Skip to content

[v0.1.3][Stage 1][11] Build and baseline single-branch history storage growth #33

Closed

Description

@yifanxuaaa

Latest Phase 1 scope amendment (2026-09-04)

The user authorized permanent Phase 1 scheduling suppressions under the committed 15-second policy. This supersedes the original full-inventory execution wording below. Keep every case implementation and historical outcome for Phase 2 testing/optimization; suppression is not deletion or a passing result.

The original 130 new timed definitions remain intact. The initial 14 suppressed scenario IDs remove 42 performance samples and their 42 associated verifiers from required Phase 1 scheduling, leaving 116 active new cases / 348 performance samples, plus 25 capped inherited samples. All 29 standalone proof recipes remain required. Any additional single performance sample exceeding 15 seconds of cumulative declared product time permanently suppresses that exact scenario across all seeds/repetitions and future Phase 1 sources. Preparation, standalone proofs and external cleanup are exempt. Still-active correctness, resource, custody and cleanup gates remain unchanged; #21 stays open.

Phase 1 completion amendment (2026-09-04)

The user requires functional failures to be repaired in Phase 1. The committed amendment supersedes earlier baseline-only closure and deferral wording below. Required functional, correctness, integrity, resource, safety-deadline and cleanup gates must pass before PHASE1_TERMINAL_PASS. Optional latency/storage optimization of correct operations remains Phase 2.

Preserve all original baseline failures and producing identities; record corrected candidate evidence separately. Fix shared causes and recollect only genuinely invalidated coverage. Do not shrink fixtures, omit cases, weaken oracles, raise budgets/timeouts or add benchmark-only bypasses. Keep affected family issues and #35 open until corrected execution and independent verification pass; #21 stays open for later phases/release.

Parent: #21
Depends on: #22 (shared infrastructure and committed specification).

Problem statement

LayerFS needs an authentic initial baseline for single-branch history storage growth before deciding product improvements.

Goal and family scope

Retain 1/10/100/500 Commits for distributed, hot-set, recurring, metadata-only and unrelated-rewrite trajectories on one fixed 1 MiB tree.

Inventory: 20 new timed cases, 60 prescribed initial performance sample slots, and 0 standalone proof recipes. The canonical planning file is docs/roadmap/0.1/0.1.3/dedup-branch-history.md. The snapshot below makes the current proposal reviewable until its committed revision is published.

Stage 1 work

  • Freeze this family and its exact fixture/schedule/oracle identities under the shared contract; implement its complete membership in the existing harness with one canonical definition and thin adapter.
  • Qualify route, call counts, size/resource bounds, independent oracles and selected-case self-checks. Resolve harness defects before collecting baseline evidence.
  • Run one selected case/seed to establish a valid execution path, then collect every prescribed initial sample and separate proof variant on the pinned unoptimized product. Required extended cases are explicit but remain required.
  • Retain raw failures/timeouts as classified product findings. Keep setup, performance, verification, cleanup and external command wall distinct.
  • Produce a source-bound family report with all outcomes and measured questions for later optimization; do not tune product algorithms within this issue.

Acceptance criteria

  • Complete family implementation, unique IDs, exact counts, deterministic fixtures/schedules and independent oracle checks.
  • Committed specification and product/harness/build/runtime/input identities linked in the evidence manifest.
  • Every required case/subcase executed with a retained initial outcome; no missing mandatory counter, unimplemented member or silent skipped slot.
  • All prescribed valid sample repetitions retained, with separate independent verification of distinct input/schedule variants. Failed/invalid outcomes remain explicit and ineligible for performance claims.
  • Preparation reuse, per-sample isolation, byte limits and cleanup checked; actual first-use/cache-hit command wall reported.
  • Family baseline summary delivered to the consolidated report issue; known product failures remain failing release gates and are recorded for Stage 2.

Files to read

docs/roadmap/0.1/0.1.3/dedup-branch-history.md, docs/roadmap/0.1/0.1.3/testing-rules.md, docs/general/benchmark_rules.md, and the existing source/test references in the specification below.

Stage 1 boundary

Build the benchmark and retain its initial outcomes. Product algorithm, storage-format and performance optimization are deferred to Stage 2. Improve benchmark preparation/reuse and correct harness defects now; do not silently change the measured product to make results pass. Pin the unoptimized product identity and distinguish harness changes from product changes.

A slow result, product failure or timeout is an initial finding, not permission to lower a gate, drop a case or start tuning. Keep outcome, harness validity, verification, resources, cleanup and coverage completeness separate. Failed/invalid cases cannot support performance claims. Unimplemented cases, missing mandatory observations and unexecuted slots remain incomplete.

Shared execution rules

  • Reuse fs-bench-pro, existing generators, runner helpers, custody, qualified pristine inputs, independent writable samples and reports. Add only needed family adapters.
  • Use 1/10/100/500 of the declared unit; no Cartesian size/count matrix. Each workload file is at most 500 MiB and aggregate logical content is strictly below 1 GiB at every intermediate state, including temporary files and Git internals. Physical Store/spool/cache budgets are separately reported.
  • Resolve the selected case/seed before preparation. Reuse compatible artifacts; prepare only required inputs. Measured creation/import/history must still execute all work.
  • Iterate one selected case and seed; ordinary warm-prepared 1–5 second command wall is a target to qualify, not a blanket guarantee. Required extended cases are explicitly selected and cannot be omitted from complete baseline coverage.
  • Separate performance and verification. Added benchmark hashes, full manifests, reopen and fault injection stay outside performance timing; intrinsic CAS/Git hashing remains measured.
  • Collect several metrics from one execution and regenerate reports from raw evidence without rerunning successful workloads. Retain all valid results and failures, with exact source, fixture, oracle, runtime and cache identities.
Planning specification snapshot — dedup-branch-history.md

Source file: docs/roadmap/0.1/0.1.3/dedup-branch-history.md. Local draft SHA-256: b2569c2a4ed165ca32be6f76b066c90e426f9254c1e4db2b7bbbdf345dc718ca. This snapshot is reviewable now; it is not a committed admission contract. The infrastructure issue must publish/freeze the source revision before implementation and collection.

Deduplication across retained history on one Branch

Status: Draft family dedup_branch_history: 20 new timed cases and no
proof-only cases. Depth 500 belongs to extended qualification; it is not a
promise of a 1–5 second execution.

Question and scope

Does retaining many exact Workspace states cost only their unique content and
required metadata? Measure accumulated storage, per-Commit latency/resources,
and exact historical readability for localized, repeated, metadata-only, and
unrelated whole-Workspace changes.

This family explicitly permits one Branch with repeated successful Commits
in v0.1.3. It supersedes the old single-Commit draft restriction for these
cases only. Branch fan-out, competing writers, Add, conflicts, and multi-Branch
deduplication remain v0.1.4. The shared testing rules (docs/roadmap/0.1/0.1.3/testing-rules.md; planning file awaiting publication) and
parent inventory (docs/roadmap/0.1/0.1.3/README.md; planning file awaiting publication) own common requirements and release boundaries.

Fixed Workspace and exact membership

Use exactly one deterministic 1 MiB, 200-file Workspace shard from the
shared fixture (docs/roadmap/0.1/0.1.3/testing-rules.md#shared-workspace-fixture; planning file awaiting publication), including its wide
directory, regular directory, deep spine, paths, generator, and metadata:

128 files × 1 KiB + 64 files × 8 KiB + 8 files × 48 KiB = 1 MiB

Files have unique independently generated bytes and frozen paths/metadata.
Qualify chunk intersections rather than assume small files contain no repeats.
Use one Store, LayerStack, genesis Layer, Branch, and real-FUSE Workspace for
the complete history. N is the number of successful retained Created
Commits
, with tiers 1, 10, 100, 500. Each row expands to four scenario IDs
using those suffixes.

Scenario prefix Work before each Commit Public surface
dedup-history-distributed One 64-byte same-length replacement, cycling a seeded permutation of all 200 files; replacement changes on later visits One singular SDK file-range edit
dedup-history-hotset One 64-byte replacement cycling the fixed eight 48 KiB files; later visits receive new bytes One singular SDK file-range edit
dedup-history-recurring Replace one fixed 48 KiB file in full, alternating exact deterministic B and its original A One singular SDK full-range replacement
dedup-history-metadata Toggle one file's mode between 0600 and 0640; preserve payload and all other metadata Ordinary FUSE chmod in one completed public Exec
dedup-history-unrelated Rewrite all 200 files with unrelated deterministic bytes of their original lengths, close/sync, then Commit One public Exec running one compiled ordinary filesystem workload

There are 20 timed IDs and 60 timed samples. A fresh tier sample executes
its entire N-Commit history; its schedule is an exact prefix of the same
500-Commit seed/profile schedule, not a resumed Store from a smaller timed
tier. Generated replacements always differ from the current bytes; mode
toggles differ from the current mode. The SDK batch API is same-file only,
so it cannot replace the stated multi-file call count.

Use zero-based Commit ordinal k. Distributed edits choose the k mod 200
member of a seed-ranked permutation of shard file ordinals; hotset edits choose
the k mod 8 member of the seed-ranked ordinals 192..199. Both replace the
64 bytes centered at file_length / 2, using distinct replacement bytes
generated from family/profile/seed/k before the operation timer. Recurring
targets file ordinal 192: even k installs B and odd k restores A, using
WorkspaceFileReplacement::Inline for the complete 48 KiB range. Metadata
targets ordinal 0: even k sets 0600 and odd k restores 0640. Unrelated rewrites
generate all 200 replacement files from profile/seed/k/file ordinal. Bind the
permutations, exact replacement hashes, modes, and manifests to the source
seal; no current result or clock chooses the next operation.

The unrelated profile is the necessary negative control for storing new
complete snapshots, not a latency peer for a single 64-byte edit. At depth
500 it performs 500 public Exec calls and 500 Commits, rewriting 100,000
files and 500 MiB in total. This preregistered POSIX/FUSE tool operation is not
an emulation or timing of an SDK edit. Compare each profile only against the
same declared profile/surface. Its actual work and wall belong in the extended
lane; do not substitute one edited file, skip intermediate publications, or
claim a few-second full history.

Byte ceilings and retention denominators

Every visible Workspace state is exactly 1 MiB; the largest file is 48 KiB.
Count genesis plus N retained states, including repeated content roots as
distinct committed snapshots with distinct Commit identities:

retained logical snapshot bytes = (N + 1) × 1,048,576
maximum at N=500 = 525,336,576 bytes = 501 MiB

No-op attempts, failed attempts, verifier reopenings, and UpToDate results
do not add a snapshot or increase the denominator. Report attempted, Created,
and UpToDate counts separately and fail membership if fewer than N Created
Commits result. Do not deduplicate repeated snapshot identities out of the
logical denominator, or include them more than once.

Stream replacement generation and historical verification; at most one bounded
replacement/verification buffer is live, and at most one historical projection
is mounted at a time. Do not materialize all historical snapshots as ordinary
files alongside the live Workspace. Retained logical snapshots, current files,
temporary logical artifacts each receive explicit ledgers under the shared
<1 GiB total and <=500 MiB per-file rules. Actual Store/spool allocation has
its separate physical disk budget, as the shared rules require. This fixture
leaves headroom even when a current state is counted separately.

Operation and timing

Prepare and seal the genesis outside the history timer, using the existing
fixture/custody infrastructure. On the same Workspace, execute the declared
mutation, wait for any Exec to terminate, and call
Client::commit_workspace_session_with_status. Require Created; record
Commit ID, parent, root, head, call counts, phase durations, and available
product counters after acknowledgement. The next mutation follows that
acknowledgement. Do not Add the Commit as another Layer.

The distributed, hotset, and recurring profiles use exactly the stated public
SDK content-edit route, with an attached real FUSE projection and no
edit-caused FUSE payload writes. Metadata and unrelated whole-tree rewriting
use their declared native operations and are not called SDK range edits.
The unrelated process independently regenerates bytes, opens each existing
file, writes its complete declared new contents, closes it, restores the frozen
mtime, and performs its declared sync before exit. No per-file process launch
or cached completed snapshot replaces this work. Active processes or pinned
handles cannot be silently bypassed to make owner-side edits work.

Record per-edit aggregate, per-Commit, history-construction, Workspace Create,
and End walls separately. Do not put transcript construction, historical
queries, complete Store scans, digesting, resource sampling not permitted by
the shared contract, reconnect, or FUSE verification inside operation timers.
Cheap acknowledgements/counters must obey the declared collection boundary.

Payload, metadata, and physical accounting

Use the exact transcript model (docs/roadmap/0.1/0.1.3/dedup-cross-file.md#exact-payload-and-storage-accounting; planning file awaiting publication)
for every retained regular-file root. Maintain distinct sets for genesis,
previously retained history before each Commit, the current state, and the
union of all retained states. The union includes chunks no longer referenced
by the latest state but still required by an older Commit.

Report for every step and cumulative history:

  • attempted/created Commit counts, roots, changed file counts and bytes;
  • regular-file payload occurrences, referenced extent bytes/source offsets,
    distinct chunk payload and canonical bytes, newly inserted bytes, emitted
    preexisting CAS reuse, and borrowed unchanged extents;
  • distinct namespace/metadata canonical objects and actual inserted/reused
    bytes, even when file payload is unchanged;
  • retained logical snapshot bytes versus distinct retained payload bytes,
    current-state logical bytes, and unique bytes introduced by this step; and
  • actual SQLite length/allocated blocks/live pages, canonical bytes by role,
    transaction maxima, spool and temporary peaks, CPU, RSS/cgroup memory,
    swaps/OOM, I/O, and execution/output/lease retention.

Savings against retained logical snapshots is an honest history-sharing
measure, not proof that the same quantity was newly emitted and CAS-deduplicated.
Borrowed unchanged extents and emitted CAS hits remain separate. Actual Store
growth also includes Commit records and metadata; monitor saved_fraction and
physical_bytes alone cannot establish retained-history disk cost.

For metadata-only histories require zero new regular-file payload chunks,
not zero total objects or zero database growth. For unrelated rewrites require
the sealed new payload set each step and preserve every older set. For local
edits retain the exact oracle-defined extents and chunk sets rather than
requiring a fresh whole-file CDC transcript that the incremental representation
does not promise.

For recurring A/B, full-file replacement makes the input content relationship
explicit. Seal the canonical transcript expected for that operation before
admission and compare each recurrence with the earlier state. A zero-new-chunk
gate is justified only when the expected transcript is the earlier transcript.
Do not generalize it to arbitrary SDK partial reversion: equal visible bytes
can be represented by different valid borrowed extents, and unchanged extent
borrowing is not an emitted CAS hit. Exact byte reappearance and retention are
mandatory regardless of that distinction.

Historical verification and completion

In a separate verification run, reconstruct the deterministic schedule and
verify every retained Commit, not only earliest/middle/latest examples.
After fresh Store reconnect, enumerate exact Commit identities using
Client::query(Query::new(QueryKind::Commits)) scoped to the LayerStack,
including continuation pages, and check parent/head/root topology. Reuse the
existing read-only canonical verifier to authenticate each root's complete
file-payload/metadata graph and stream all expected bytes. Query returns Commit
records, not a new historical file-read API.

For real-FUSE historical readback, use the existing public
Client::fork_branch(name, LocalForkSource::Branch { branch_id, commit_id })
route and create/end one
historical Workspace at a time. Verification-only Branches/projections are not
timed history members or a Branch fan-out claim; their allocation and readback
walls stay in verification receipts. Freeze the original one-Branch storage
measurements before these verifier-created Branches, never include them in
history-sharing denominators, and clean up their runtime resources. Do not add
a product API solely for the verifier.

The final state, all historical states, recurrent A/B bytes, untouched paths,
canonical roots, metadata, and released cleanup behavior must be exact. No
GC or compaction may remove history to improve this family's growth metric.
No additional persistent index, sidecar, format, or benchmark-only API is
authorized.

Run three fresh seed samples per ID. Reuse pristine genesis preparation and
existing family selection, source sealing, result reports, and SDK runner
adapters; extend their operation-count schema for this declared repeated-Commit
family rather than weakening inherited one-edit admission. Small selected
histories may target the shared fast loop. Freeze realistic extended budgets
for depth 500, retain all valid samples, and report baseline limitations without
silently truncating the retained history.

Source references

Activity

  1. self-assigned this
    on Sep 4, 2026
  2. yifanxuaaa commented on Sep 4, 2026

    @yifanxuaaa
    ContributorAuthor

    Performance-results checkpoint for #33: 51 independently validated active VM8 performance samples. The report validates 0 retained full proof(s) for this family at this checkpoint; further full checks are running. This comment does not claim global Phase1 terminal pass.

    Published report, exact rows, source identities and verification status, phase/counter evidence.

    VM8 performance source: 7948df2de269e5ffd47a232ffd8091ff83f8869f; fixed 8-CPU Docker VM, unchanged 2-CPU/2-GiB benchmark-container caps. No VM4/VM8 timing pooling. Every listed performance row passed the independent authenticity, custody, observation, resource and cleanup validation; workload admission still requires its separate full proof.

    Values below are medians across the prescribed three seeds. Product time is the declared cumulative public-operation time; preparation and complete sample wall are separate. Store growth is the observed physical Store-file delta, not a deduplication ratio or verified canonical byte count.

    Case n Product ms Preparation s Complete sample s Store delta MiB Full proofs in checkpoint
    dedup-history-distributed-1 3 18.987 3.811 4.218 0.000 0/3
    dedup-history-distributed-10 3 57.229 0.512 1.102 0.062 0/3
    dedup-history-distributed-100 3 1421.546 0.515 4.048 0.750 0/3
    dedup-history-hotset-1 3 18.270 0.483 0.891 0.000 0/3
    dedup-history-hotset-10 3 198.303 0.519 1.247 0.062 0/3
    dedup-history-hotset-100 3 2331.693 0.502 4.854 0.812 0/3
    dedup-history-hotset-500 3 12519.133 0.494 21.460 3.688 0/3
    dedup-history-metadata-1 3 25.448 0.480 0.903 0.000 0/3
    dedup-history-metadata-10 3 86.168 0.486 1.085 0.000 0/3
    dedup-history-metadata-100 3 716.550 0.490 3.121 0.000 0/3
    dedup-history-metadata-500 3 3783.155 0.476 12.483 0.125 0/3
    dedup-history-recurring-1 3 18.588 0.499 0.888 0.062 0/3
    dedup-history-recurring-10 3 52.336 0.511 1.094 0.062 0/3
    dedup-history-recurring-100 3 412.714 0.478 2.831 0.062 0/3
    dedup-history-recurring-500 3 2518.989 0.486 11.316 0.188 0/3
    dedup-history-unrelated-1 3 977.117 0.480 1.877 1.375 0/3
    dedup-history-unrelated-10 3 7527.608 0.489 8.637 12.875 0/3

    Explicit Phase1 suppressions (not passing coverage): dedup-history-distributed-500, dedup-history-unrelated-100, dedup-history-unrelated-500. Definitions, historical failures and incomplete attempts remain retained for Phase2.

    All prior actual failures, source incompatibilities, invalidations and preparation interruptions remain in the linked report and raw evidence. The fast development verifier contributes zero full-verification credit. Completed family tickets will close only after their remaining mandatory full checks pass; #21 remains open for later phases.

  3. yifanxuaaa commented on Sep 4, 2026

    @yifanxuaaa
    ContributorAuthor

    The user has amended Phase 1 acceptance: qualified fast verification is sufficient for routine benchmark verification. The published amendment supersedes earlier comments requiring every routine exhaustive proof before closure. The reference qualification contract preserves independent expected states, affected-content/metadata/alias/absence checks, authenticated references and deterministic witnesses.

    The posted performance results remain valid at their original source/environment identities. Completed full proofs remain accepted and are reused. At the scope checkpoint,48 routine slots already had full proofs; the305 unrun routine exhaustive counterparts are now deferred to Phase2 while their qualified fast checks are implemented and collected. Fast evidence keeps its actual assurance and explicit skipped coverage; it will not be labelled fully_verified.

    Required targeted reliability checks, the CDC boundary proof, exact expected-error oracles, resource and cleanup gates still apply. Observed failures still require focused repair confirmation. The already-qualified600-second proof is retained rather than repeated. This amendment does not manufacture missing evidence or itself establish terminal pass.

    Remaining routine families will close after their accepted fast/full coverage and applicable gates pass. #21 remains open for later phases and release.

  4. yifanxuaaa commented on Sep 4, 2026

    @yifanxuaaa
    ContributorAuthor

    Read-only pin repair performance results for #33 — 12 corrected qualified PASS; 0 corrected non-PASS outcome(s). The corrected passing samples have completed independent validation with zero issues or resource violations. It does not claim terminal pass.

    Corrected runtime source: 30d13deeec72b46ff7bc411f1ec08a46990541e1. Other rows retain their original qualified checkpoint at 7948df2de269e5ffd47a232ffd8091ff83f8869f with their original numbers and source. Exactly 45 historical slots across #23/#25/#29/#33 were affected. The correction produced 42 independently qualified PASS slots; directory-content-scan-100 seed 1 exceeded the cumulative 15-second limit and its seeds 2/3 were not run. That case is now suppressed for all Phase 1 seeds/sources: 15 disabled case IDs, 370 active performance slots (328 retained plus 42 corrected). No old/new timings are pooled.

    The changed product branch now acknowledges the existing read-only regular-file Pin before exposing the Open handle. Historical affected rows remain actual observations of their old source; they are superseded for the repaired candidate, not deleted or called failed. No speedup claim is made.

    Product ms, preparation s, complete sample s and physical Store-file delta MiB use the prior table definitions. Corrected medians use only the listed completed qualified PASS seeds; incomplete three-seed groups are explicitly partial. They are medians of independently qualified samples. The full-proof column is copied from the prior checkpoint and is not a refreshed current proof count. Current routine acceptance follows the authorized fast-verification amendment; a fast result remains fully_verified=false. Targeted reliability gates are unchanged.

    Case n Product ms Preparation s Complete sample s Store delta MiB Full proofs in prior checkpoint Source Disposition
    dedup-history-distributed-1 3 18.987 3.811 4.218 0.000 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-distributed-10 3 57.229 0.512 1.102 0.062 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-distributed-100 3 1421.546 0.515 4.048 0.750 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-hotset-1 3 18.270 0.483 0.891 0.000 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-hotset-10 3 198.303 0.519 1.247 0.062 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-hotset-100 3 2331.693 0.502 4.854 0.812 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-hotset-500 3 12519.133 0.494 21.460 3.688 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-metadata-1 3 23.566 2.353 2.769 0.000 0/3 30d13dee Qualified PASS
    dedup-history-metadata-10 3 90.263 2.389 2.969 0.000 0/3 30d13dee Qualified PASS
    dedup-history-metadata-100 3 755.511 2.376 5.129 0.000 0/3 30d13dee Qualified PASS
    dedup-history-metadata-500 3 3768.286 2.368 14.415 0.125 0/3 30d13dee Qualified PASS
    dedup-history-recurring-1 3 18.588 0.499 0.888 0.062 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-recurring-10 3 52.336 0.511 1.094 0.062 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-recurring-100 3 412.714 0.478 2.831 0.062 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-recurring-500 3 2518.989 0.486 11.316 0.188 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-unrelated-1 3 977.117 0.480 1.877 1.375 0/3 7948df2d Retained prior qualified checkpoint
    dedup-history-unrelated-10 3 7527.608 0.489 8.637 12.875 0/3 7948df2d Retained prior qualified checkpoint

    Historical affected rows (preserved exactly from the prior qualified table; excluded from corrected medians):

    Case n Product ms Preparation s Complete sample s Store delta MiB Full proofs in prior checkpoint
    dedup-history-metadata-1 3 25.448 0.480 0.903 0.000 0/3
    dedup-history-metadata-10 3 86.168 0.486 1.085 0.000 0/3
    dedup-history-metadata-100 3 716.550 0.490 3.121 0.000 0/3
    dedup-history-metadata-500 3 3783.155 0.476 12.483 0.125 0/3

    Exact corrected outcomes and historical supersession mapping:

    • dedup-history-metadata-1 seed 1: dedup-history-metadata-1-s1-performance-bad704e14934 (independently qualified PASS), replaces timing selection dedup-history-metadata-1-s1-performance-8b3f4c1dd518.
    • dedup-history-metadata-1 seed 2: dedup-history-metadata-1-s2-performance-645312188d78 (independently qualified PASS), replaces timing selection dedup-history-metadata-1-s2-performance-e8b8acd68a56.
    • dedup-history-metadata-1 seed 3: dedup-history-metadata-1-s3-performance-7ed6a26cfac9 (independently qualified PASS), replaces timing selection dedup-history-metadata-1-s3-performance-567c675dae2b.
    • dedup-history-metadata-10 seed 1: dedup-history-metadata-10-s1-performance-3e4ca62c047b (independently qualified PASS), replaces timing selection dedup-history-metadata-10-s1-performance-828034eb7b80.
    • dedup-history-metadata-10 seed 2: dedup-history-metadata-10-s2-performance-093ca9b60f42 (independently qualified PASS), replaces timing selection dedup-history-metadata-10-s2-performance-fcd6233c40db.
    • dedup-history-metadata-10 seed 3: dedup-history-metadata-10-s3-performance-a66af497bc30 (independently qualified PASS), replaces timing selection dedup-history-metadata-10-s3-performance-6f1adc458fe2.
    • dedup-history-metadata-100 seed 1: dedup-history-metadata-100-s1-performance-f6a10321ab57 (independently qualified PASS), replaces timing selection dedup-history-metadata-100-s1-performance-a2b3266c8d1f.
    • dedup-history-metadata-100 seed 2: dedup-history-metadata-100-s2-performance-b4b756714849 (independently qualified PASS), replaces timing selection dedup-history-metadata-100-s2-performance-ba78cc981d2a.
    • dedup-history-metadata-100 seed 3: dedup-history-metadata-100-s3-performance-6f1b91c2b5e3 (independently qualified PASS), replaces timing selection dedup-history-metadata-100-s3-performance-d69da3573573.
    • dedup-history-metadata-500 seed 1: dedup-history-metadata-500-s1-performance-84ab78bc55d5 (independently qualified PASS), replaces timing selection dedup-history-metadata-500-s1-performance-8c78da5492ad.
    • dedup-history-metadata-500 seed 2: dedup-history-metadata-500-s2-performance-e3ed97d6a3ee (independently qualified PASS), replaces timing selection dedup-history-metadata-500-s2-performance-02860d45823b.
    • dedup-history-metadata-500 seed 3: dedup-history-metadata-500-s3-performance-36a0de9e48ed (independently qualified PASS), replaces timing selection dedup-history-metadata-500-s3-performance-27c521702cd0.

    All attempt paths are under benchmark-results/fs-bench-pro/phase1-v013/attempts/. Source, custody, observations, resources and cleanup were independently checked; matching routine correctness status is reported separately. Original failures and suppressions remain preserved; central #21 stays open.

    Published corrected timing checkpoint and per-case medians.

  5. yifanxuaaa commented on Sep 4, 2026

    @yifanxuaaa
    ContributorAuthor

    Phase 1 is closed under the user's explicit withdrawal of all remaining verification tests. This is a scope withdrawal, NOT a claim that the original verification terminal criteria passed.

    The assistant designed an excessively expensive verification suite and spent too long expanding it after the user requested a much faster finish. That design is withdrawn and must be reassessed in Phase 2. No more verification work is scheduled.

    Final performance results:370 independently qualified active observations, with actual source identities and unoptimized timings. All15 suppressed case definitions, original failures and incomplete attempts remain preserved. Completed29 targeted proofs,48 routine full proofs and qualified fast evidence remain historical evidence; unrun verification is withdrawn/deferred, never relabeled passing.

    Explicit withdrawal and accountability record. Separate Phase 2 backlog. Central#21 stays open for Phase 2/release. No release tag was created.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

new featureA new user-facing capabilityv0.1.3LayerFS v0.1.3 checkpoint and delivery

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions