Skip to content

Epic review fixes for #1206 - #1208

Merged
neuromechanist merged 16 commits into
feature/issue-1181-epic-zarr-servingfrom
feature/issue-1181-review-fixes
Sep 3, 2026
Merged

neuromechanist merged 16 commits into
feature/issue-1181-epic-zarr-servingfrom
feature/issue-1181-review-fixes

Conversation

@neuromechanist

@neuromechanist neuromechanist commented Sep 3, 2026 •

Copy link
Copy Markdown
Contributor

Part of epic #1181, addresses the six-lens review of #1206.

Sixteen atomic commits, grouped below by the finding ids they close (rounds 1 to 4 first, then rounds 5 and 6). Every MUST and SHOULD from the type, comment, converter, silent-failure, backend, test-coverage and editor-diagnostic lenses is applied; nothing was skipped as a false positive.

Commits by finding id

Types (T)

  • 69a9c89f Validate zarr-ready counts as integers — T1, T2, T7. An integer count primitive in zarr-ready.ts (z.number().int().finite().nonnegative()) for every count-shaped field, so a fractional store_count is dropped and logged like any other malformed field instead of being bound into an INTEGER column as a REAL. data_failures is retyped unknown[] to match its validator and ZarrDataFailure is deleted (nothing read its fields). not_attempted_count and non_raw_dropped join the summary log line; both were validated and then read by nothing.
  • f7409876 Note the zarr index trust boundary and one cast — T3, T9. One-line trust-boundary notes on ZarrIndexJson (s3.ts) and ZarrFidelityIndexJson (zarr-fidelity-sweep.ts): the coverage equation is the producer's invariant (check_index_invariant), not something these types assert. A SAFETY: comment on the rateLimiter context cast in zarr-data.ts.
  • T5 and T6 ride in 285213b5 below, because they touch the same file as V1: PendingEntry.reason becomes the schema's closed three-value enum (normalized through _pending_reason when read off a published index, so a bad value cannot fail validation at the END of a conversion), and EventsPublication.file becomes a ManifestFileEntry mirroring the manifest schema. FailureEntry.code stays str, an open set.

Comments (C)

  • 284cdcdf Correct the dev-D1 cron allowlist rationale — C2. The comment called dev D1 a partial production mirror; the catalog was purged to the curated fixtures. The allowlist survives that purge for the reasons AGENTS.md gives: the users table was not purged, the dev worker holds a real RESEND_API_KEY, and the nemarDatasets org is shared.
  • 590d2cb5 Reconcile the staging cron and two ops claims — C1, V2, V3. systems-inventory and validated_workflows contradicted each other on the staging nightly cron; both now carry the installed crontab line verbatim, including the explicit ZARR_DRIVER_REF override and the note that it flips to dev once the epic merges.

Converter and ops (V)

  • 285213b5 Publish index.json as a conditional write — V1, V4 (plus T5, T6). Two processes write <id>/zarr/index.json and both read it, work for minutes to hours, then write back what they merged from that read, so an unconditional PUT from either silently reverted the other with both exiting 0. The PUT now carries --if-match with the ETag read at the top of the run (--if-none-match "*" when there was no document). A 412 is re-read, re-merged onto the newer document and retried once; a second 412 abandons the publish through write_failed_callback. The stub aws in test_generate_zarr.py gains the real ETag semantics the purge stub already had; tests cover a clean write, one 412 then success, and two 412s failing loudly. The pre-existing F811 duplicate import is removed.
  • V2 and V3 in 590d2cb5: the --test guard-rail claim is corrected (only API_BASE and TEST_API_URL are case-folded; the dirs are path-normalized and the bucket/profile are exact equality), and --preview-engine-bump's "no writes" claim is reworded, since it connects and so runs migrate_schema.

Silent failures (S)

  • 6db500f8 Serve the zarr summary keys D1 already stores — S2 (with B4, B5).
  • e22bcca0 Stop erroring datasets starving the fidelity sweep — S1 (with X3). Candidacy clears only on a verdict and the batch was ordered by dataset_id, so roughly 25 alphabetically early datasets that error every run held the whole quota forever and nothing behind them was ever swept, while each run reported errors and looked like progress. Every attempt now stamps zarr_verify_attempted_at, verdict or not, and the batch is ordered by it (never-attempted first, then oldest attempt). An unreachable dataset still records NO verdict and stays a candidate.
  • 3bb48710 Fail loudly when an S3 listing fails — S3, S4 (with P1 to P4). _s3_child_prefixes returned [] for both "no subdirectories" and "the call failed"; list_dataset_ids read that as "this bucket holds no datasets". It now raises, _rm_recursive keeps its tolerance at the call site, and purge --all with zero datasets exits 2 like the snapshot-directory check.
  • 898f7de7 Say what a 403 and a redirect row actually mean — S5, S6 (with B3, X5). The sweep's "index.json is absent" message now names the other reading of the same response, so an access regression is not read as a per-dataset converter bug; chunk-redirect access rows are documented as intent rather than delivery.

Backend (B)

  • 6db500f8 — B4, B5. The callback wrote pending/discovered into zarr_data_failures (Per-dataset Zarr coverage: every raw recording accounted for in index.json, counts in D1, detail off-database #1197), the catalog projection kept only count/detail_ref, and the contract stripped what it did not declare, so counts that existed in D1 reached nobody while every layer's own tests passed. Projects them, declares them, and adds events_upload_failed/manifest_upload_failed to the same bounded object (previously visible only as a console.warn in the Worker log). Both are written only when true, so a later run that publishes both clears the claim. No new columns (ADR 0034). Tested end to end: real callback route into a real D1, out through the real detail route, parsed by the shipped contract.
  • 898f7de7 — B3. shared/facets.ts now lists hasZarrVerified among the bespoke filters it excludes.
  • d611d236 Compile the published zarr schemas in CI — B2 (with X1).
  • 633ff68f Run the converter suite CI actually skips — B1 (with X2, X4). Adds the missing shared/** paths filter: a PR touching only shared/** matched no filter and ran zero jobs.

Test coverage (X)

  • e22bcca0 — X3. The sweep_stamps JSON paths move into one module the sweep and its three readers import: json_extract returns NULL for a path that matches nothing, so a typo read as "never swept" forever with no error anywhere. A seam test now runs the real sweep and then both readers' real SQL.
  • d611d236 — X1. shared/zarr-index.schema.json and zarr-manifest.schema.json are what the converter validates every document against before upload, and nothing on the TypeScript side ever compiled them, so a broken schema document failed at the next Hallu run with every TS test green. An ajv test compiles both, validates checked-in v3 fixtures, and proves the schemas discriminate by mutating one field each. The consumer fixture in backend/test/zarr-index-v3.test.ts now reads the same file: the inline one carried v1's source_key and neither layout nor updated_utc, so every consumer assertion about "a v3 index" was made against a document the producer would have refused to publish.
  • 633ff68f — X2, X4. zarr-python-test installs only jsonschema and pyarrow, so five classes (43 tests covering real-recording conversion, events.parquet publication, convert_one and the sample-index rate cases) raised SkipTest on every CI run and were reported as "skipped=9" in a green job. patch-duration-test already installs the real stack, so the full suite runs there now, under pytest, with a tripwire that fails the job above two skips (the real-data download and the bash-3.2 test). A tripwire test asserts every paths filter is exported as a job output and that shared gates the jobs that compile it. The engine-bump ack file is now covered in bash: armed for exactly one run, consumed before reconcile, and the pending-ack notice re-raised as its own error line.
  • 898f7de7 — X5. Pins the redirect-before-cache ordering: a chunk already primed by an allowlisted Origin must still 302 for a library request. Nothing else in that file fails if the order inverts.

Editor diagnostics (P)

  • 3bb48710 — P1 to P4. Typed containers where the values really are TypedDicts (and a comment on fails, where they deliberately are not, since it mixes _failure_entry results with entries read verbatim out of a published index), a str head_commit, a documented _row_group return, and two unused bindings removed (_soft, and stat_prefix's ignored bucket parameter).

Round 5 (final comment and doc corrections)

Three commits, same gates, all green.

  • 6b6619fd State the real sample_index formula — M1. Both sample_index_for's docstring and AGENTS.md said round(onset_s * rate); the code is math.floor(onset_s * rate + 0.5). They differ at an exact .5 tie, which is not exotic (any onset landing on a half sample at the serving rate hits one): Python's round() is banker's rounding and ties to even, this ties up. A client reimplementing the column from the docs would disagree with the published value on exactly those rows, which is the failure the column exists to prevent. The formula is now written out verbatim in both places, with the tie rule stated. The other round(onset hits in the repo (.context/draft-eegdash-zarr-issues.sh, and a variable literally named wrong in test_generate_zarr.py) describe what a CLIENT does wrong and are correct as written, so they are left alone.
  • 9525f3ba Log provenance_fetch_failed, fix four comments — M2, M3, M5, M6, M7. provenance_fetch_failed was validated and read by nothing, exactly like the two counts the earlier pass put on the summary log line: it says this wave's stores carry null doi/license/citation/hed_version because the CATALOG READ failed rather than because the data lacks them, and it has no column. It joins that line, with the summary-log test extended. c04f1200 restores the on008083 provenance the comment had lost: an earlier edit in this commit shortened "the on008083 shape (Per-dataset Zarr coverage: every raw recording accounted for in index.json, counts in D1, detail off-database #1197)" to "the pending-with-no-failures shape", which read as tidying a garbled token but in fact dropped the real dataset id behind Per-dataset Zarr coverage: every raw recording accounted for in index.json, counts in D1, detail off-database #1197 (43 raw recordings, 2 stores, 36 failures, five in no published list at all). It now reads "the pending-with-no-failures shape of on008083 (Per-dataset Zarr coverage: every raw recording accounted for in index.json, counts in D1, detail off-database #1197)", matching the two neighbouring citations by id in backend/src/services/s3.ts and backend/test/zarr-index-v3.test.ts. application/schema+json is registered by RFC 9358 (published, not a draft); the purge script's bare "PR dataset status/list don't authenticate, hiding private datasets #98" is qualified with the repo it was filed in (nemarDatasets/.github, before ADR 0029 repatriated the converter) and a truncated docstring sentence is finished. .context/decisions/README.md now says what an Amendment YYYY-MM-DD: block is for — a clarification that leaves the verdict standing — and that anything changing the verdict supersedes instead.
  • M4 needed no change: 590d2cb5 already rewrote that sentence, and it was re-verified line by line against hallu-zarr.sh — _is_prod_api_host for API_BASE/TEST_API_URL (lowercased, scheme/path/query/port/trailing-dot stripped), _normalize_guard_path for STATE_DIR/WORK_DIR (slash-collapsed and realpath'd, NOT case-folded), and plain == for S3_BUCKET/AWS_PROFILE.

Round 6 (delta re-review residuals)

Two commits. The delta re-review verified every must-fix and called the branch mergeable; these are the six residuals it listed.

  • c93adc50 Type the arrow rows, catch the listing failure — P1, P2, P3 and the S3 residual. merge_index's failures parameter becomes list[FailureEntry] | None; converted, pending and prior_pending deliberately stay list[dict] with a comment saying why (two are inputs this function does not validate, and prior_pending comes verbatim out of a published index). PriorEventRows._row_group returned object and was then indexed and .filter()ed, so both were unchecked accesses; it is now typed with an ArrowTable Protocol declaring the three members actually used — a Protocol rather than a TYPE_CHECKING import because pyarrow is an optional lazy dependency, so there is no module-scope name to annotate with and an import would make a checker's answer depend on whether the package happens to be installed. purge --all no longer hands an operator a traceback when the listing fails: list_dataset_ids raising is right (a failed listing must not read as an empty bucket), but letting it out of main was not, since exit 1 is what a per-dataset purge failure already uses. It now prints [purge] ERROR: --all could not list s3://<bucket>/: <reason> and returns 2 like the empty case, with a test driving the stub aws failure injection and asserting the message, the cause, and the exit code. A comment describing an or head expression where the code is a ternary is corrected to explain what the extra and prior_commit term is actually doing.
  • 6e03e2e4 Correct two docs the earlier rounds left stale — residuals 1 and 6. .context/validated_workflows.md still called the engine-bump preview "Read-only, no network, no writes"; hallu-zarr.sh's own comment was corrected in round 1 and this was not. migrate_schema runs for every subcommand and its additive ALTER plus engine_version seed are idempotent, so a preview on a node that has never run the current queue schema writes those, exactly as the next reconcile would; nothing is enqueued, parked or reset. catalog.ts said the contract declares the object CLOSED, which stopped being true when B5 made it .passthrough(); the three-edit rule still holds for the reason that actually carries it — a key this projection never reads is never in the object to begin with — so the sentence now says that instead of resting on a strictness that is gone.

M1 was already done in round 5 and needed no further change.

Decisions worth a reviewer's attention

The index publish is now a conditional write, and that has a CLI requirement. The PUT carries --if-match / --if-none-match, so the Hallu node's aws CLI must support conditional writes on put-object (S3 gained them in November 2024). purge_non_raw_stores.py already relied on this, but only on manual runs; it now gates every conversion. An unsupported CLI fails loudly (failed callback, exit 1) rather than silently, but the node's aws --version is worth checking before this reaches production.

An index upload failure now returns 1 instead of raising. It used to escape main() as an uncaught RuntimeError, so no callback file was written, hallu-zarr.sh POSTed nothing, the converting signal was never superseded, and D1 sat at zarr_status='pending' with nothing running. It now goes through write_failed_callback like every other exit that returns 1. test_a_clean_run_with_nothing_to_convert_publishes was rewritten to assert the new, stronger control: rc 1 with an error naming "index publish failed", not "refused".

read_index_with_etag and write_index moved into generate_zarr.py. purge_non_raw_stores.py imports them rather than keeping its own copies: both writers of this one object have to agree on how the ETag is read and sent back, or the condition protecting them from each other is decorative. write_index now returns the PUT's own ETag, which the converter uses instead of a follow-up head-object that could observe another writer's version; the purge ignores the return value.

The nested zarr_data_failures object got .passthrough(), not a closed shape. That matches the module doc's stated rule for every other object in shared/contract/dataset.ts ("the contract asserts REQUIRED fields and their types, a lower bound, so additive backend fields never break it"). Declaring the four fields is what fixes the stripping; passthrough only stops the next key from being invisible the same way. The contract tests assert both directions.

CI scope choice

The new shared filter gates lint, unit-pure and integration-dev. api-test was deliberately left off it: that job hits the live dev backend, and a shared-only change (a comment, a schema description) would otherwise burn live-backend quota on every such PR. Say the word if it should be included. patch-duration-test keeps its job name so the required-check configuration is untouched, even though its scope is now wider than the name suggests; a comment in the workflow says so.

Gates

  • cd backend && bun test: 1637 pass, 0 fail.
  • CI's pure tier, run locally with the exact file-selection loop from test.yml (255 files): 3876 pass, 20 skip, 0 fail.
  • bunx tsc --noEmit root and backend: clean. bun run lint: clean. bunx biome check on all 18 changed TS files: clean.
  • Converter suite, uv run ... pytest -q in scripts/zarr: 753 passed, 2 skipped (both Linux-only cases, skipped on the macOS dev box; they run in CI).
  • uvx ruff@0.15.4 check scripts/zarr: clean, with no new findings versus the epic branch (the pre-existing F811 is removed).
  • shellcheck --severity=warning scripts/zarr/hallu-zarr.sh: clean.

Known gap, not fixed here

The callback to D1 to detail-route round-trip test has to stringify id before datasetDetailSchema.parse(): the detail route serves SELECT d.*, so id arrives numeric while the contract declares a string. That is the systemic finding already filed as #1207 (no live response is parsed against these schemas today); the test names it in a comment rather than papering over it.

Every number on the zarr-ready callback body is a count, so the schema
primitive carries .int(): a fractional store_count validated cleanly and
was bound into an INTEGER column as a REAL. .catch(undefined) makes it
behave like any other malformed field -- dropped and logged, the rest of
the body still persisted.

Also retypes data_failures as unknown[] to match its validator (nothing
reads an element's fields, only the length), and puts not_attempted_count
and non_raw_dropped on the summary log line: both were validated and then
read by nothing, and neither has a column.
The index.json shapes in s3.ts and zarr-fidelity-sweep.ts are all-unknown
because the document is fetched from the serving bucket, written by a cron
on another host: the coverage equation is the producer's invariant
(check_index_invariant), not something these types assert. Says so, and
says where each file re-derives rather than trusts it.

Also tags zarr-data.ts's rateLimiter context cast with SAFETY: it widens
the static Variables map on the same runtime object, and rateLimiter reads
no Variable at all.
The comment called dev D1 a partial production mirror; the catalog was
purged to the curated fixtures. The allowlist survives that purge for the
reasons AGENTS.md gives: the users table was not purged, the dev worker
holds a real RESEND_API_KEY, and the nemarDatasets org is shared.
Two processes write <id>/zarr/index.json -- a converter run and
purge_non_raw_stores.py -- and both read it, work for minutes to hours,
then write back what they merged from that read. The converter's PUT was
unconditional, so whichever finished second silently reverted the other,
both exiting 0.

The PUT now carries --if-match with the ETag read at the top of the run
(--if-none-match '*' when there was no document). A 412 is re-read,
re-merged onto the newer document and retried once; a second 412 abandons
the publish through write_failed_callback so the driver posts a failed
callback rather than leaving D1 at converting forever. An upload failure
now takes that same path instead of escaping as an uncaught RuntimeError.

read_index_with_etag/write_index move into generate_zarr.py and the purge
imports them: the two writers of one object must agree on how the ETag is
read and sent back. The ETag now comes from the PUT's own response rather
than a follow-up head-object, which could observe another writer's version.

The stub aws in test_generate_zarr.py gains the real ETag semantics the
purge test stub already had.

Same file, same pass: PendingEntry.reason becomes the schema's closed
three-value enum (normalized through _pending_reason when read off a
published index, so a bad value cannot fail validation at the END of a
conversion), EventsPublication.file becomes a ManifestFileEntry mirroring
the manifest schema, and a duplicate import is dropped.
systems-inventory and validated_workflows contradicted each other on the
staging nightly cron; both now carry the installed crontab line verbatim,
including the explicit ZARR_DRIVER_REF override and the note that it flips
to dev once the epic merges.

Also corrects the --test guard-rail claim (only API_BASE and TEST_API_URL
are case-folded; the dirs are path-normalized and the bucket/profile are
exact equality) and hallu-zarr.sh's 'no writes' claim for
--preview-engine-bump, which connects and so runs migrate_schema.
The callback wrote pending/discovered into zarr_data_failures (#1197), the
catalog projection kept only count/detail_ref, and the contract stripped
what it did not declare -- so counts that existed in D1 reached nobody and
every layer's own tests passed.

Projects them, declares them, and adds events_upload_failed and
manifest_upload_failed to the same bounded object: an index published
without its sibling was previously visible only as a console.warn in the
Worker log. Both are written only when true, so a later run that publishes
both clears the claim. No new columns (ADR 0034).

Tested end to end -- real callback route into a real D1, out through the
real detail route, parsed by the shipped contract -- which is the only
shape that fails when one of the three layers is missed.
Candidacy clears only on a verdict, and the batch was ordered by
dataset_id, so ~25 alphabetically early datasets that error every run
(unreachable index, rotated credentials, their own fetch budget spent)
held the whole 25-row quota forever and nothing behind them was ever
swept -- while each run reported errors and looked like progress.

Every attempt now stamps zarr_verify_attempted_at, verdict or not, and the
batch is ordered by it (never-attempted first, then oldest attempt). An
unreachable dataset still records NO verdict and stays a candidate; it
just costs one slot per cycle instead of all of them.

The sweep_stamps JSON paths move into one module the sweep and its three
readers import: json_extract returns NULL for a path that matches nothing,
so a typo read as 'never swept' forever with no error anywhere. A seam
test now runs the real sweep and then both readers' real SQL.
_s3_child_prefixes returned [] for both 'no subdirectories' and 'the call
failed'. _rm_recursive is correct either way (it falls back to one
unsharded delete, and keeps that tolerance at the call site), but
list_dataset_ids read the same [] as 'this bucket holds no datasets' and
discover_excluded_stores as 'no stranded stores here' -- a clean empty
result for a run that could not see the bucket.

It now raises, and purge --all with zero datasets exits 2 like the
snapshot-directory check: the bucket holds ~800 prefixes, so an empty
listing is a misconfiguration, not an all-clean verdict.

Also adds the missing purge test for a non-412 write failure (classified
as an error, never as a conflict and never as a rewrite), and fixes the
editor diagnostics on the converter: typed containers where the values
really are TypedDicts, a str head_commit, and two unused bindings.
The fidelity sweep reported an unfetchable index as 'index.json is
absent'. S3 also answers 403 for a missing key when the caller cannot list
the bucket, so a rotated credential or a changed bucket policy produces
that same message for every candidate at once; the message now names both
readings so an incident is not read as a per-dataset converter bug.

zarr-data.ts's chunk-redirect access rows carry the byte count the Range
header ASKED for, recorded before the 302, so they are intent and not
delivery -- said so at the call site. shared/facets.ts now lists
hasZarrVerified among the bespoke filters it excludes.

Pins the redirect-before-cache ordering: a chunk already primed by an
allowlisted Origin must still 302 for a library request. Nothing else in
that file fails if the order inverts.
shared/zarr-index.schema.json and zarr-manifest.schema.json are what the
converter validates every document against before upload, and nothing on
the TypeScript side ever compiled them: a broken schema DOCUMENT (a typo'd
keyword, a dangling $ref) failed at the next Hallu run with every TS test
green.

Adds an ajv test that compiles both, validates checked-in v3 fixtures, and
proves the schemas discriminate by mutating one field each. The consumer
fixture in backend/test/zarr-index-v3.test.ts now reads the same file: the
inline one carried v1's source_key and neither layout nor updated_utc, so
every consumer assertion about 'a v3 index' was made against a document
the producer would have refused to publish.
zarr-python-test installs only jsonschema and pyarrow, so five classes --
43 tests covering real-recording conversion, events.parquet publication,
convert_one and the sample-index rate cases -- raised SkipTest on every CI
run and were reported as 'skipped=9' in a green job. patch-duration-test
already installs the real stack, so the full suite runs there now, under
pytest, with a tripwire that fails the job above two skips (the real-data
download and the bash-3.2 test).

Adds the missing shared/** paths filter: a PR touching only shared/** --
the wire contract, the facet vocabulary, the two JSON Schemas -- matched no
filter and ran zero jobs. A tripwire test now asserts every filter is
exported as an output and that shared gates the jobs that compile it.

Also covers the engine-bump ack file in bash: armed for exactly one run,
consumed before reconcile, and the pending-ack notice re-raised as its own
error line.
Both the docstring and AGENTS.md said round(onset_s * rate); the code is
math.floor(onset_s * rate + 0.5). They differ at an exact .5 tie, which is
not exotic (any onset landing on a half sample at the serving rate hits
it): Python's round() is banker's rounding and ties to even, this ties up.
A client reimplementing the column from the docs would disagree with the
published value on exactly those rows, which is the failure the column
exists to prevent. Written out verbatim in both places.
provenance_fetch_failed was validated and read by nothing, like the two
counts the earlier pass added: it says this wave's stores carry null
doi/license/citation/hed_version because the CATALOG READ failed, not
because the data lacks them, and it has no column. On the summary log line
now, with the test extended.

Comment corrections: the garbled 'on008083 shape' becomes
'pending-with-no-failures shape'; application/schema+json is registered by
RFC 9358, published, not a draft; the purge script's 'PR #98' is qualified
with the repo it was filed in (nemarDatasets/.github, before ADR 0029
repatriated the converter) and a truncated docstring sentence is finished.

The ADR README now says what an 'Amendment YYYY-MM-DD:' block is for -- a
clarification that leaves the verdict standing -- and that anything
changing the verdict supersedes instead.
@github-actions

github-actions Bot commented Sep 3, 2026 •

Copy link
Copy Markdown

📦 Package Preview

Your package is ready to test!

bun install -g nemar-cli@PR1208

Version: 0.9.12-PR1208.6e03e2e
Commit: 6e03e2e


🤖 Updated on each push.

on008083 is the real dataset behind #1197 -- 43 raw recordings, 2 stores,
36 failures, and five in no published list at all -- not a garbled token.
Dropping it cost the comment its provenance and left a shape named after
nothing. s3.ts and zarr-index-v3.test.ts both still cite it by id.
merge_index's failures param is list[FailureEntry]; converted, pending and
prior_pending stay list[dict] with a note, since two of them are inputs
this function does not validate and one is a worker-built store entry.

PriorEventRows._row_group returned object and was then indexed and
filtered, so both were unchecked accesses. Typed with an ArrowTable
Protocol declaring the three members used: pyarrow is an optional lazy
import, so there is no module-scope name to annotate with and a
TYPE_CHECKING import would still depend on it being installed.

purge --all no longer hands an operator a traceback when the listing
fails. list_dataset_ids raising is right; letting it out of main was not.
Same message shape and exit 2 as the empty case, with a test driving the
stub aws failure injection.

Also corrects a comment that described an 'or head' expression where the
code is a ternary.
validated_workflows still called the engine-bump preview read-only with no
writes; hallu-zarr.sh's own comment was corrected in round 1 and this was
not. migrate_schema runs for every subcommand and its ALTER plus
engine_version seed are idempotent, so a preview on a node that has never
run the current schema writes those. Nothing is enqueued, parked or reset.

catalog.ts said the contract declares zarr_data_failures CLOSED; it is
.passthrough() since the B5 fix. The three-edit rule still holds -- a key
this projection never reads is never in the object to begin with -- so the
sentence now says that rather than resting on a strictness that is gone.
@neuromechanist
neuromechanist merged commit 50f92e7 into feature/issue-1181-epic-zarr-serving Sep 3, 2026
12 checks passed
@neuromechanist
neuromechanist deleted the feature/issue-1181-review-fixes branch September 3, 2026 00:38

This branch was previously deployed

1 inactive deployment
npm — 6e03e2e4 Deployed Sep 3, 2026 by neuromechanist via publish #2202
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.

1 participant