Remove top-level engineering directory

engineering/reviewed-public-api-diff-0.8.1.md → docs/release-evidence/public-api/reviewed-public-api-diff-0.8.1.md

@@ -1,6 +1,6 @@
 # Reviewed public API diff for j2k 0.8.1
 
-This report is generated by `cargo xtask semver --write-report`. Normal `cargo xtask semver` regenerates it in memory and fails if this committed file is stale. Every ordinary added/removed fingerprint and every full rustdoc-hidden candidate-inventory fingerprint requires an exact reviewed entry in `engineering/public-api-review-0.8.1.yml`; report regeneration never updates that review config.
+This report is generated by `cargo xtask semver --write-report`. Normal `cargo xtask semver` regenerates it in memory and fails if this committed file is stale. Every ordinary added/removed fingerprint and every full rustdoc-hidden candidate-inventory fingerprint requires an exact reviewed entry in `docs/release-evidence/public-api/public-api-review-0.8.1.yml`; report regeneration never updates that review config.
 
 - Baseline registry version: `0.8.0`
 - Baseline source snapshot: `v0.8.0` peeled to `53e0ad3d4f75f492af55413e0dab5a5834bd09c6`

docs/release.md

@@ -45,36 +45,38 @@ that are absent from the selected registry releases. The `cpu` feature is
 unaffected. The published `0.8.0` adapters instead perform accelerator codec
 decode, explicit decoded-pixel readback, and ordinary Burn tensor upload using
 released public APIs. Do not recommend the defective 0.7.5 accelerator
-features. Any Burn community notice still requires the post-publication
-exact-version consumer and benchmark evidence listed in the notice draft.
+features.
 
 Version `0.8.0` was intentionally source- and behavior-incompatible
 with `0.7.5`: decode entry points become strict by default, explicit leniency
 is limited to the documented JP2/JPH metadata recoveries, warnings report
 actual recovery rather than lenient configuration, and
 `J2kDecodeWarning::LenientDecodeMode` becomes
-`J2kDecodeWarning::LenientMetadataRecovery`. The
-[reviewed API report](../engineering/reviewed-public-api-diff-0.8.0.md) records
-the release's generated signature diff, and the adjacent
-[review configuration](../engineering/public-api-review-0.8.0.yml) contains the
-exact source- and behavior-break ledger with migrations.
-
-Version `0.8.1` compares directly with published `v0.8.0`. Its
-[reviewed API report](../engineering/reviewed-public-api-diff-0.8.1.md) and
-[review configuration](../engineering/public-api-review-0.8.1.yml) record an
-additive-only public surface: exact-resolution and sRGB/ICC decode APIs,
-encode-stage context, and the shared irreversible midpoint calculation. The
-one-time `0.7.5` to `0.8.0` transition allowance has been removed.
+`J2kDecodeWarning::LenientMetadataRecovery`. The [reviewed API
+report][v0.8.0-api-report] records the release's generated signature diff, and
+the adjacent [review configuration][v0.8.0-api-review] contains the exact
+source- and behavior-break ledger with migrations.
+
+Version `0.8.1` compares directly with published `v0.8.0`. Its [reviewed API
+report](release-evidence/public-api/reviewed-public-api-diff-0.8.1.md) and
+[review configuration](release-evidence/public-api/public-api-review-0.8.1.yml)
+record an additive-only public surface: exact-resolution and sRGB/ICC decode
+APIs, encode-stage context, and the shared irreversible midpoint calculation.
+The one-time `0.7.5` to `0.8.0` transition allowance has been removed.
 
 Version `0.7.3` retained the API contract introduced by `0.7.1`, which
 intentionally contracted parts of the published pre-1.0 `0.6.2` API. It does
 not claim source compatibility with `0.6.x`. The
-[`CHANGELOG`](../CHANGELOG.md) provides migration notes, and the
-[reviewed API report](../engineering/reviewed-public-api-diff-0.7.3.md)
-records the additions, removals, and changed signatures. That report was
-regenerated, independently reviewed, and verified for the published tag.
-Any report prepared for a future release remains provisional until it is
-regenerated and verified after that release's final source freeze.
+[`CHANGELOG`](../CHANGELOG.md) provides migration notes, and the [reviewed API
+report][v0.7.3-api-report] records the additions, removals, and changed
+signatures. That report was regenerated, independently reviewed, and verified
+for the published tag. Any report prepared for a future release remains
+provisional until it is regenerated and verified after that release's final
+source freeze.
+
+[v0.7.3-api-report]: https://github.com/frames-sg/j2k/blob/v0.7.3/engineering/reviewed-public-api-diff-0.7.3.md
+[v0.8.0-api-report]: https://github.com/frames-sg/j2k/blob/v0.8.0/engineering/reviewed-public-api-diff-0.8.0.md
+[v0.8.0-api-review]: https://github.com/frames-sg/j2k/blob/v0.8.0/engineering/public-api-review-0.8.0.yml
 
 ## Candidate freeze and exact-SHA evidence
 

docs/stable-api-1.0.md

@@ -47,7 +47,8 @@ the candidate directly with published 0.8.0. Both reports also record each
 package's complete hidden-inventory count and fingerprint.
 Every semver invocation collects both live passes, compares both committed companions, and
 requires exact ordinary added/removed fingerprints plus the hidden
-count/fingerprint in `engineering/public-api-review-0.8.1.yml`.
+count/fingerprint in
+`docs/release-evidence/public-api/public-api-review-0.8.1.yml`.
 Nonempty hidden inventories also require a package-specific hidden rationale.
 
 The 0.8.0 review file contains the reviewed 0.7.5-to-0.8.0 break ledger. The
@@ -73,11 +74,13 @@ uses Rust `1.96` and does not accept the former `J2K_SEMVER_TOOLCHAIN` override.
 The snapshots record the published workspace's public items and the CLI exit-code
 contract expectations. Manual prose in this file must not duplicate that
 inventory. The completed 0.7.5-to-0.8.0 comparison is in the generated
-[`0.8.0` reviewed API report](../engineering/reviewed-public-api-diff-0.8.0.md).
+[`0.8.0` reviewed API report][v0.8.0-api-report].
 That report became release evidence after source freeze and the exact-SHA
 local, hosted, Metal, and CUDA gates completed. The current comparison is in
 the generated
-[`0.8.1` reviewed API report](../engineering/reviewed-public-api-diff-0.8.1.md).
+[`0.8.1` reviewed API report](release-evidence/public-api/reviewed-public-api-diff-0.8.1.md).
+
+[v0.8.0-api-report]: https://github.com/frames-sg/j2k/blob/v0.8.0/engineering/reviewed-public-api-diff-0.8.0.md
 
 The currently published stable contract is the `0.8.x` line. Version `0.8.0`
 intentionally changed the strict-decoding behavior and one warning variant