Junio C Hamano <gitster@xxxxxxxxx> writes: > The "previous" in the context of running "git apply" (which does not > look at the index) is _not_ what is in the index, on the other hand. > Instead of "struct index_state *", convert_to_git() needs to be > fixed to take something else that can be queried for the "previous" > version to use as a hint, if "safe CRLF" wants to work correctly. I left it unsaid by mistake, but I think the right thing to use as the "previous" to take hint from in the context of "git apply" is what is in the working tree, i.e. the result of applying patch in step (4) to create a file F in the sample scenario. While applying patch in step (5), convert_to_git() should "imagine" adding the file F currently in the working tree (i.e. the result of step (4)) to the index---if the resulting object in the index would have CR, then the safe CRLF logic should refrain from doing CRLF->LF conversion. And it should do so without actually adding neither the preimage or the postimage to the index, of course. When we are doing "git apply --index", then we _require_ that the indexed contents and what is in the working tree matches before applying the patch, so it is perfectly fine to let convert_to_git() to look at the current index---that is the "previous" one we want to take hint from while using the safe CRLF logic.
