Skip to content

t27b ledger: bless from master a7fa547a6, two reference-bug rows, cap 144 -> 125 (Closes #11924) - #11925

Merged
gHashTag merged 1 commit into
masterfrom
night/36-t27b-ledger
Oct 24, 2026
Merged

gHashTag merged 1 commit into
masterfrom
night/36-t27b-ledger

Conversation

@gHashTag

Copy link
Copy Markdown
Collaborator

Closes #11924

docs/reports/t27b_expectations.json is blessed from the t27b-native-run artifact of master run 40038516068 (a7fa547). That is the same chore as #11723. Master tip 3e47410 differs from a7fa547 only in seals.

Ratchet checked with the new ledger against the three CI artifacts:

run mode verdict
a7fa547 (master) plain green, all counts 0
668f73b (#11922) --pr green (UNEXPECTED PASS 1 = its own fix)
1d4325b (#11920) --pr green (UNEXPECTED PASS 1 = its own fix)

Once #11920 and #11922 land, master shows UNEXPECTED PASS on those two rows until the next bless. That means the ledger is behind, not that something regressed.

No spec changed, so there is no t27c/t27b leg to run. Both compilers' verdicts are the ones in the master run artifact.

🤖 Generated with Claude Code

…ap 144 -> 125 (Closes #11924)

The ledger came from d800c56. It did not name 56 specs that have
landed since, so master t27b native was red. Every PR was also red in
--pr mode on a jit/interp mismatch in a file it does not touch: #11920
on frame_render, #11922 on lanes_view. Each fix was held red by the
other one.

Both mismatches are defects in the specs: the tests read bytes they
never wrote. They go in by hand with reason reference-bug, as the
bless refusal asks.

Source: artifact t27b-native-run of run 40038516068. Ratchet with the
new ledger: master run green (all counts 0); --pr on 668f73b (#11922)
and on 1d4325b (#11920) green.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-24 00:04:27 UTC

Summary

Status Count
Total Open PRs 44
PRs with Failing Checks 39
PRs with All Checks Green 5
READY 4
FAILING 39
PENDING 0
NO CHECKS YET 0

These columns do not partition: 4 + 39 + 0 + 0 = 43, and there are 44 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=5aec047d42ec != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@tri-claw

tri-claw Bot commented Oct 24, 2026

Copy link
Copy Markdown

Summary by t27-bees

This PR updates the t27b expectations ledger to a new baseline from master commit a7fa547. The cap for allowed non-passing specs drops from 144 to 125, many previously blocked specs now pass, and two specs are explicitly marked as reference-bug mismatches where t27b correctly faults on reading unwritten bytes.

Worth a look

  • docs/reports/t27b_expectations.json:1717 (specs/t27f/frame_render.t27) — Marked mismatch with reason reference-bug; blocker shows JIT returns Ok(123) but interpreter faults on "1-byte read at stack+0x0 of bytes never written". This is intentional per the PR description, but will show as UNEXPECTED PASS once test(t27f): frame_render clears the raster before it blends (Refs #3) #11922 lands until the next bless.
  • docs/reports/t27b_expectations.json:1825 (specs/t27os/lanes_view.t27) — Marked mismatch with reason reference-bug; blocker shows JIT returns Ok(1) but interpreter faults on "1-byte read at stack+0x952 of bytes never written". Same intent as above; will show UNEXPECTED PASS once fix(t27os): lv_copy_row copies the row's written bytes, not the dead tail (Closes #11919) #11920 lands.
  • docs/reports/t27b_expectations.json:877 (specs/port/browseros/trios/agent-server/apps/server/src/tools/memory/update-soul.t27) — Status frontend with a truncated typecheck error blocker mentioning SoulUpdateTool field execute at line 53. The blocker text appears cut off mid-sentence; verify the full error is captured or the ledger entry will be misleading.
  • docs/reports/t27b_expectations.json:13 — max_not_pass reduced from 144 to 125. Current ratchet shows 0 non-passing, but any future regression pushing non-pass count above 125 (previously 144) will now fail the gate.
  • docs/reports/t27b_expectations.json:679, 682, 683, 1169, 1299 — Several blockers changed from ExprCall(undeclared fn) to ExprCall(@​sin) (e.g., specs/nn/attention.t27, specs/nn/hslm.t27, specs/port/trinity/src/phi-engine/scheduler/phi_spiral.t27, specs/t27f/particle_engine.t27, specs/port/trinity/src/vibeec/flash_benchmark.t27). This reflects a more precise diagnosis (missing sin intrinsic) rather than a generic undeclared function; no functional change but worth noting the specificity shift.

Read: 1 files, +89 -33, the whole diff. Head 1b19ae5. Model: openai-compatible/nvidia/nemotron-3-ultra-550b-a55b. Commands: @t27-bees review | summary | help | pause | resume. Free for every repository.

@gHashTag
gHashTag merged commit d8dd43f into master Oct 24, 2026
28 checks passed
@gHashTag
gHashTag deleted the night/36-t27b-ledger branch October 26, 2026 03:45
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

t27b ledger lags master: 56 unlisted, and two PRs each red on the other's mismatch

1 participant