[PATCH 0 of 4 phases] secret changeset creation and exchange

[Pierre-Yves David]

On 22 déc. 2011, at 00:19, Matt Mackall wrote:

> On Tue, 2011-12-20 at 01:52 +0100, Pierre-Yves David wrote:
>> Here is a resent of patch 4 and 6 with the ui option implementation.
>> 
>> First patch is a small doc update independent from 3 others patches.
>> 
>> I was not very happy with adding the secret argument to commitctx and I'm glad
>> to replace it with a ui option. This options is named "phases.new-commit".
>> While writing this down I'm thinking "phase.min-commit-phase" might be better.
>> What do you think ?
>> 
>> However I fell like a secret switch on commit are still pretty handy (based on
>> my experimentation of replacing mq with obsolete). If there is a clear
>> opposition to such switch, patch 3 may be ignored if "--secret" switch in tests
>> are replaced by proper --config call. Test update from patch 3 should be
>> applied in all case.
> 
> I think I'd rather introduce a command to manipulate phases manually
> (which we'll already need) rather than introduce command line options to
> a bunch of commands.

I fell like the --switch *on commit* will be handy (and less common usecase can use the manual phase movement). However such decision might be delayed to more generic UI talk. Some of my Logilab fellow might have started playing with phase from then and provide feedback.

> 
> If I do a commit with no options on top of a secret changeset, what
> happens? Do we inherit the secret state?

Yes, that was my idea. (and a free consequences from storing roots)

-- 
Pierre-Yves David