As shown in this patch.  This *should* be fixed by Petr's 3-way
partitionFL patch which I'm currently reviewing.

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9

Attachments

• failing-cvs-duplicate-patchinfo.sh

Oops, wrong test (no actual dependency).
I'll push the right one in later.

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9

The following patch updated issue issue1879 with status=resolved;resolvedin=2.6.0 HEAD

* Resolve issue1879: notice unexpected commute failure on merge. 
Ignore-this: ee25219b305f2d01d21404c55fa24342
Fixed by Petr's 3-way partitionFL.

The following patch updated issue issue1879 with status=resolved;resolvedin=2.5.0 CURRENT

* Resolve issue1879: notice unexpected commute failure on merge. 
Ignore-this: ee25219b305f2d01d21404c55fa24342
Fixed by Petr's 3-way partitionFL.

I am marking this as invalid/wont-fix.

First, what it complains about is missing lipstick on a pig. The solution it 
suggests protects against spoofed patches only if they happen to depend on 
patches not in our repo. This can easily be avoided simply by making the spoofed 
patch in a current clone. We should be able to /reliably/ detect spoofed patches 
or not at all.

Second, and more importantly, the "fix" actually breaks Darcs in an absolutely 
horrible way. See issue2605 for details.

I take back my last remark. It is very sensible to fail in this case 
even if it does not reliably detect all patch manipulation. We 
currently call error which is not ideal because the user then gets 
an error message about this being a bug in darcs. Instead the 
message should hint at the possibility that someone manually crafted 
a patch in *one* of the repos, we can't tell which. My other 
argument is even less convincing considering that "merge by value" 
can be easily shown to be inconsistent with permutivity.