[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [PROPOSAL] Move sleeping for timestamps to libsvn_wc

From: Erik Huelsmann <ehuels_at_gmail.com>
Date: 2007-03-14 13:54:41 CET

On 3/13/07, Peter Lundblad <plundblad@google.com> wrote:
> Ivan Zhakov writes:
> > On 3/13/07, C. Michael Pilato <cmpilato@collab.net> wrote:
> > > Ivan Zhakov wrote:
> > > > I like this idea. Actually I've never understood why
> > > > sleep_for_timestamps stuff is in libsvn_client.
> > >
> > > I could be wrong, but I think it lives in _client at the moment because we
> > > wanted to sleep the minimum amount for a given high-level operation. The
> > > worry was with a single person running 'svn' in a scripted situation, not
> > > running any-ol-WC-library-consumer. And rather than sleep once per changed
> > > file, for example, we much preferred to sleep once per update, for example.
> > >
> > Michael, thanks for clarification. But I think that wc operations is
> > general enough to sleep only once for operation. For example in update
> > we can sleep in close_edit(), instead in sleeping after each file.
> >
> Firstly, I agree that the WC library is better of making a decision of
> if/how long to lseep. For example, it could use the last timestamp put into
> an entries file as the bases for the sleep instead of the current time.
> I actually had an experimental patch for this some time (read years)
> ago but I abandonded it because it got messy and I decided it wasn't
> worth it.
>
> But wait, I think sleeping when closing a working copy or an edit would be a
> mistake. This would mean that update/switch would sleep after each
> svn:external and target, instead of once at the end as it is now.
> Copy would sleep after each target.
>
> Peter Samuelson writes:
> >
> > Hmmm, how about moving the sleep functionality into a new
> > svn_wc_sleep_for_timestamp called from libsvn_client? The idea is that
> > apps could also call it directly if they want.
> >
> > libsvn_wc would have to keep an internal global for "latest timestamp
> > written to an entries file", but that's trivial enough.
>
> Not for a library that can be invoked from multiple threads simultaneously.
>
> So, no, please don't hide the sleeping inside libsvn_wc!
> I think I spent enough time on this extra 0.55 seconds we sleep on average
> now;)

Right. Understood. Maybe it's not worth it.

bye,

Erik.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Mar 14 13:54:56 2007

This is an archived mail posted to the Subversion Dev mailing list.