I have a repository at http://code.haskell.org/~galbolle/darcs2-bug/darcs-doc/
which tracks http://darcs.net with some (quite a lot of) conflicts.

I resolved these conflicts and recorded the resolutions, but when I try to pull
>Wed Aug 27 12:58:43 CEST 2008  Dmitry Kurochkin <dmitry.kurochkin@gmail.com>
>  * Refactor get_matcher and apply_inv_to_matcher functions from Darcs.Match
>module.
i get the following error:
darcs: mergeAfterConflicting failed in geteff

then a huge where ix is, which i attach.

Hi, I think you may have forgotten the attachment :-)

The attached repository has recorded conflict markers, but they are not the
cause of the bug: unpulling the whole «resoliving conflicts» sequence and
re-recording it the right way hit the same bug.

Here is a nice reproducible darcs 2 bug.

Steps: get http://code.haskell.org/~galbolle/darcs2-bug/darcs-doc/ and pull
http://darcs.net (2 patches) and boom!

Florent, would you be interested in trying to boil this down to a minimal test
case, along the lines of those in bugs/ ?

The trouble code is here:

src/Darcs/Patch/Real.lhs

geteff :: [Non RealPatch C(x)] -> FL Prim C(x y) -> ([Non RealPatch C(x)], FL
RealPatch C(x y))
geteff _ NilFL = ([],NilFL)
geteff ix (x:>:xs) | Just ix' <- mapM (remP (Normal x)) ix
                               = --traceDoc (greenText "I got rid of x" $$
showPatch x) $
                                 case geteff ix' xs of
                                 (ns,xs') -> (non (Normal x) : map (addP (Normal
x)) ns,
                                              Normal x :>: xs')
geteff ix xx = case mergeConflictingNons ix of
               Nothing -> errorDoc $ redText "mergeConflictingNons failed in
geteff with ix" $$
                          showNons ix $$ redText "xx" $$ showPatch xx
               Just rix -> case mergeAfterConflicting rix xx of
                           Just (a,x) -> (map (addPs (reverseFL a)) $ toNons x,
                                          a +>+ x)
                           Nothing -> errorDoc $ redText "mergeAfterConflicting
failed in geteff"$$
                                      redText "where ix" $$ showNons ix $$
                                      redText "and xx" $$ showPatch xx $$
                                      redText "and rix" $$ showPatch rix

The following patch updated the status of issue1043 to be resolved:

* resolve issue1043: avoid use of sort_coalesceFL in Patch.Real. 
Ignore-this: 79003144c8571ef746108e09192cb4ed
The trouble was that sort_coalesceFL does the "coalescing" which isn't
governed by strict rules.  This might have been fixed by changing
sort_coalesceFL to also run canonize, but the cleaner solution is to
simply acknowledge that what I really wanted was just to join together
pairs of inverses, which is much easier than the problem
sort_coalesceFL is trying to solve.