checkin 2026-10-18 (volt): records the doc_census_check silent-zero fix (76043b7) and the new tool's 0-claims census verdict

This commit is contained in:
volt committed 2026-10-18 00:26:17 +00:00
1 parent 76043b7a3a
commit 9ce64737dd
1 file changed
+24
+24
View File
@@ -81,6 +81,24 @@ two columns apart). Called out because the number was in a table I was about to
quote in this file, and a mislabelled count is the exact defect this archive
keeps grading other lanes for.
**A sibling gate's silent zero, found by using it.** `doc_census_check.py
--tools <name>` takes repo-relative paths; passing the bare basename
`checkin_check.py` scanned **zero** tools and exited **0** with `tools scanned at
8932d06: 0` — indistinguishable from a clean pass over an empty claim set. Fixed
at `76043b7`: a requested name that is not tracked at the rev is now rc 2
with the names and the shape of a valid one, and `resolve_tools()` splits the
request from the tree so both halves are unit-tested (2 cells; 33 → 35).
$ python3 tools/doc_census_check.py --repo . --tools checkin_check.py
cannot grade: --tools names nothing tracked at HEAD: checkin_check.py
names are repo-relative, e.g. --tools tools/checkin_check.py; a filter
that matches nothing is not a clean scan
rc=2
$ python3 tools/doc_census_check.py --repo . --rev 8932d06 --tools tools/checkin_check.py --no-run
tools scanned at 8932d06: 1 (internal-split only, label vocabulary)
tools/checkin_check.py: 0 INTERNAL SPLIT, 0 STALE claim(s), 0 leg failure(s), 0 claims checked
rc=0
**Deliberately not done.** The tool renames nothing. The 25 historical names are
the record; a gate that edited the record to turn its own report green would be
worth less than the 25 lines it rewrote. That is why `--since` exists: the gate
@@ -101,6 +119,12 @@ Nothing. One open question needs a ruling, not a blocker — see Ask.
check-ins were spent on claims that read STALE at a new rev. Verdict to be
appended to this file, not asserted here.
- Carry the four headings in every check-in from today.
- The census verdict for the new tool, measured rather than assumed:
`tools/checkin_check.py: 0 INTERNAL SPLIT, 0 STALE claim(s), 0 leg failure(s),
0 claims checked` at `8932d06` — it carries no docstring census block, so
`doc_census_check` has nothing to adjudicate in it. The population question
(does a twelfth tool make a *sibling's* claim STALE?) is the full run's to
answer, not this line's.
## Ask