Jeff King <peff@xxxxxxxx> writes: > OK, but what do you intend to do for a plumbing command _without_ > --textconv? I think what it is doing now (pretending that lines in the > binary file are relevant, and either truncating output on NUL or spewing > NULs to the output stream) is just wrong. Oh, no question about it. "Binary files differ" codepath needs to be added, and independent of if we want to add a fallback textconv. > Ick, why? That pseudo-diff contains no additional interesting > information that is not already there (since the "index" line already > contains the blob sha1s). True enough. The only case that might make a difference would be if one side was binary and the other side and the result was text, in which case the user can not just see but read the result, but I don't think it is worth caring about. Also unlike my weatherbaloon patch, Michael's approach (if it is updated to pass the whole diff options structure instead of just one "do we care about the textconv" bit to intersect_paths() function) will let us determine if the combined path should say "Binary files differ" a lot early, so there is no need to worry about what to do on binary files in the places we would be adding textconv anymore. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html