Bug : uncommit and deleted file

[Aaron Bentley]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Martin Pool wrote:
> On 5/1/07, Aaron Bentley <aaron.bentley at utoronto.ca> wrote:
>> > The way I look at it, the bug is that _commit_ alters the tree state.
>> > The bug isn't in uncommit.
>>
>> I want that behavior.  People who don't want it can use --strict.
> 
> I don't take issue with you wanting that behaviour; it's a reasonable
> thing to want (which is why I wrote it that way in the first place.)
> At least it's a reasonable thing to have as an option.

> My point was just that conceptually the problem is that commit is
> throwing information away.  It is a bit strange that the only case
> where it updates the user's model of the working inventory (as opposed
> to the last-modified and similar bits which are internal) is for
> missing files.

Okay, so this is interesting, partly because I'm a bit wrong.

If Erik had been using --strict, the commit would have failed because
b.txt would be unknown.  But there are circumstances where that would
not be the case, like if *.txt was ignored.

It would be possible to have commit fail if there were dangling
ids/missing files as well.  Personally, I don't like that, but it would
be more reliable than --strict.

I certainly would welcome that more than changing uncommit in the way
that was proposed.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGN1B90F+nu1YWqI0RAqC7AJ9+InfvpC721MGBqmocpE3jwpDqzgCdH2EO
Aeb9YP7uG5iDra1PvR4tWqk=
=ioXd
-----END PGP SIGNATURE-----