[PATCH 1 of 1] rebase: hard-code 'hg pull --rebase' to use internal:fail merge tool

[Adrian Buehlmann]

On 2011-05-26 17:45, Sune Foldager wrote:
> On 2011-05-26 15:45, Adrian Buehlmann wrote:
>> On 2011-05-26 15:22, Sune Foldager wrote:
>>> On 2011-05-26 15:18, Adrian Buehlmann wrote:
>>>> On 2011-05-26 14:55, Sune Foldager wrote:
>>>>> Right.. but that's kinda a third workflow on top of rebase and merge isn't it?
>>>>> I mean, at Edlund, no one knows about resolve. People either use rebase or
>>>>> merge and resolve all conflicts in kdiff3. Ok, SOME people know, but many
>>>>> don't. Not sure this makes it easier.
>>>>
>>>> Didn't you claim that you guys don't use rebase --pull? So I think you
>>>> shouldn't be affected by this patch.
>>>
>>> No, we don't, but I can still have an opinion on the workflow :). You claimed
>>> you didn't use it either :p.
>>
>> Oh. I'm not say you shouldn't have an opinion on this :-). But you
>> sounded like you were seeing a potential problem for Edlund.
> 
> Ah, no I didn't mean that of course. I dare not change our workflow at work
> on a whim, hehe.
> 
>> My point is, only people really using pull --rebase are affected by
>> this. So, if a user actually uses pull --rebase, he already does have
>> some special expectations about his desired workflow.
> 
> This is true, but this does introduce a new step, hg resolve.

Yes. But only if there are conflicting changes. Which gives a chance for
the user to turn it into a merge.

> But this is not
> a big deal, really. I am more concerned with my other question:
> 
> What happens when it aborts halfway through a multi-changeset rebase?

Good question. We have to investigate that for sure and I agree I
haven't answered that (yet).

But at this point, let me reply with (an incomplete) counter question:

Is the state that rebase leaves if it stops rebasing due to conflicts an
"invalid" state?

If it is (an invalid state), why does it let me commit, after I have
resolved the conflicts? :p