beginner question: networkless use of hg

[Tim Bell]

Giorgos Keramidas wrote:

  [...]

> My experience so far with pulls from `other' trees seems to indicate
> that it's a good idea to use a separate `integration workspace', as I
> described above.  This way if you botch the integration, you can simply
> wipe the integration workspace and start from scratch.

I completely agree.  Use temporary space for these 'integration' or
'merge' areas (anything you do with more than one parent) - no matter
what SCM you are using.  Use a throw away area so you can burn up the
entire thing and start over quickly if something is not right.

If the merge goes OK, also after your build and test runs on the temporary
area are successful (you checked, right? ;-), then push/pull changes out to
the other more permanent repositories, either uphill or downhill, where
others will pick them up.

> The other option, of pulling from a bundle directly into a personal
> workspace is a bit risky.  You may accidentally pull hundreds or even
> thousands of changesets with a *very* complex history graph into an
> otherwise `clean' workspace and then decide that this was a mistake.

Risky - you bet - I would rather use some lunch money to buy the extra
disk needed and add scratch space for a disposable buffer.  You will be
glad you did.

> Cleaning up is still possible, but it's far trickier than a simple
> 
>     rm -fr /var/tmp/integration-workspace

Right.

Cheers-

Tim Bell
'integrator' for http://hg.openjdk.java.net/jdk7/tl