Repository navigation
spec(widgets): analog-v01-acceptance, the SAR ADC acceptance scoreboard of t27 analog v0.1 as a spec-generated widget (Refs #11972) - #12025
Conversation
…oard of t27 analog v0.1 as a spec-generated widget (Refs #11972) One table (9 of 13 blocks pass, 0 fail, 4 not modelled; 26 / 0 / 4 of 30 rows) and one diagram (the dynamic comparator decision time, 613.07 ps in t27 against 602.54 ps in ngspice, +1.75 %). Every number is copied from docs/analog/v0.1-release-notes.md (PR #11973) and pinned by the spec's own tests: the sums (9 + 0 + 4 = 13, 26 + 0 + 4 = 30), each table cell and flow step against its K_ constant, and the 1.75 % recomputed from the two decision times in integers. A wrong number fails the compile. The page says the four not-modelled blocks stay open, that integration is a behavioural converter fitted to the golden deck, and that nothing here is a speed claim. The page, tool.js and card live in gHashTag/trinity; this is the byte-identical source. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
🤖 Generated with Claude Code |
|
Three checks are red on this PR; none of them is caused by
I did not touch the seal ledger or the dupe ledger in this PR, to keep it to one spec file. The new checks that are still running ( 🤖 Generated with Claude Code |
Summary by t27-beesChecked by the t27 compilert27c parse, parse-complete and typecheck on each changed .t27 file at
Read by the modelThis adds a spec-generated widget ( Worth a look
Read: 1 files, +141 -0, the whole diff. Head |
…what the row test can carry, state the gap window's unit (Refs #11972) Review of the bot on the first push: - the Total row's last cell (index 15) was the one table cell no test pinned; it is now asserted - the row test was named for more than the counts can prove (the release notes count rows per class, not per block); renamed to each_class_has_at_least_as_many_rows_as_blocks, with a line saying so - the comparator gap window is 50 ppm either side (half of the last digit shown), not 0.5 %; a comment now states the unit Not changed: K_TABLE_CELLS stays u16 and DATA_SOURCES keeps its caveat entry, as in the other widget specs. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… regenerated (Refs #1377) The spec is byte-identical to specs/widgets/analog_v01_acceptance.t27 in Trinity-S-AI-DNA/t27#12025 after its first review: the Total row's last table cell is now pinned by a test, the row test is renamed to what its counts can prove, and the gap window's unit is stated. The page text is unchanged; the generated copy and the page stamp follow the spec. 604 spec tests, 1937 asserts, all hold. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
PR DashboardGenerated at: 2026-10-24 15:09:00 UTC
Summary
Seal Status
|
… regenerated (Refs #1377) The spec is byte-identical to specs/widgets/analog_v01_acceptance.t27 in Trinity-S-AI-DNA/t27#12025 after its first review: the Total row's last table cell is now pinned by a test, the row test is renamed to what its counts can prove, and the gap window's unit is stated. The page text is unchanged; the generated copy and the page stamp follow the spec. 604 spec tests, 1937 asserts, all hold. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
Merged on explicit owner instruction: the owner overrode the |
What this adds
specs/widgets/analog_v01_acceptance.t27: the source of the spec-generated tool widgetanalog-v01-acceptance(category fpga) that shows the v0.1 acceptance scoreboard of t27 analog emulation, and one comparator number beside it. The page, the card and the gallery entry live in gHashTag/trinity (companion PR below); this repository holds the spec, as for the other widget specs here.What the widget shows:
Where each number comes from
Every number is copied from
docs/analog/v0.1-release-notes.md(PR #11973, merged at head37eb0fad2, re-checked on master just now): its "Acceptance" table (the per-block checks ofspecs/analog/accept_sar.t27; counts pinned there by the testsas_all_blocksandas_all_rows) and its M4b-q row (#11325). Nothing is measured on this page and nothing else is used as a source.A wrong number fails the compile of the spec, not a review. The
K_*constants hold each count once; the spec's 10 tests add them up (9 + 0 + 4 = 13, 26 + 0 + 4 = 30), pin every table cell and every diagram step to its constant, and recompute the 1.75 % from the two decision times in integers (61307 and 60254 hundredths of a picosecond). Two negative controls, each changing one number in its constant only, fail the compile as intended.Checks run
widget-pages-from-spec.mjs --checkup to date, 194 pages, 604 spec tests, 1937 asserts, all hold. Companion PR has the rest of the local gates.t27c test-reporton this file (master-built t27c, run in a T27 OS shell pane, not on the Mac's shell): 10 tests, 10 pass, 0 fail. Runtime asserts executed per test: 15, 1, 3 for the three tests that compare strings (the table cells, the note, the diagram steps); 0 for the seven tests that compare only integer constants. The compiler folds integer-only asserts, so the report marks those 7 as vacuous (T730). Not hidden: the same holds for the template tests of every widget spec (for examplebasics-arrays: 2 of 2 vacuous). I tried to move the sums and the 1.75 % intopub fnso they run at runtime; the page generator's test evaluator refuses calls in tests (unsupported expression ExprCall), so the integer proofs hold in the generator's evaluation (which is where the page is gated) and stay constant-folded in the native report. The string-based tests, which pin the page text to the numbers, do execute natively.specs/widgets/carry no seal in.trinity/seals, and a seal saved from a worktree under/Users/...can embed the path; if the maintainers want one, it comes from a master-built t27c.tri widget doctor --repo <trinity checkout>: 164 anomalies, exit 1 before the deploy. 162 are the same pre-existing "css missing or wrong" line printed for widgets of the one-importtool.jsshape (for examplebasics-arrays), 1 is the same line for this widget (same shape), and 1 islive: analog-v01-acceptance 404 200(the page is not deployed yet; a merged trinity PR is live only after the release PR in 999). Nothing it prints is new in kind. Full output is attached as a comment.specs/; no hand-written non-t27 code is added here (only-t27).After the first review
The bot review of the spec found the Total row's last table cell unpinned (now asserted), a row test named for more than its counts prove (renamed
each_class_has_at_least_as_many_rows_as_blocks) and a gap window whose unit was not stated (now a comment: 50 ppm either side, half of the last digit shown, not 0.5 %).K_TABLE_CELLSstaysu16and the secondDATA_SOURCESentry stays, as in the other widget specs. The page text is unchanged; asserts went from 1936 to 1937.Blog
No analog post exists in trinity's
bodies/yet; the embed goes in the blog PR that announces v0.1 (Trinity-S-AI-DNA/trinity#1733, postanalog-emulation-v0-1.ts, body and ruBody). Embed address once live:https://t27.ai/widgets/analog-v01-acceptance/?embed=1(tool embeds are 640 wide, 460 high).Refs #11972, see #10486. Companion PR (page, card, gallery, course lesson, with the v0.1 post and the embed): Trinity-S-AI-DNA/trinity#1733 (the first companion, trinity#1735, was closed as superseded by it)
No auto-merge is enabled.
🤖 Generated with Claude Code