[merge] Fix commit --strict regression

[Martin Pool]

On 23/11/2006, at 3:12 PM, Robert Collins wrote:

> On Wed, 2006-11-22 at 11:35 -0500, Aaron Bentley wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Commit --strict is still broken.  John's fix of changing the default
>> ignores isn't good enough because it doesn't fix all the bzr installs
>> that already have user ignores.
>>
>> Since this is a regression, and we're halfway to a release candidate,
>> I'm submitting a patch to revert the changes.
>>
>> I don't think it's the right approach anyway.  A better approach  
>> is not
>> to prompt the user until all checks are complete.  One way is to  
>> supply
>> a callback to commit.  Another approach would be to split commit into
>> "prepare" and "finalize" steps.
>>
>> I prefer the callback approach, because it doesn't require many  
>> changes
>> to existing code, like returning a state object.  But with either
>> approach, we can delay prompting for the message until we're actually
>> ready to write the revision into the knit, so even network failures
>> would prevent us from prompting for commit messages that we can't  
>> use.
>
> +1 to the merge and the described better approach.

+1 too

-- 
Martin