Re: [ANNOUNCE] cvsps support for parsing BK->CVS kernel tree logs

[Andrea Arcangeli]

On Wed, Mar 19, 2003 at 07:19:32PM -0500, David Mansfield wrote:
> On Wed, 19 Mar 2003, Andrea Arcangeli wrote:
> 
> > this is a critical feature for me, note that if Larry agreed of tagging
> > the tree with the atomic-date of the patchset extracting those patchset
> > would be an order of magnitude faster, I would pay for the RTT latency 1
> > per patchset, not once per file and it would be possible to teach cvsps
> > to learn diffing against the tag if it matches. Anyways stuff works now
> > so I'll just wait for the RTT all the time w/o being able to use the
> > real bandwidth provided by my link. As soon as the CVS tarball is
> > available I'll use it for the large patch extractions.
> 
> I just looked briefly at the 'cvs server' protocol.  It looks fairly easy 
> to have a 'keepalive' session with a server where multiple diffs of 
> multiple files are requested.  This isn't a 5 minute fix though, but I've 

this sounds an excellent plan if technically doable

> been bothered by the RTT per file as well, because we use 'ssh' to 
> authenticate cvs access here, and each file is a complete ssh handshake 
> which is very slow.

with ssh the latency would be exploded indeed, it takes quite some cpu too

> > One more feature wish (besides the python inteface for a quick gui) is
> > the C^c killing cvsps too during a -g patch extraction, you probably
> > should check the return code of cvs when extracting the patches. Right
> > now I press C^z and then I killall cvsps ;)
> 
> Already fixed in my code.  I won't bother you with a diff.

I'll fetch it with the next release, thanks.

Andrea