[PATCH 4 of 9 phases] mq: qimporting revision set them to secret

[Pierre-Yves David]

On Thu, Jan 19, 2012 at 11:53:36AM -0600, Kevin Bullock wrote:
> On Jan 19, 2012, at 11:37 AM, Pierre-Yves David wrote:
> 
> > Moreover, I'm really not convinced mq changeset should be anything else
> > that secret:
> > 
> > (1) It's much more simple: "mq changeset are secret" vs "mq changeset are
> > secret most of the time but…"[1]
> > 
> > (2) we could ditch all the dedicated mq logic and fallback to "standard"
> > phase related behavior.
> > 
> > (2) There is currently no way to exchange mq changeset with other people
> > without blowing you mq state, so preventing it as much as possible,
> > expecially on pull, if something we want.
> 
> 
> There is also currently no way to exchange MQ changesets with other people
> without either deliberately publishing (in the pre-phases sense—putting it
> somewhere that someone can pull from it) an MQ repo, or a --force switch on
> push. I think minimally adding phaseyness to MQ will be an _improvement_ to
> the current safety level without disrupting anyone's working patterns.

with the current 2.1 state, pushing mq changeset won't work without a prior
"hg phase --draft" call, event with -f.

This example is given in a simple situation where nothing else moved mq phases

> Until we have garbage collection in place, those of us comfortable with
> history editing can use strip to get rid of draft changesets that get
> abandoned. Those who aren't, we can keep recommending that they only share
> draft changesets in throwaway clones.

With secret only mq, we would haven't have to recommend anything to ensure mq
changeset won't mess with those who are not comfortable. They just won't.
People who are comfortable with sharing mutable history could use dedicated
option to share their mq changest on push.

> The recommended workflows don't change, but they do get incrementally safer,
> and the basis is laid for 2.2 to make phases really whizbang and cool for
> making a whole heap of things safer.

I don't think anything more will happen in 2.1 in the freeze. My point is that
I considerer a much better approach to have secret only mq changeset. Not
having this in 2.1 is okay, but aiming to this for 2.2 is the way to go in my
opinion. Matt is willing to backout the above changeset because he does not
want to goes in this direction. That's what I'm arguing about.



-- 
Pierre-Yves David

http://www.logilab.fr/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://selenic.com/pipermail/mercurial-devel/attachments/20120119/08d87a0d/attachment.pgp>