lot of files:
> 22500)
> (cvs commit is able to commit those 22500 files.)
> Do you know how to correct that problem ?

There were a number of memory and file descriptor leaks that were fixed
in CVS 1.11; hopefully this is one of them.

> I was thinking of upgrading my repository to cvs 1.11.
> I was not unable to see through the mailing list, or on cvshome.org
> if the edit -c patch was part of cvs 1.11.
> Is the reserved edit (edit -c) part of 1.11 ?

No.

> What are the benefits of 1.11 over 1.10.8 for a :pserver-only ?
> Should I upgrade my solaris 2.7 to cvs 1.11 ?
> When reading NEWS, I did not find any new feature worth the move...

There are lots of bug fixes.  I definitely recommend upgrading.

> I was also thinking of moving the :pserver: to a Linux RedHat 7.0 machine.
> Would everything be smooth ?

Probably not.  There have been a lot of reports of NFS interoperability
problems with Linux machines (note that this does *not* imply that there
are bugs in the Linux NFS code -- the problems may well be bugs in the
other machine's NFS code that Linux just happens to trigger) which
typically result in a corrupted repository.  Also, Linux has an annoying
habit of setting $HOME for inetd/xinetd which causes havoc when pserver
inherits it.

On the other hand, if you dump NFS and deal with $HOME, it will work
just fine.

-Larry Jones

All girls should be shipped to Pluto--that's what I say. -- Calvin