obliterate functionality?

[Adrian Buehlmann]

On 19.03.2008 10:19, Ollivier Robert wrote:
> According to Adrian Buehlmann:
>> Would it be helpful if we could first think about removing *all* revisions
>> of a file?
> 
> If I take the other side for a moment (Devil's advocate and all that), what
> does happen if the file belongs to a much larger changeset (like whitespace
> changes on 1000s of files)?

Good question.

The same question can be asked about partially cloned repos (see PartialClone
on the wiki).

My main point was that PartialClone may lead to the same problems. If PartialClone
can be implemented, then it might be used/extended for obliteration.

Back to your question:

I would say, if you do a "hg update" to that cset, you would not get jerk.c into
your workspace. But of course you should get the other files. Maybe, "hg update"
should print a reminder that jerk.c has been obliterated and is thus not updated.

The bigger question is how to store the info that jerk.c has been obliterated.
PartialClone hints:
"add a new file like .hgignore to the .hg/ directory that specifies which files
should be ignored for clone/checkout."

I understand "ignoring" here as Mercurial should not complain about the missing
revlog for jerk.c.

This is somewhat of a perennial. See also earlier threads about closing heads.

Maybe we need something like a meta-repository (somewhat like MQ).