gate: make green mean green -- fail-closed evidence, and one product profile - #2

Closed

ebt55 wants to merge 29 commits into

mainfrom

claude/exactdoc-pr1-gate

Commits on Jul 29, 2026

identity: reset to 0.1.0a1/Alpha and reconcile every claim with STATUS.md

ebt55
and
claude
committed
95fcb9d
testkit: a missing oracle is a skip, a broken one is an error (corpus 15 -> 16)

ebt55
and
claude
committed
cab3939
scripts: one command provisions the harness, and probes what it found

ebt55
and
claude
committed
a3139f1
git: keep shell scripts LF on checkout

ebt55
and
claude
committed
11ee43f
golden_ir: record the environment a golden was frozen in

ebt55
and
claude
committed
0e368d4
golden_ir: re-freeze on Linux to attach the manifest -- provably content-neutral

ebt55
and
claude
committed
03c6948
harness: count CJK by character, not by whitespace (c4_i18n recall .83 -> 1.00)

ebt55
and
claude
committed
b8bcf30
testkit: gate on NEW failures, so the gate can fail at all (exit 1 -> 0 on Linux)

ebt55
and
claude
committed
41e5064
ci: the gate can now fail, so stop ignoring it

ebt55
and
claude
committed
98d8e60
docs: record the Linux/CI baseline and close M1

ebt55
and
claude
committed
d4ca28f
testkit: a span-level microscope, and --only, for the code-heavy pair

ebt55
and
claude
committed
24376a7
docs: the parity gate is the contract, the golden IR is a microscope

ebt55
and
claude
committed
34d9183
pdfium: put back the leading indentation PDFium does not report (c7_code .16 -> .59)

ebt55
and
claude
committed
60b9c29
SESSIONS: 03_tech traced to the stroke-bbox convention; the fix measured worse

ebt55
and
claude
committed
dadb36e
docs: CI is the number of record, and it has now actually run

ebt55
and
claude
committed
eced9ee
testkit: probe which space PDFium reports path points in (612 paths, 578 transformed)

ebt55
and
claude
committed
52ced06
pdfium: put path geometry in page space (c7_code .59 -> .76, 8 regressions held)

ebt55
and
claude
committed
c0d09b7
testkit: plot the distributions the block-split threshold separates (§12.6)

ebt55
and
claude
committed
bb39e2a
pdfium: split blocks on the BODY pitch, not the median (8 regressions -> 6)

ebt55
and
claude
committed
c10b3a7
pdfium: take x from the metric box, as y already did (6 regressions -> 3)

ebt55
and
claude
committed
49a12ff
pdfium: a span ends where the style ends, and a generated space has a width

ebt55
and
claude
committed
41770e1
pdfium: drop the end-of-line space PDFium invents (03_tech text diff 33% -> 0%)

ebt55
and
claude
committed
9976d2a
pdfium: one space is one space, however far justification stretches it

ebt55
and
claude
committed
4a5d2a3
pdfium: body pitch per type size, not per page (3 regressions -> 2)

ebt55
and
claude
committed
d2a0356
testkit: histogram the vertical error, and locate the last constant offset

ebt55
and
claude
committed
aeec173
git: ignore the planning docs by name shape, not one at a time

ebt55
and
claude
committed
1ac6304
infer: the granted line-box escalation, built and reverted -- fallback (c)

ebt55
and
claude
committed
b1c90d2
docs: ROADMAP.md, and bring every committed doc up to where the project is

ebt55
and
claude
committed
0cd7d11
gate: make green mean green -- fail-closed evidence, and one product profile

ebt55
and
claude
committed
6d8c47b

Navigation Menu

gate: make green mean green -- fail-closed evidence, and one product profile - #2

gate: make green mean green -- fail-closed evidence, and one product profile#2
ebt55 wants to merge 29 commits into
mainfrom
claude/exactdoc-pr1-gate

Commits on Jul 29, 2026

identity: reset to 0.1.0a1/Alpha and reconcile every claim with STATUS.md

testkit: a missing oracle is a skip, a broken one is an error (corpus 15 -> 16)

scripts: one command provisions the harness, and probes what it found

git: keep shell scripts LF on checkout

golden_ir: record the environment a golden was frozen in

golden_ir: re-freeze on Linux to attach the manifest -- provably content-neutral

harness: count CJK by character, not by whitespace (c4_i18n recall .83 -> 1.00)

testkit: gate on NEW failures, so the gate can fail at all (exit 1 -> 0 on Linux)

ci: the gate can now fail, so stop ignoring it

docs: record the Linux/CI baseline and close M1

testkit: a span-level microscope, and --only, for the code-heavy pair

docs: the parity gate is the contract, the golden IR is a microscope

pdfium: put back the leading indentation PDFium does not report (c7_code .16 -> .59)

SESSIONS: 03_tech traced to the stroke-bbox convention; the fix measured worse

docs: CI is the number of record, and it has now actually run

testkit: probe which space PDFium reports path points in (612 paths, 578 transformed)

pdfium: put path geometry in page space (c7_code .59 -> .76, 8 regressions held)

testkit: plot the distributions the block-split threshold separates (§12.6)

pdfium: split blocks on the BODY pitch, not the median (8 regressions -> 6)

pdfium: take x from the metric box, as y already did (6 regressions -> 3)

pdfium: a span ends where the style ends, and a generated space has a width

pdfium: drop the end-of-line space PDFium invents (03_tech text diff 33% -> 0%)

pdfium: one space is one space, however far justification stretches it

pdfium: body pitch per type size, not per page (3 regressions -> 2)

testkit: histogram the vertical error, and locate the last constant offset

git: ignore the planning docs by name shape, not one at a time

infer: the granted line-box escalation, built and reverted -- fallback (c)

docs: ROADMAP.md, and bring every committed doc up to where the project is

gate: make green mean green -- fail-closed evidence, and one product profile