Re: [glob2-devel] Hg cleanup status

[Michael [Plouj] Ploujnikov]

On Tue, Nov 25, 2008 at 4:22 AM, Leo Wandersleb <address@hidden> wrote:
> Stéphane Magnenat wrote:
>> What is the status of the hg cleanup of the wrongly commited files? I will be
>> in traveling next week so I would like to have useable glob2 tree ;-)
>>
> http://www.selenic.com/mercurial/wiki/index.cgi/Strip
> would, if aplied on 697e24c18cc6 remove this branch.
> unfortunately it would remove all the wanted changes, too.
> didn't test what happens if a new branch was created to contain
> the wanted results before stripping.
> testing is not dangerous as a backup is written automatically that
> can be restored easily.

Since there are "valid" commits after 697e24c18cc6 the 'strip' command
is not enough. 'hg strip 697e24c18cc6' would simply delete all commits
after and including 697e24c18cc6. You would then have to re-commit the
lost wanted commits. A better alternative would be to convert commits
up to and including 697e24c18cc6 into MQ patches
(http://www.selenic.com/mercurial/wiki/index.cgi/MqExtension). Then
'hg fold' the 9ff115d7065a commit into 697e24c18cc6.
>
> The "strip" has to be done directly on the server repo as the result
> obviously can't be a committable changeset

That's right. All of this (strip, fold, etc) has to be done on the server.

> Also all committers should avoid recommitting those changes but
> i guess that part is manageable with only a hand full of devs.

Having to do this kind of people management is one reason why editing
a public repository is a bad idea.

Read the following articles for more reasons:
https://lists.ubuntu.com/archives/bazaar/2008q2/041647.html
http://lwn.net/Articles/269210/
http://hgbook.red-bean.com/hgbookch9.html#x13-1700009.1  - section 9.1.4
http://hgbook.red-bean.com/hgbookch9.html#x13-1860009.4

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