Re: [glob2-devel] Hg cleanup status

[Michael [Plouj] Ploujnikov]

On Tue, Nov 25, 2008 at 9:11 AM, Kai Antweiler <address@hidden> wrote:
> I tried "hg strip" last weekend. I ran into a problem and didn't have
> the time to fix it.
>
> My plan was:
> 1. using "hg export" with the last clean revision and the most recent
> revision in that branch.
> 2. using "hg strip" on the first dirty revision.
> 3. using "hg import" to get the revision back.
>
> As you might remember, "hg export" does not store the individual
> revisions unless you tell it too. I just stores the overall change
> between the two revisions.
> So the scheme above should do what we want.
>
> But I ran in to a problem with the strip command.
> It worked well when I stripped the revisions after the dirty one. But
> when I used "hg strip" on the dirty revision, it created two new
> revisions somewhere. (At least that is what mercurial told me.)
>
> I don't know if this caused my plan to fail. Maybe I used import wrong
> afterwards. I haven't used it for a long time.
>
> Somebody else could try.
> In terms of logic, it should work.

The logic seems correct, but I don't know what went wrong since I
avoided ever using export/import thanks to the MQ extension's
flexibility.

I've tried to first create MQ patches which could be later manipulated
with 'hg qimport --git -r 3881:tip'. However, that failed with "abort:
revision 3890 has unmanaged children". It looks like the repo has
multiple heads, which makes things more complicated.

> How to prevent people from repushing?
> I don't know. Just hope that they don't do it.

This could be done with a precommit hook, in theory.
(http://hgbook.red-bean.com/hgbookch10.html)

Here is more info about editing Mercurial histories:
http://www.selenic.com/mercurial/wiki/index.cgi/EditingHistory

May I suggest that someone gets IRC support for this problem from
#mercurial on irc.freenode.net.

I wish you the best.

-- 
Michael Ploujnikov
http://plouj.com/