Re: cvs [update aborted]: received broken pipe signal

On Wed, Mar 5, 2008 at 5:18 AM, David Wolfskill <david_at_catwhisker.org> wrote:
> On Wed, Mar 05, 2008 at 03:46:53PM +0800, David Xu wrote:
>  > What's wrong with our cvs server? I can not receive updates:
>  >
>  > cvs update: Updating .
>  > cvs update: Updating alpha
>  > ...
>
> > cvs update: Updating amd64/conf
>  > cvs update: Updating amd64/ia32
>  > cvs update: Updating amd64/includeRead from remote host
>  > ncvs.freebsd.org: Connection reset by peer
>  >
>  > cvs [update aborted]: end of file from server (consult above messages if
>  > any)
>
>  You might need to take a look at the server, but I think I recall seeing
>  similar symptoms when I created /tmp as a swap-backed md-resident file
>  system with the default inode density and /tmp didn't actually become
>  "full," but ran out of inodes during a period of heavy (for that
>  environment) usage.
>
>  Adjusting the tmpmfs_flags specification in /etc/rc.conf (for the CVS
>  server) to say 'tmpmfs_flags="-i1024 -o async -S"', then re-creating
>  /tmp  alleviated the pain.

In this case, repoman's /tmp is far from full:
Filesystem  1K-blocks Used   Avail Capacity  Mounted on
/dev/da0s1e   7858549   65 7229801     0%    /tmp

There's nothing in dmesg about cvs crashing, nor log messages anywhere
obvious that I can find that might give a clue as to what is going on.

I tested a cvs update myself, and it worked for me.  The best guess I
can come up with is that the local checkout is corrupt somehow.

-- 
Peter Wemm - peter_at_wemm.org; peter_at_FreeBSD.org; peter_at_yahoo-inc.com
"All of this is for nothing if we don't go to the stars" - JMS/B5
"If Java had true garbage collection, most programs would delete
themselves upon execution." -- Robert Sewell