Re: cvs lock stuck?

[Joern Rennecke <amylaar at redhat dot com>]

> 
> On Thu, Aug 30, 2001 at 12:21:35PM +0200, Michael Matz wrote:
> > On Thu, 30 Aug 2001, Joern Rennecke wrote:
> > 
> > > cvs server: [02:26:43] waiting for anoncvs's lock in /cvs/gcc/gcc/gcc
> > > cvs server: [02:27:13] waiting for anoncvs's lock in /cvs/gcc/gcc/gcc
> > > cvs server: [02:27:43] waiting for anoncvs's lock in /cvs/gcc/gcc/gcc
> > > cvs server: [02:28:13] waiting for anoncvs's lock in /cvs/gcc/gcc/gcc
> > > ...
> > > cvs server: [03:08:44] waiting for anoncvs's lock in /cvs/gcc/gcc/gcc
> > > cvs server: [03:09:14] waiting for anoncvs's lock in /cvs/gcc/gcc/gcc
> > > cvs server: [03:09:44] waiting for anoncvs's lock in /cvs/gcc/gcc/gcc
> > 
> > Something definetly isn't good on the server.
> 
> AFAIK it is typical cvs behaviour. Try e.g. doing cvs diff -rX -rY | less
> with something that overflows the pipe buffer and keep it that way for a
> while. The lock will be held all the time...

Of course, this is easy to avoid by skipping to the end straight ahead
with G and then going back to the start.  But we can't expect any kind
of instructions to followed consistently with an anoncvs server.  So
maybe we should do something in the server instead:
- set the output buffer size to a large value.
- set a timeout for all jobs that are expected to produce a lot of output
  (we don't want a timeout for checkins because there could be a lot of disk
  I/O).