| Message ID | 20201215235027.10401-3-avarab@gmail.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 3eab6be40114931d575cf1ed85499adeba51aad6 |
| Headers | show |
| Series | [1/2] strmap: do not "return" in a void function | expand |
Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes: > Remove this unreachable code. It was found by SunCC, it's found by a > non-fatal warning emitted by SunCC. It's one of the things it's more > vehement about than GCC & Clang. This is a borderline Meh to me. I am even tempted to suggest that, unless all other case arms return, iow, if there is even a single arm that breaks, it may even be more future-proof to end any and all case arms that do not fall-thru to consistently end with break. If there is some way to fix the compiler, that may be preferrable, but as I said, this is borderline Meh and I do not care too deeply either way. Thanks. > These return/break cases are just unnecessary however, and as seen > here the surrounding code just did a plain "return" without a "break" > already. > > Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> > --- > apply.c | 2 -- > builtin/fast-export.c | 1 - > 2 files changed, 3 deletions(-) > > diff --git a/apply.c b/apply.c > index 4a4e9a0158c..668b16e9893 100644 > --- a/apply.c > +++ b/apply.c > @@ -3948,10 +3948,8 @@ static int check_patch(struct apply_state *state, struct patch *patch) > break; /* happy */ > case EXISTS_IN_INDEX: > return error(_("%s: already exists in index"), new_name); > - break; > case EXISTS_IN_INDEX_AS_ITA: > return error(_("%s: does not match index"), new_name); > - break; > case EXISTS_IN_WORKTREE: > return error(_("%s: already exists in working directory"), > new_name); > diff --git a/builtin/fast-export.c b/builtin/fast-export.c > index d2e33f50052..0a60356b06e 100644 > --- a/builtin/fast-export.c > +++ b/builtin/fast-export.c > @@ -923,7 +923,6 @@ static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name) > if (!tag) > die("Tag %s points nowhere?", e->name); > return (struct commit *)tag; > - break; > } > default: > return NULL;
Junio C Hamano <gitster@pobox.com> writes: > Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes: > >> Remove this unreachable code. It was found by SunCC, it's found by a >> non-fatal warning emitted by SunCC. It's one of the things it's more >> vehement about than GCC & Clang. > > This is a borderline Meh to me. > > I am even tempted to suggest that, unless all other case arms > return, iow, if there is even a single arm that breaks, it may even > be more future-proof to end any and all case arms that do not > fall-thru to consistently end with break. > > If there is some way to fix the compiler, that may be preferrable, > but as I said, this is borderline Meh and I do not care too deeply > either way. > > Thanks. Heh, this patch cleanly applies even to the tip of 'maint', which means it is not a new issue at all. Is this caused by more recent SunCC than you used in the past?
diff --git a/apply.c b/apply.c index 4a4e9a0158c..668b16e9893 100644 --- a/apply.c +++ b/apply.c @@ -3948,10 +3948,8 @@ static int check_patch(struct apply_state *state, struct patch *patch) break; /* happy */ case EXISTS_IN_INDEX: return error(_("%s: already exists in index"), new_name); - break; case EXISTS_IN_INDEX_AS_ITA: return error(_("%s: does not match index"), new_name); - break; case EXISTS_IN_WORKTREE: return error(_("%s: already exists in working directory"), new_name); diff --git a/builtin/fast-export.c b/builtin/fast-export.c index d2e33f50052..0a60356b06e 100644 --- a/builtin/fast-export.c +++ b/builtin/fast-export.c @@ -923,7 +923,6 @@ static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name) if (!tag) die("Tag %s points nowhere?", e->name); return (struct commit *)tag; - break; } default: return NULL;
Remove this unreachable code. It was found by SunCC, it's found by a non-fatal warning emitted by SunCC. It's one of the things it's more vehement about than GCC & Clang. It complains about a lot of other similarly unreachable code, e.g. a BUG(...) without a "return", and a "return 0" after a long if/else, both of whom have "return" statements. Those are also genuine redundancies to a compiler, but arguably make the code a bit easier to read & less fragile to maintain. These return/break cases are just unnecessary however, and as seen here the surrounding code just did a plain "return" without a "break" already. Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> --- apply.c | 2 -- builtin/fast-export.c | 1 - 2 files changed, 3 deletions(-)