[colbymchenry]

Summary

A struct member whose type a macro renames now links what the macro names. redis-cli.c's sds members are hiredis's hisds, not src/sds.h's namesake typedef.

Second of a stack of four, on top of #3172.

Root cause

redis-cli.c and redis-benchmark.c include hiredis's sdscompat.h, which has #define sds hisds. The parser sees sds, and a macro with no parameters has no node, so #3144 linked 24 members to the only sds the project declares.

The fix

matchCppDeclaredType asks what the name is renamed to where it is written (src/resolution/c-macro-renames.ts).

• Finding renames. Headers are read once per resolver, by a search for define, into an index by macro name. A name nothing renames costs one lookup. The scan stops at 16 MB of headers: betaflight has 109 MB, nearly all vendor register maps.
• When one counts. As the preprocessor applies it: the translation unit from the asking file must have certainly defined the macro by that line. isCppMacroDefinedAt is the existing include walk, following one object-like macro. So not above the #include or the #define, and not under a flag no file sets.
• When it doesn't. Two arms that disagree say nothing, and a new name that is no project type leaves the name as written.

The same rule corrected two more that nobody had read:

• redis's core sets REDISMODULE_CORE, so RedisModuleString is robj in module.c, 19 edges. A module that includes the same header keeps the typedef.
• jq's #define YYLTYPE location: bison's struct is compiled out, 3 edges.

Validation

Windows 11, main 4499d00.

Tests. 5 of the 7 new tests fail on main and all pass. 9 mutations each fail a test.

A/B against main.

Repo	Language	Result
redis	C	43 moved: 24 to hisds, 19 to robj, each checked against the includes and the flag. Nothing lost
jq	C	3 moved to location, right
glfw, betaflight, rocksdb, leveldb	C, C++	nothing from this commit
gin, flask	controls	identical

No extraction change.

Cost. A resolver's scan is 26 ms on redis and 81 ms on rocksdb, once per worker; 49 ms of a redis index's profile in all. Whole-index CPU in the same run: redis +0.6%. rocksdb, 3 interleaved rounds on a busy machine: 132.3 to 129.7 s, level.

Left out. 551 calls edges in redis's CLI files have the same problem on both arms: #define sdslen hi_sdslen means sdslen(x) is hiredis's, and it links to src/sds.c's. The index added here would serve for that, but it is a change to call resolution and not on my list.

Suite. 857 files passed and 6 skipped, 10,651 tests passed, 2 expected fail, 87 skipped, 4 workers, with the kernel, on the stack one clean rebase back (719ea417 on main 3bd9949). After the rebase onto 087de4c, whose hunks are apart from these: 138 files and 2,224 tests of the C, C++, Objective-C, Python and extraction suites pass on the tip.

🤖 Generated with Claude Code