Skip to content

A false zero punished a bee for a generator defect - #4468

Merged
gHashTag merged 1 commit into
masterfrom
fix/false-zero
Sep 20, 2026
Merged

gHashTag merged 1 commit into
masterfrom
fix/false-zero

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

Closes #4467

Issue #4446 went out with:

- 5. `t27c test-report specs/fpga/linker.t27 ... | grep -c BLOCKED` prints `0`
     - this spec compiles today and must still compile (today: 0)

It has never compiled. Every criterion run against the bee's own branch:

1 ok   2 ok   3 ok   4 ok   5 FAILS — and fails identically on master

The bee added a correct four-line test, met 4 of 5, and was sent back for a
generator defect it could not fix inside its boundary: the Zig backend emits
.align = 4, and align is a reserved word (#4466).

grep -c prints 0 for an empty input, so a t27c that did not run produces
the same 0 as a spec that compiles. The report must now say it ran before
its count is believed.

Repaired across the open backlog with the distinction that matters:

5 issues spec exists and does not compile → criterion removed, it asks for what the boundary cannot deliver
49 port issues spec does not exist yet → criterion kept; a file a bee is about to create has no excuse for not compiling

The first pass removed all 54. The correction reads the filesystem rather than
the exit code.

🤖 Generated with Claude Code

Issue #4446 went out carrying "this spec compiles today and must still compile
(today: 0)" for specs/fpga/linker.t27, which has never compiled: its generated
Zig writes `.align = 4`, and `align` is a reserved word in Zig, so the file
fails with `expected field initializer` at spec.zig:26:84.

The bee added a correct four-line test, met four criteria of five, and was sent
back for a defect it did not cause and could not fix inside its boundary.
Checked by running every criterion against the bee's own branch: 1 ok, 2 ok,
3 ok, 4 ok, 5 fails - and 5 fails identically on master.

`grep -c` prints 0 for an empty input, so a t27c that did not run produces the
same `0` as a spec that compiles. The report must now SAY it ran - the words
`test report:` - before its count is believed.

Repaired in the open backlog with the distinction that matters: 5 issues whose
spec exists and does not compile lost the criterion, and 49 port issues whose
spec does not exist YET kept it, because a file a bee is about to create has no
excuse for not compiling. The first pass removed all 54; the correction reads
the filesystem rather than the exit code.

The generator defect is #4466.

Closes #4467

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@gHashTag
gHashTag enabled auto-merge (squash) September 20, 2026 16:03
@github-actions

Copy link
Copy Markdown
Contributor

📓 NotebookLM Notebook linked to this PR

This notebook contains session context, decisions, and artifacts for this work.

@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-09-20 16:03:57 UTC

Summary

Status Count
Total Open PRs 21
PRs with Failing Checks 19
PRs with All Checks Green 2
READY 2
FAILING 19
PENDING 0
NO CHECKS YET 0

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=403499176a5d != 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).

@gHashTag
gHashTag merged commit 93e2c3a into master Sep 20, 2026
30 of 34 checks passed
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.

A false zero became a criterion: grep -c prints 0 for an input that never arrived

2 participants