No-one's defended the behaviour in the forwarded message, so I claim
that it is a bug.

Thanks
Ian

Attachments

• unnamed

Here's the attachment, for the convenience of people using the bug tracker:

Hi all,

gen_summary at the bottom of PatchViewing.lhs seems to very deliberately
show all of the moves before everything else. This seems to have been
the case ever since it first showed moves:
darcs annotate --match='hash 20040622092818-53a90-c683ef72f83e38ad8c6bc4b5909233
e44b2d97b6'
Is there a reason for this?

It's causing problems for some people on IRC as they are doing things
with the output that don't like files being moved into directories before
the directory is created.


Thanks
Ian

I confirmed the behavior still exists with Darcs2, but am re-classifying it as a
"wishlist" item, since I don't believe darcs ever claimed that the "summary"
output would appear in a re-playable order, although I agree that could make it
more intuitive or useful in some cases.

Re-opening (to be overruled by Thorkil as needed, as I haven't checked the
details yet).

See http://bugs.darcs.net/issue1281 for evidence on why this could be considered
a bug and not just a feature request.

Thanks. :)

Note the closely related issue1519.

Whereas that bug deals with the sorting within move patches, this deals with the
sorting of the move patches themselves (ie. before all the hunks, etc).

OK, I think to fix this we need to do some more research and answer some questions:
 - why do we sort moves first? is it just to help coalescing somehow?
 - what would we lose by not doing this?

I imagine it's safe not to do, so maybe one approach would be to try NOT sorting
and see how that works out in practice.  In any case, this needs somebody to
pick up the bug and play with things...

I think my patch will resolve this bug

Mon Aug 17 23:58:14 BST 2009  Eric Kow <kowey@darcs.net>
  * Resolve issue183: Do not sort changes --summary output.
  This adds move patches to our high-level representation of summary output.

But we need somebody (probably me) to write a regression test showing an example
of non-replayable output, eg with an adding a directory and moving a file into it.

Test written and pushed (this was resolved a long time ago, just never
got around to writing that test)