Re: [Monotone-devel] New 'update <rev>' command?

[Derek Scherger]

Nathaniel Smith wrote:
> Heya, Joel, noticed the new version of 'update' that you added, that
> takes a revision argument.  Can you describe what the intended
> semantics of this are?

I originally logged the associated bug after a discussion with njs 
actually... at the time I beleive I had done an update and decided that 
I wanted to back up the tree for some reason that I can't recall.

I have another thought (described below) on where this would possibly be 
useful, related to the other bug (update with multiple candidates) that 
is also addressed in the same commit. I happened to have logged that bug 
as well :) so thanks to Joel for fixing them both!

The other day I was at a revision above a fork and did an update which 
actually chose one of the two available heads (by joel)  and updated to 
it. I was left wondering how that head was chosen over the other 
available one (by njs). I don't have trust hooks set up to exclude 
either head.

When I originally logged the bug I don't think I had a clear 
understanding of what update was complaining about. Now I beleive the 
problem is/was that there's no clear way to choose which head to be the 
update target. If the two heads were merged then an update could 
presumably take either path to the single merged head and be successful.

So how is the decision to pick one head from many made in this fix?

The other fix (update to specified revision) might be useful in the case 
where update picks one head and you actually want the other. That is, "I 
don't like this update, please move me over the fork to the other head". 
Whether this is necessary or not probably depends somewhat on whether 
choosing an arbitrary head as the update target was good in the first 
place.

In my case I ended up at joel's head, which appeared to have failing 
tests so I wanted to move over to njs's head where the tests were all 
passing or failing where they were expected to.

Cheers,
Derek

... with yet another long winded post...