From: Johannes Sixt Date: Wed, 06 Jan 2010 08:55:21 GMT Subject: Re: [PATCH 9/9] rerere forget path: forget recorded resolution Message-ID: <4B444FF9.1000004@kdbg.org> In-Reply-To: <7v4on0oxcs.fsf@alter.siamese.dyndns.org> Junio C Hamano schrieb: > Johannes Sixt writes: >> When you simply call ll_merge(), will it obey any merge drivers that >> are defined in .gitattributes? Do we care about them? >> >> I already had an implementation of "rerere forget" before you >> presented this solution, but it relies on that the user calls >> "checkout --conflict=merge" first. One reason (besides its simplicity) >> was that it does not have to care how the merge is computed. > > Doesn't "checkout --conflict=merge" use the same ll_merge() machinery? It does, without setting up .gitattributes, either (IIUC), which is a bug IMO. That said, I consider your solution superior because it works without depending on conflict markers in the file in the worktree. Nevertheless, we should think about whether merge drivers of .gitattributes should be obeyed. I think they should: For example, a specialized XML merge driver could leave conflicts that are different from those that merge-recursive generates. -- Hannes