Hi,

Let's say you create a repo:

dr init
echo a > a
dr add a 
dr rec (1)
dr mv a b
rm b
dr rec (2)
echo a > a
echo b > b
dr add a b
dr rec (3)
rm b
dr rec (4)
dr mv a b
dr amend-rec (4)

Basically create two special patches.

(2) is about: rename A B, remove B - so B was not in the repo and after
the patch A won't be either.

(4) is about: remove B, rename A B - so A and B were in the repo and
after the patch only B will be, with the contents of A.

So you can see the order of remove and move matters a lot.

Now if you do a darcs changes --xml -s, then you *always* get the moves
first, so you can't decide what situation are you in.

The solution would be to leave the "move" commands where they are
physically in the patch and not move them to the top.

Thanks.

Thank you for this report. I assume that the "dr" command that you use is simply an 
alias for "darcs". In order to understand this better, could you please supply some 
information about how you intend to use this changed behavior of darcs? At the same 
time, I request that someone who is better at home in darcs internals take a look 
at this and comment: I expect that thorough interpretation of the data produced by 
darcs (perhaps without the -s option?) actually will enable a suitable decision on 
the matter that you seem to be concerned about.

Thanks everybody and best regards
Thorkil

On Mon, Jan 05, 2009 at 02:09:17PM -0000, Thorkil Naur <bugs@darcs.net> wrote:
> I assume that the "dr" command that you use is simply an alias for
> "darcs".

Yes - I forgot to mention it, sorry.

> In order to understand this better, could you please supply some
> information about how you intend to use this changed behavior of
> darcs?

Basically tailor tries to workaround the current problematic behaviour,
but it can't do all the the time, because of the problem I described.
The result of this is that if you try to convert the 4 patches mentioned
above using tailor, you get a corrupted result. And I think this is not
a tailor bug, it's a darcs one.

(For more info, search for "darcs idiosyncrasies" in the tailor source,
it does several tweak, since replaying a patch in the given order is not
possible, because of this issue.)

> At the same time, I request that someone who is better at home in
> darcs internals take a look at this and comment: I expect that
> thorough interpretation of the data produced by darcs (perhaps without
> the -s option?) actually will enable a suitable decision on the matter
> that you seem to be concerned about.

Thanks!

I'm pretty sure this is a duplicate of another issue (in which move patches are
systematically presented first).  If so, the next actions would likely be to
locate that bug (I tried doing a search, no luck) and close this bug as a duplicate.

Marking need-volunteer since we have a next action.

Actually, I think I've got it:
http://bugs.darcs.net/issue183

vmiklos: is that the same issue, in your opinion?

On Mon, Jan 05, 2009 at 02:32:44PM -0000, Eric Kow <bugs@darcs.net> wrote:
> Actually, I think I've got it:
> http://bugs.darcs.net/issue183
> 
> vmiklos: is that the same issue, in your opinion?

Yes, it is. Though I think I provided a good example on why this is a
bug - not just a nice feature, even if it was ever this way.

Thanks!

Closing as a duplicate.