Sebastien Roy wrote:
> On Thu, 2008-10-30 at 09:11 -0700, Erik Nordmark wrote:
>> At the ipobs putback you'll have to collapse all its deltas into a
>> single one. Have you worked out what impact this will have on other CWDs
>> that are children of it?
>
> The repo that Phil will use for integration is:
>
> /net/zhadum.east/export/ws/pk34663/ipobs-putback
>
> The changesets are collapsed there. This is the first time that I will
> have gone through an exercise like this with hg, but I can run an
> experiment by taking a clone of clearview-noipmp, then re-parenting it
> to this ipobs-putback repo. From there I'll do an hg pull and hg merge
> and see what kind of chaos ensues.
What I found (again with parent and child having identical file content)
was that after I had done a reci in the parent I could do
pull -u
merge
commit
in the child and now the child would have a bunch of "deltas" (whatever
they are called in hg) but no active files.
Doing a reci in the child would get rid of those deltas so that hg
outgoing would be empty.
In the case of the identical file content the merge is trivial since it
doesn't even try to modify the file content. I can imaging that with
different file content in parent and child the merge might mess up the
content. Checking the content against a snapshot of the child prior to
the pull, and just copying in the correct content should the merge mess
it up is something that should work.
This shouldn't be that hard if you have two versions of the parent, one
with the full "deltas" and one with collapsed where both versions have
identical file content.
> Me neither. This particular series of events was very disruptive with
> teamware (perpetual merges for files that are not different in content
> from the parent), and I don't know if hg makes it any better.
I'm hopeful. The fact that I could just do the above (again, with
identical content) with hg is a lot easier than trying to wx reset a
bunch of files to get rid of the extra SCCS history from the child in
teamware.
Erik
