Re: [Gnu-arch-users] Re: <<< conflict markers

[Jan Hudec]

On Wed, Apr 14, 2004 at 17:10:34 -0400, Stefan Monnier wrote:
> >> > diff3-style conflict markers really only make sense when the program
> >> > creating them has three files on hand: ANCESTOR, MINE, and YOURS.
> >> > `replay', `redo', etc. all work (currently) on the basis of diffs, not
> >> > 3-way merges.
> >> One exception: `update' could very well use diff3.
> > It currently does not keep a copy handy, so it's not _that_ simple.
> 
> Well, I don't know much about internals, so I'm only speaking from
> the point of view of the functionality.
> 
> >> But it would have the same effect as `tla star-merge --three-way', so it
> >> would be a bit redundant.
> > But star-merge is far less efficient for this.
> 
> Sounds like a good reason to improve star-merge's performance ;-).

star-merge is principialy less efficient. Star-merge must build the
ancestor and mine versions always, because of it's generality.
replay/update only need them if a conflict occurs (update always builds
ancestor, replay does not build anything) and thus can be faster in the
common case.

> >> I wish there was a tool that would do the guessing and turn the .rej
> >> into conflict markers.  Can't be done reliably in general, but it
> > There is: wiggle
> 
> Ha!  Way cool!  Thanks!
> 
> 
>         Stefan
> 
-------------------------------------------------------------------------------
                                                 Jan 'Bulb' Hudec 
<address@hidden>

signature.asc
Description: Digital signature