Drew T
987 B
§356 — MEASURE A DRAFT IN THE TU IT WILL LIVE IN, NOT IN THE STANDALONE PROBE (P31 S68; measured over 47 stored drafts)
match_one compiles a draft against common.h only. A real TU carries a whole file-scope
declaration environment. Sizing a lane by the standalone verdict undercounts it badly:
Of 47 backlog drafts reported
cc1-fail, 43 failed with`D_XXXXXXXX' undeclared. 39 of those 43 name symbols that ARE present in the draft's own target TU. The failure was the probe's environment, not the draft. Only 4 were genuinely missing.
The honest instrument is recover_integration --probe-only, which compiles in the REAL TU and
reports MATCH / DIFF / CC1-FAIL separately from the static guess. Here the standalone verdict
said 22 candidates; the real-TU population is up to 61. Extends R35 — fix the instrument before
trusting the measurement — to the specific case where the instrument is narrower than reality
rather than broken.