Darcs creates inconsistent conflictors in a conflict fight scenario
when using darcs-2 format.

In conflict fight scenario we have two repositories A and B. After
pulling from A and finding conflict, B resolves it unilaterally. This
is repeated with each new change pulled from A.

Bug is probably triggered because B reverts to the situation before
first pull. In other words, all the new patches recorded by B are
identical.

Please find attached a test script, output from two test runs, first
with --hashed, second with --darcs-2. The exact-version is included.

Darcs pull --verbose output:

Pulling from "../RA"...
darcs: Inconsistent patch:
conflictor {{
|hunk ./file 1
|-0
|+let it b
|conflictor [
|hunk ./file 1
|-0
|+let it b
|]
||:
|hunk ./file 1
|-0
|+a0
|:
hunk ./file 1
-0
+let it b
|hunk ./file 1
|-0
|+a0
|:
hunk ./file 1
-a0
+a1
|:
hunk ./file 1
-0
+let it b
|:
hunk ./file 1
-0
+a0
}} [
hunk ./file 1
-0
+let it b
]
|hunk ./file 1
|-0
|+a0
|hunk ./file 1
|-a0
|+a1
|:
hunk ./file 1
-a1
+a2
mm conflicts with im in is_consistent where nmm is
{{
|hunk ./file 1
|-0
|+a0
|hunk ./file 1
|-a0
|+a1
|conflictor [
|hunk ./file 1
|-0
|+a0
|hunk ./file 1
|-a0
|+a1
|]
||:
|hunk ./file 1
|-0
|+let it b
|:
hunk ./file 1
-0
+let it b
}}
Command exited with non-zero status 1

Escalating (+droundy)

I notice you have not attached the test script [or it didn't make it to the
tracker].  Could you submit it as a patch to darcs, under the bugs/ directory? 
Or barring that, attach it?  Thanks!

2008/4/30, Eric Kow <bugs@darcs.net>:
>  I notice you have not attached the test script [or it didn't make it to the
>  tracker].  Could you submit it as a patch to darcs, under the bugs/ directory?
>  Or barring that, attach it?  Thanks!

Here you go.

Attachments

• demobug817.dpatch

Alright, I've submitted your patch.  I'm guessing that either Jason or David are
the best people to look at this.  Assigning to David.

I've isolated the bug here to Darcs.Patch.Real.geteff, for what it's worth. 
This is not surprising, as I'd say that this is the trickiest function to get
right in all the new formalism.  :(

The following patch updated the status of Issue817 to be resolved in the =
unstable branch:

* resolve Issue817: fix bug in conflict-handling with darcs-2 semantics.=20
This bug was due to the buggy use of a buggy function called
depends_uponFL.  I've removed this function, and am making this note
explicit so that noone else (including myself) will make the mistake of
resurrecting this function from the past.

You can view the patch details online here:=20
http://darcs.net/cgi-bin/darcs.cgi/unstable/?c=3Dannotate&p=3D20080507144=
950-72aca-25383dc9aa6240304203aac32eed1ddf563aeb9a.gz