This fixes a common issue when for e.g. generating patches with $ diff -u foo.orig foo where 'got patch' failed because 'foo.orig' has an 'unexpected status'. prodded by naddy, ok stsp

check_file_status used got_worktree_status to check if the file was in an allowed state, but it's wrong since the callback is not invoked on unchanged files. While here also fix a relate bug: unlink(newpath) is in the wrong spot and ends up removing files even when it shouldn't, so move it early in the got_worktree_schedule_* error handling. Finally, update the appropriate test case. It was passing before because got_worktree_schedule_add returned GOT_ERR_FILE_STATUS, not because check_file_status failed. ok stsp@

ok stsp

ok thomas_adam millert

Augment got_patch_progress_cb by providing the hunks that were applied with offset (or that failed) and the recoverable error encountered during the operation (bad status, missing file, ...) got_patch now proceeds when a file fails to be patched and exits with GOT_ERR_PATCH_FAILED if no other errors are encountered. While here, also add a test for the 'hunk applied with offset' case and shrink test_patch_dont_apply and illegal_status by taking advantage that 'got patch' doesn't stop at the first error. (And add some other cases to illegal_status too.) discussed with and ok stsp@

ok stsp@

ok stsp@

This introduce a proper got_patch specific progress callback that is invoked after processing a patch. It also drops the hackish printf in the modified case and takes `nop' out of the struct got_patch. ok stsp

ok op@

otherwise we call got_repo_close with repo being NULL and segfault there. ok stsp

Avoids overhead due to readdir calls while searching a pack index. ok op@

with lots of help from stsp for the manpage bits!

ok op@

Simplify some phrasing, explain what preconditions `got patch' has and what happens to the work tree when an error occurs.