| Message ID | 20230807170935.2336730-1-oswald.buddenhagen@gmx.de (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v2] sequencer: rectify empty hint in call of require_clean_work_tree() | expand |
Oswald Buddenhagen <oswald.buddenhagen@gmx.de> writes: > The canonical way to represent "no error hint" is making it null, which > shortcuts the error() call altogether. This fixes the output by removing > the line which said just "error:". > > Alternatively, one could make the function treat empty strings like null > strings, which would make it resemble its original script implementation > more closely, but that doesn't seem like a worthwhile goal. If anything, The analysis I gave you in the previous round was primarily to explain how this bug came to be, i.e. a misconversion of the scripted original, and that would be a lot more pertinent information than "is this closer to or further from the scripted original?" IOW, "equating NULL and an empty string is how the original implemented the logic correctly, but the misconversion made the current source to fail to do so" is the take-home message. You could correct the current version by making it follow the same "NULL and an empty string are the same" convention to implement the logic correctly, or you can build on the misconverted callee that treats only NULL specially hence pass NULL instead of "". Either one is a valid implementation, but the reason to choose the former would not be to "make it resemble the original". IOW, "doesn't seem like a worthwhile goal" is beating a strawman. Nobody advocates "if (!hint || !*hint)" because "the code must look exactly like the original". It is merely that it is one known way to achieve correctness, as it was how the original achieved its correctness. > I'd go in the opposite direction and assert() that the argument is not > an empty string. If we were writing this program from scratch, "NULL means something different from any sttring, and an empty string does not have anything special compared to any other strings" would probably be a good semantics to give to this helper function. I'd be OK with the change. Also, adding "if (!*hint) BUG(...)" would be a good future direction, I would think. > diff --git a/sequencer.c b/sequencer.c > index cc9821ece2..d15a7409d8 100644 > --- a/sequencer.c > +++ b/sequencer.c > @@ -6182,7 +6182,7 @@ int complete_action(struct repository *r, struct replay_opts *opts, unsigned fla > if (checkout_onto(r, opts, onto_name, &oid, orig_head)) > goto cleanup; > > - if (require_clean_work_tree(r, "rebase", "", 1, 1)) > + if (require_clean_work_tree(r, "rebase", NULL, 1, 1)) > goto cleanup; > > todo_list_write_total_nr(&new_todo);
Oswald Buddenhagen <oswald.buddenhagen@gmx.de> writes: > The canonical way to represent "no error hint" is making it null, which > shortcuts the error() call altogether. This fixes the output by removing > the line which said just "error:". Sorry, but I forgot to point out a rather obvious thing in my review. We would want to see a reproduction recipe described here in the proposed log message at least. Even better is an addition to an existing test to ensure that there is no such empty "error:" line, which will make sure that we will notice when anybody by mistake (this includes a mismerge of other topics) breaks this fix. Thanks. > sequencer.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/sequencer.c b/sequencer.c > index cc9821ece2..d15a7409d8 100644 > --- a/sequencer.c > +++ b/sequencer.c > @@ -6182,7 +6182,7 @@ int complete_action(struct repository *r, struct replay_opts *opts, unsigned fla > if (checkout_onto(r, opts, onto_name, &oid, orig_head)) > goto cleanup; > > - if (require_clean_work_tree(r, "rebase", "", 1, 1)) > + if (require_clean_work_tree(r, "rebase", NULL, 1, 1)) > goto cleanup; > > todo_list_write_total_nr(&new_todo);
diff --git a/sequencer.c b/sequencer.c index cc9821ece2..d15a7409d8 100644 --- a/sequencer.c +++ b/sequencer.c @@ -6182,7 +6182,7 @@ int complete_action(struct repository *r, struct replay_opts *opts, unsigned fla if (checkout_onto(r, opts, onto_name, &oid, orig_head)) goto cleanup; - if (require_clean_work_tree(r, "rebase", "", 1, 1)) + if (require_clean_work_tree(r, "rebase", NULL, 1, 1)) goto cleanup; todo_list_write_total_nr(&new_todo);
The canonical way to represent "no error hint" is making it null, which shortcuts the error() call altogether. This fixes the output by removing the line which said just "error:". Alternatively, one could make the function treat empty strings like null strings, which would make it resemble its original script implementation more closely, but that doesn't seem like a worthwhile goal. If anything, I'd go in the opposite direction and assert() that the argument is not an empty string. Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de> --- v2: - expanded commit message Cc: Junio C Hamano <gitster@pobox.com> --- sequencer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)