please add an --explict-files config option

[LaMont Jones]

There are use models where requiring the user to explicitly list files to be committed would be a benefit.

Obviously, this does not want to be the default, nor does it want to be the default for all branches on the same machine, or even for the same user. Hence it wants to be a config option that can be set for specific branches.

thanks

[Dan Watkins (oddbloke)]

I'd imagine versioning ~ would be a use case for this.

[John A Meinel (jameinel)]

LaMont, could you try to explain the use-case a little bit more?

Is it saying you don't want the files added? Or that you just want to disable the "commit everything by default"? (So that you always have to supply a list of file names to be commited.)

This seems pretty rare to request, so I would be tempted to put it into a plugin. And it seems like a pre commit hook *might* be able to do it. (Our current pre-commit hook would probably not fit, as it happens too late in the commit process.)

Off-hand I put something together in about 25 lines of python. I turned it into a bzr branch and uploaded it to my +junk and associated it with this bug.

I tested it here, and it seems to work. Does it satisfy what you would want?

[James Westby (james-w)]

On Thu, 2008-03-13 at 20:38 +0000, John A Meinel wrote:
> LaMont, could you try to explain the use-case a little bit more?
>
> Is it saying you don't want the files added? Or that you just want to
> disable the "commit everything by default"? (So that you always have to
> supply a list of file names to be commited.)

Hi,

I was part of the IRC discussion the proceeded this, and was the
one that recommended that LaMont file a bug.

I believe LaMont was looking for a way to have bzr not commit any
file by default, and so fail if none were specified on the command
line.

I asked him to file a bug as I didn't think this would be very
intrusive, as it wouldn't change any default behaviour. If it
is doable in a plugin that is great.

Thanks,

James

[John A Meinel (jameinel)]

Well that seems to be the behavior of the plugin I wrote. I'll let LaMont tell me if it is the behavior he was thinking of.

[Aaron Bentley (abentley)]

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

Daniel Watkins wrote:
> I'd imagine versioning ~ would be a use case for this.

Why not just ignore everything?

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFH4dec0F+nu1YWqI0RAnV4AJ0dwU4eAfo85MG4RU68LY/azHWrlgCfaMY9
R51qqp30RgAQbGdptOtG1cI=
=cJYf
-----END PGP SIGNATURE-----

[John A Meinel (jameinel)]

I think it is more the case where you have a lot of things versioned, and when you go to commit, you want to separate out what gets committed at one time.

It isn't so much that you don't want them versioned, you just don't want all of your changes committed together.

[Martin Pool (mbp)]

So basically it's asking that commit with no arguments should give an
error, and maybe that commit should be non-recursive?