On Wed, Jul 16, 2008 at 04:30:55PM -0700, Eric Schrock wrote: > On Wed, Jul 16, 2008 at 04:16:34PM -0700, Stephen Hahn wrote: > > * Eric Schrock <eric.schrock at sun.com> [2008-07-16 21:30]: > > > After editing a single file and switching to SSH, I now get the > > > following times: > > > > > > hg status: 14 seconds > > > hg list: 29 seconds > > > hg pdiffs: 41 seconds > > > > > > I don't see how this is viable as an everyday development tool. > > > > Could you give a bit more detail about how you're measuring these > > operations? That's more than 5x worse for hg status on a local > > filesystem than I've ever measured, assuming the working copy has been > > warmed up. > > This is just: > > $ hg clone ssh://elapso.eng//export/gate-hg' > $ vi usr/src/Makefile > $ hg status > > I also did a clone of the closed gate into usr/closed, but I don't think > that affects the main gate. If you think the machine may be at fault, > this is on zday.sfbay (and the hg workspace is > /export/eschrock/on-fixes). I updated the SFW and onbld packages just > the other day, so it should all be up to date.
Interestingly, the times seem to have gone down dramatically: hg status: 3.6 seconds hg list: 8.1 seconds hg pdiffs: 11.7 seconds And now there is no difference between the real time and user+sys time - perhaps there was some unexplained network latency before? I ran multiple invocations beforehand, so I doubt it could be any kind of filesystem caching (and the machine was idle at the time). It's still impressive that we can burn alsost 12 seconds of CPU time to find one character difference in my workspace :-( - Eric -- Eric Schrock, Fishworks http://blogs.sun.com/eschrock
