[darcs-devel] Bugs in RepoPatchV3

[Ben Franksen]

Am 24.06.19 um 22:21 schrieb Ian Lynagh:
> OK, so we actually have
>     X + A = X A'
>     X A' B + X C = X A' B C'
> Where A' and C' are conflictors. Then I think you're right that in a
> case like:
> 
>     X touches line 1.
>     A touches lines 1-5.
>     B touches line 3.
>     C touches lines 5.
> 
> that C does not record B amongst the patches that it conflicts with, and
> that B remains a non-conflictor even if you reorder to
>     X A' C'' B'

Yes, exactly. And this also shows why I needed to extend the test case
generator to see an example of this: B must be recorded after the
conflictor A' to get into this situation.

This is good. It means we can continue to regard a normal patch that
depends on a conflictor as resolution for that conflict, i.e. ignore it
when it comes to marking unresolved conflicts.

Cheers
Ben