A larger v2, this: * Renames string_list.cocci to string-list.cocci (to be consistent with the string-list.[ch]) * Answers Elijah's question about why the previous 2/2 left out one case, it's due to --disable-worth-trying-opt. That's now split out into its own commit. FWIW it's the only case with "make coccicheck" that we've "missed", including existing rules, but we just got lucky there. I have some follow-up patches to fix various such bugs in our coccicheck, but let's leave the larger fixes for now. * The new 4/6 duplicates Junio's https://lore.kernel.org/git/xmqq7d471dns.fsf@gitster.g/, but now we've got a coccinelle rule for it and similar inverse cases. It's still the only in-tree case, but it's probably good to keep the rule around regardless. I could also eject that *.C change and say this is based on Junio's, or both could make the same change and "merge" will sort it out. I do think 4/6 is a bit easier to read with an actual case where we change things, but then again there's the added tests... * A new 5/6 tweaks some "strdup_strings" dancing to a simpler pattern, this is adjusted from my local WIP branch of larger string_list API fixes. * A new 6/6 fixes a confusing case of dup v.s. nodup to use a more obvious callback pattern where we don't flip-flop. Ævar Arnfjörð Bjarmason (6): string_list API users + cocci: use string_list_init_dup() cocci: apply string_list.cocci with --disable-worth-trying-opt reflog-walk.c: use string_list_init_dup() cocci: add "string_list" rule to swap "DUP" <-> "NODUP" string-list API users: don't tweak "strdup_strings" to free dupes notes.c: make "struct string_list display_notes_refs" non-static bisect.c | 7 +++--- builtin/remote.c | 3 +-- contrib/coccinelle/string-list.cocci | 26 ++++++++++++++++++++++ contrib/coccinelle/tests/string-list.c | 20 +++++++++++++++++ contrib/coccinelle/tests/string-list.res | 18 +++++++++++++++ notes-utils.c | 4 ++-- notes.c | 28 ++++++++++++++---------- reflog-walk.c | 2 +- refs.c | 4 ++-- resolve-undo.c | 8 +++---- revision.c | 4 ++-- 11 files changed, 95 insertions(+), 29 deletions(-) create mode 100644 contrib/coccinelle/string-list.cocci create mode 100644 contrib/coccinelle/tests/string-list.c create mode 100644 contrib/coccinelle/tests/string-list.res Range-diff against v1: 1: c89758491e7 ! 1: 61a62bdf8e9 string_list API users + cocci: use string_list_init_dup() @@ Commit message Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> - ## contrib/coccinelle/string_list.cocci (new) ## + ## contrib/coccinelle/string-list.cocci (new) ## @@ +@@ +struct string_list *P; @@ contrib/coccinelle/string_list.cocci (new) +- (P)->strdup_strings = 1; ++ string_list_init_dup(P); - ## contrib/coccinelle/tests/string_list.c (new) ## + ## contrib/coccinelle/tests/string-list.c (new) ## @@ +int init(void) +{ @@ contrib/coccinelle/tests/string_list.c (new) + list->strdup_strings = 1; +} - ## contrib/coccinelle/tests/string_list.res (new) ## + ## contrib/coccinelle/tests/string-list.res (new) ## @@ +int init(void) +{ 2: 5d8baa9cbc4 ! 2: 33e551a2f4c string-list API users: manually use string_list_init_*() @@ Metadata Author: Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> ## Commit message ## - string-list API users: manually use string_list_init_*() + cocci: apply string_list.cocci with --disable-worth-trying-opt - Convert various code that didn't use string_list_init_*() to do so, in - cases where the only thing being allocated was the string list we can - change from CALLOC_ARRAY() to ALLOC_ARRAY(), the string_list_init_*() - function will zero out the memory. + Apply the new string-list.cocci added in the preceding commit with + --disable-worth-trying-opt. For optimization purposes we run spatch in + a mode where even though we run it with --all-includes we'll miss some + changes because we don't use --disable-worth-trying-opt. - This covers cases that weren't matched by tho coccinelle rule in the - preceding commit, which is conservative enough to care about the type - of what we're modifying. + This is because without that option it'll take a look at + notes-utils.c, and conclude that it doesn't need to process + it (irrelevant output excluded with "[...]"): + + $ spatch --sp-file contrib/coccinelle/string-list.cocci --patch . notes-utils.c + [...] + (ONCE) Expected tokens string_list strdup_strings CALLOC_ARRAY + Skipping: notes-utils.c + + This is just one of the known (and probably some unknown) issues where + our "make coccicheck" fails to include changes for whatever + reason. That should be fixed more generally, but let's just fix this + manually for now. Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> @@ notes-utils.c: struct notes_rewrite_cfg *init_copy_notes_for_rewrite(const char c->refs_from_env = 0; c->mode_from_env = 0; if (rewrite_mode_env) { - - ## reflog-walk.c ## -@@ reflog-walk.c: struct reflog_walk_info { - void init_reflog_walk(struct reflog_walk_info **info) - { - CALLOC_ARRAY(*info, 1); -- (*info)->complete_reflogs.strdup_strings = 1; -+ string_list_init_dup(&((*info)->complete_reflogs)); - } - - void reflog_walk_info_release(struct reflog_walk_info *info) -: ----------- > 3: 62aab32ae77 reflog-walk.c: use string_list_init_dup() -: ----------- > 4: 2d858c49243 cocci: add "string_list" rule to swap "DUP" <-> "NODUP" -: ----------- > 5: 8c0ac6cbd96 string-list API users: don't tweak "strdup_strings" to free dupes -: ----------- > 6: b0de7a63d1c notes.c: make "struct string_list display_notes_refs" non-static -- 2.37.1.1095.g64a1e8362fd
