On Mon, Apr 5, 2021 at 3:23 PM Junio C Hamano <gitster@xxxxxxxxx> wrote: > > Jerry Zhang <jerry@xxxxxxxxxx> writes: > > > I can see what you mean about the user safety issue. However, > > my specific use case (see cover letter) involves an index that does not > > match HEAD, and wouldn't be possible at all if we forced the index to > > match HEAD. Furthermore git-apply --cached even without --3way > > doesn't force the index to match HEAD either, so why force it now? > > Primarily because we tend to be extra careful before mergy operation > than any other operation. Especially without --3way, apply (with or > without --cached/--index) is extra careful to make itself all-or-none > operation to be safe, so that there is no mixed mess that requires > manual intervention (which would further increase the risk of mistakes). > > It is OK to introduce a new option to allow a dirty index, and your > tool can pass that option when it calls "apply --cached --3way", but > it would be safe to require a clean index (it does not matter how > dirty the working tree is ;-) by default. > Sure adding the staged files will definitely clobber whatever the user had in the cache at stage 0. This will probably be unexpected. But the normal invocation of --3way also does this without warning, since it touches the cache as well. It just seems odd to me to be adding a safety check on some paths that aren't there on other very similar ones. Maybe another option would be to add a very stern warning for users of --3way? Unrelatedly would you have context on why --3way falls back on 3way rather than trying 3way first then falling back on apply_fragments if blobs don't exist? I see some cases where the normal patch application will succeed but apply the patch incorrectly, while 3way will apply the patch correctly. In these cases it's impossible for the user to force 3way. Are there downsides to 3way that aren't solved by falling back on apply_fragments?
