Quoth Mark J. Nelson on Fri, May 30, 2008 at 04:52:15PM -0600:
> SCM Migration Project Tool Usage Survey
> 
> What tools have you used from the SCM Migration Project?
>
> http://opensolaris.org/os/project/scm-migration/

Cadmium, ws, webrev, and hgsetup.

> When was the last time you used them?

Today.

> What project(s) did you use them for?

Enhanced SMF Profiles.

> Are you a project or consolidation gatekeeper?

Project gatekeeper.

> What consolidation are you targeting?

O/N.

> Did you use teamware or mercurial?

Mercurial.

> Did you have to do any merges or "tricky" operations?

Yes.  Sync'ing with the gate can be obtuse.  For one, I like to pull
over the changeset which includes the new tag, which means I have to
look up that changeset's ID, not the tag.

Second is merging.  I don't remember why I stopped using gpyfm, and
I haven't investigated installing any other tools, but in lieu of that
all I want is a list of the files which still have conflict markers in
them, and a way to get sensible diff's.  One way I've found is to use
set $rev to the revision of the ancestor (which I think I use hg log -m
for) and run

        hg st --rev $rev -mard -n | HG_PARENT=$rev webrev -

> Did you use the tools inside or outside SWAN?

Inside.

> Did you find any issues?

When I create a workspace by pulling (necessary to put the workspace in
its own ZFS filesystem), ws emits some shell-test (you know, [ test ] )
errors.  Turns out it's because the .hg/hgrc isn't automatically
populated with a default path.  I don't know whether anything can be
done about that, except maybe making hg clone ZFS-aware.

Oh, and putting the workspace and the closed workspace in separate ZFS
filesystems is a pain.

And when I use hg on machines without Cadmium, it complains about not
being able to find it.  I've considered putting a copy in my home
directory, but I suspect that's not worth the pain.

> Anything else that you think the SCM team should know
> about before we integrate?

There should probably be a recommended workflow somewhere.  Including
how to track commit comments (I'm using frequent hg commit's for that),
how to merge, and how to sync with a gate.

And I've got some changes that shouldn't go back into ON.  So I suspect
I should use mq.  But I'm not sure whether it's recommended or not, or
if it's worth the complexity and learning curve.  So I guess consider an
official "when to use mq" recommendation.


David

Reply via email to