Repository navigation
NaN-any: a NaN matches any NaN in the bit-exact corpus comparison (PM ruling) - #147
Merged
Merged
Conversation
…nes-host on xcode-27), regen --skip/--only, ci.yml workflow_dispatch
…ery Chrome round (it is a sink, ~50 min on 3 cores)
…he result of the job they need)
…hes are invisible to other branches); never pushes there
…ilies, animation kinds and derived paint unions Each feature owns its own file and one sorted line in the central list: diagnostics/codes/<feature>.ts (codes with their catalogue entries, pinned per feature in test/diagnostic-codes/<feature>.json after the frozen legacy list), faults/<feature>.ts, css/animation-kinds/<family>.ts, and one family(...) line per property family in properties.ts. The paint write and plant unions are derived from the registered modules, so a paint module edits only its own files. Values are unchanged (same codes, order, catalogue, faults, longhand tables and kinds); registry-claims.test fails when two features claim one name.
…ion, sorted .gitattributes fixtures.ts registers each group on one line of GROUPS, sorted by id, and runs the legacy groups in their frozen order, then every later group by id. scripts/regen.ts keeps the legacy steps and MANUAL entries and gathers each feature's steps (placed after the step they name), extra outputs and MANUAL entries from scripts/regen-steps/<feature>.ts, one sorted line per feature. .gitattributes is sorted (same patterns, same attribute). STEPS, MANUAL, FIXTURE_GROUPS and FIXTURES are unchanged; registry-claims.test fails when two features claim one group, fixture id, step name, output or command.
…o features fails at load
…own files (entry.ts, diagnostics/codes/<feature>.ts)
…itattributes keeps master's animator line in sorted order
…lap, and registry-claims checks no tracked file has two writers; fixtures.ts documents the run order of later groups
…ide it The parallel scheduler took a whole-tree git add -A snapshot when one step finished, while another was deleting and rewriting its outputs, and aborted on 'unable to stat'. Snapshots now exclude (git exclude pathspecs, same glob semantics) the declared outputs of every step still running; those paths were already left out of change detection, so the stray-write guard is unchanged.
…ode (8x faster compiles; same digests)
…ct on failure, per-PR merges in order
…ly on its parent's position
…ules, so ua:capture's regen inputs leave out the compiler
…umented node:crypto fast path
…ault) planted-swift.test.ts and planted-kotlin.test.ts compiled and ran a planted harness for each of the 7 FAULTS in one test (planted-swift took 734 s on the CI runner, planted-kotlin 329 s). Each (target, fault) pair is now its own test file (planted.ts holds the shared test), so vitest schedules them across workers and shards. planted-union.test.ts proves the files cover FAULTS on both targets exactly once.
The determinism describe of parity.test.ts (356 tests, 290 s of its 548 s on the CI runner) runs from parity-determinism-<k>.test.ts, one chunk of FIXTURES per file under the same describe titles, each with its own Chrome and its own coverage check. parity.test.ts keeps a coverage test proving the chunk files together check every fixture once.
…ish; non-test proof errors stop the driver; clean other trees' ignored outputs before a proof
Commands: pnpm regen
… reruns failing files alone; skip the master proof for a tree this run proved
… with any NaN; corpus inputs, results and digests write every NaN as 7ff8000000000000 (x86-64 makes some fff8...)
regen: a snapshot never reads the outputs of a step still running beside it
Commands: pnpm regen
land: batched landing (one proof per batch, prefix bisect, per-PR merges in order)
… SIGTERM kills the running step through a supervisor and records INTERRUPTED, failing no PR
…a master left unproved; the lock records the driver and kills an orphan; cleanup after any abnormal death, including the run's devices
Contributor
Author
|
Verified at 6d78ce6 (this head): 🤖 Generated with Claude Code |
The quiet gate required zero other heavy-slot holders and load under 20. A near-idle holder (a profiling run) kept the gate shut at load ~1 while a batch's single failing file waited to be rerun. Below load 6 the machine counts as quiet regardless of holders; the quiet request still stops new jobs from starting. land.test pins both sides.
digest: SHA-256 through Node's native hash when available (8x faster compiles, same digests)
regen: ua:capture's inputs leave out the compiler (platform.ts imports the reference platforms from their own modules)
SLOW-SPLIT 1: planted translator fault tests, one file per (target, fault)
…publish, never returned without it (the landing driver's ignored-file cleanup leaves such entries)
…utputs
packages/translate/out/{kotlin,swift}/ are content-keyed caches valid for any tree. #155's clean removed their files but not
their directories, which left stale entries that blocked every later build of the key (fixed at the source in #164); keeping
them also saves the harness rebuilds on every proof. land.test adds both to the kept set.
…try root cause), where each target stopped
Commands: pnpm regen
Commands: pnpm regen
Commands: pnpm regen
Commands: pnpm regen
translate: a stale harness cache entry (artifact deleted, directory kept) is replaced, never returned; root cause of the land-147 'Unable to access jarfile' failures
land: keep the translate harness build caches when cleaning ignored outputs
land: an idle machine is quiet even with a heavy slot held
SLOW-SPLIT 2: parity determinism shards in 4 chunk files
Commands: pnpm regen
thejackshelton
added a commit
that referenced
this pull request
Oct 5, 2026
…c-only android vectors lane; merge takes the CI outcomes alone
This was referenced Oct 5, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What changed (PM ruling, 2026-10-04): in the native corpus comparison (
packages/translate/src/native.tsrunSuite), a NaN result now matches any NaN result. Nothing else is loosened:8000000000000000) is still distinguished from +0;canonicalNan(packages/translate/src/corpus.ts) writes every NaN bit pattern as7ff8000000000000. It is applied to the corpus inputs written for the native harnesses, to the corpus digests (inputs and TypeScript results) and to the comparison.sameResultis exact equality or equality aftercanonicalNan.Why: JavaScript gives NaN sign and payload bits no meaning, and hosts differ in them. x86-64 makes
0 * Infinityfff8000000000000; arm64 makes it7ff8000000000000. On the x86_64 emulator, the android layout-vectors-device lane failed 5 of 120,000units-m2cases with["ok","fff8000000000000"]against["ok","7ff8000000000000"]. The reason names the first two: #45946 and #46835, bothzoomFontSize(sizePx, zoom)=Math.fround(Math.fround(sizePx) * zoom)(packages/layout/src/units.ts:74) with NaN inputs. The other three are not printed, since the lane lists at most 2.The corpus no longer depends on the host:
fff8NaNs in its inputs and results: in units, 507 inputs and 118 results; in units-m2, 175 and 86; in units-calc, 25 and 253. Its digests differed from the committed manifests (P1dac4914…againstf4ff5014…, extended30681251…againstb2be6747…).7ff8000000000000, so canonicalizing changes nothing on the Mac. The committedcorpus.jsonandcorpus-dpr.jsondigests are unchanged, so there is no regen output.Next, once this lands: the hybrid in #132 can drop its Mac-only lane, so all device lanes run on CI.
Tests:
packages/translate/test/nan-canonical.test.ts(new) covers:pnpm typecheckand the test file are queued under the heavy lease. The Mac is at load 67+, and the results will follow in a comment.🤖 Generated with Claude Code