Repository navigation
PerfChecker 1.0.3 uses Core 1.1.1 and fixes explicit native test discovery after saved edits: Refresh invalidates the previous inventory, waits for the active scan to physically close, and starts one coalesced follow-up scan. Shutdown rejects queued discovery before controller teardown.
Known limitation: Pluto on macOS Intel
The macOS Intel / VS Code stable MCP + Pluto + discovery group did not qualify. During fresh installation, the first native Pluto notebook was still absent after the one-hour harness availability wait. The cause is unknown. Later configured/prepared checks do not turn that failed fresh-install case into a pass. This limitation is accepted for 1.0.3 and will be addressed in the next version.
Validation scope
Both general CI campaigns passed all four jobs. Native qualification has 34 physically verified PASS groups out of 35: all Linux 1.96/stable, Windows stable and Windows 1.96 groups, plus six macOS Intel groups. Windows 1.96 is non-blocking by release decision, and all seven of its groups passed. This is not a claim of 35/35 qualification. The additional Intel retry is not used as passing evidence; its cancellation was requested when publication was authorized without further CI.
The product passed local TypeScript compilation and 373 Node tests (345 passed, 28 skipped, zero failed/cancelled); the reviewed Linux preflight correction passed 47 targeted Node tests. The 45 fresh bootstrap cases were physically qualified. Native receipts retain phase-specific Core provenance: registered General 1.1.1 in fresh Controller and Pluto; pinned candidate d079efd5c24c3dcac5b7550781822038e0ff780f in configured/prepared, with the same Core source tree. External authentication, GPU rendering and arbitrary callback behavior remain outside the demonstrated scope.
Current download: H4, icon-only packaging correction
Marketplace rejected the original H3 icon because it was1254×1254 pixels. The same transparent logo is now1024×1024. H4 is a deterministic packaging derivation of H3: only extension/media/perfchecker.png changes; all49 other archive members, including the manifest and every compiled module, and their compressed data are byte-identical. No product rebuild, TypeScript compilation, test or CI was performed for this packaging correction.
The icon-only source commit is a91d2b7, tree788420ef60394923ff3377ef144bd293db21ef9a, on main after the release merge. The existing v1.0.3 tag remains unchanged at06b6d36281bd208270595c95ddbd2ea8f0657e30. The provenance explicitly records this after-tag packaging correction and its original H3 source/run.
| Current asset | Bytes | SHA256 |
|---|---|---|
| candidate.vsix | 826236 | c1eca4caf5a18a1884079cac60cece6d5bb753701bbc0ebb820e15447032d68a |
| candidate.vsix.sha256 | 81 | 4d8ca1311b789a4abb234bea03c9051245262fff5b42c6c9d95e3e3810ded169 |
| candidate.provenance.json | 1813 | 7240b757ac64346a23399cd5d0673e343596b9f42c93f6bdaf70b0cc535f2284 |
Historical H3 archive evidence — superseded download, retained unchanged
Exact archive
The release reuses the original H3 archive, packaged once on Linux by source 4a27ae0 (tree a20c19dc66670bf39bb3e203ede0081f6048ddcf), from origin run 38769699307. Production source is byte-identical to that producer. No rebuild or repackage was performed. The v1.0.3 tag points to reviewed merge 06b6d36, tree 5121ee16d7ddcc9839c9416d13cefa9d5f988604.
| Asset | Bytes | SHA256 |
|---|---|---|
| candidate.vsix | 1079814 | 416ea51fa31471545b022ca37f7ab30901a2e36b871ba230b54bbeadb97943ec |
| candidate.vsix.sha256 | 81 | 969312065029ae5763370a0687a50c2e385af0afea0379bf3a3aeffc5efbf7ab |
| candidate.provenance.json | 267 | 5755d71840acb366010171792889fce565bf1f1a800ee89d74832b7863390e4c |