Richard Lowe wrote: <snip>
> I'd imagine, given the above, that you're using backup/restore to move > changes between machines? If so, why would you do that rather than Exactly. My workstation sits under my desk in Europe and has only 2 CPUs so nightly takes forever to complete (besides, the machine is needed for other tasks). It uses local clone mirror (served over NFS over LAN) to speed up initial 'hg clone' [*]. After I am done with some set of changes (verified by partial build) the changes are transferred to a build machine in the US which sports with 16 CPUs and finishes pre-RTI nightly build under 1 hour. The repository on the build machine also uses local clone (the build machine is on different coast than the gate machine). After I am done with the building/testing/RTI, one of the repositories is reparented and the changes are pushed to the gate. > pushing or pulling to the other machine? Because it fits the Teamware model I've been using and which was so convenient and fast [**] : 'wx backup -t' on my workstation, followed by 'wx restore -f && nohup nightly incremental.sh &' on the build machine did the trick just fine. Moreover, multiple invocations of the same command were idempotent. What would be the equivalent of this operation in Mercurial in the above scenario ? (Anything which requires to remember location(s) does not really count.) v. [*] and that is further sped up with ZFS snapshots done on the workstation [**] did I mention that having to specify backup directory as argument to 'hg restore' is cumbersome ?
