Rightsize ci runners

.changes/checksum-diagnostics.md

@@ -0,0 +1,10 @@
+---
+"rscrypto" = "minor"
+---
+
+Reduce CRC-32 and CRC-64 selection diagnostics to the polynomial, input length, architecture,
+selection reason, effective force setting, and selected kernel. Remove obsolete policy thresholds,
+stream counts, capability flags, and placeholder values that no longer described active dispatch.
+Remove the unused `SelectionReason::BelowSmallThreshold` and `BelowSimdThreshold` variants.
+The numeric discriminants of the remaining selection reasons change.
+Checksum computation and backend selection are unchanged.

.changes/docs-cleanup.md

@@ -2,4 +2,7 @@
 "rscrypto" = "patch"
 ---
 
-Consolidate user documentation, migration guidance, and README assets.
+Remove the migration guide and broad API tours. Keep current API contracts, feature selection,
+secret ownership, and focused executable examples in their owning documentation.
+Correct platform coverage and the limits of retained evidence.
+Update installation examples for v0.10 and remove temporary development-version notes.

.changes/ecdsa-secret-selection.md

@@ -0,0 +1,9 @@
+---
+"rscrypto" = "patch"
+---
+
+Preserve masked ECDSA point selection on AArch64 and Windows
+and masked secret selection in portable P-256.
+Use fixed-bound ECDSA table traversal
+and retain unconditional RISC-V generator-table loads under LLVM optimization without changing
+signature semantics.

.changes/fallible-ecdsa-secret-owners.md

@@ -1,5 +1,9 @@
 ---
-"rscrypto" = "patch"
+"rscrypto" = "minor"
 ---
 
-Add fallible blinded P-256/P-384 signing and public derivation, plus zero-copy and in-place construction for rscrypto-owned secret buffers and UTF-8 strings.
+Expose fallible P-256/P-384 blinded signing and public-key derivation with caller-provided entropy.
+Remove `public_key_blinded` and `try_sign_blinded`; the supported entry points are
+`try_public_key_blinded_with` and `try_sign_blinded_with`, which report entropy failure before private arithmetic.
+Add direct-fill constructors for SecretBytes and SecretVec,
+plus allocation-preserving ownership transfer from Vec and String.

.changes/feature-target-builds.md

@@ -0,0 +1,8 @@
+---
+"rscrypto" = "patch"
+---
+
+Restore Rust 1.91 Ed25519 and X25519 builds on x86-64 Linux when assembly owns fixed-base dispatch,
+and keep standalone AEAD features lint-clean on Linux.
+Exclude SIMD backends from scalar WebAssembly hash, AEAD,
+and Argon2 dispatch when SIMD128 is disabled.

.changes/internal-evidence-boundary.md

@@ -0,0 +1,14 @@
+---
+"rscrypto" = "minor"
+---
+
+Limit ordinary `diag` builds to capability and backend-selection introspection. Remove benchmark,
+forced-kernel, constant-time, zeroization, and component operations from ordinary public module paths,
+re-exports, and associated methods across AEADs, MACs, KDFs, password hashing, signatures, key agreement,
+RSA, ML-KEM, and cryptographic hashes. This includes RSA seeded diagnostic encryption methods,
+BLAKE3 diagnostic selectors, the SHA-256 benchmark compression helper, and the diagnostic-only
+`Argon2Error::BackendUnavailable` variant. Removing that variant changes the numeric discriminant of
+`VerificationLimitTooLow` in ordinary `diag` builds with `phc-strings`.
+Repository evidence tools retain explicit internal access. Internal PBKDF2 verification probes exercise
+the primitive instead of rejecting their fixed parameters through the application password policy.
+Application cryptographic operations are unchanged by this boundary cleanup.

.changes/release-lockfiles.md

@@ -0,0 +1,5 @@
+---
+"rscrypto" = "patch"
+---
+
+Synchronize standalone fuzz and evidence-tool lockfiles with the rscrypto version during release preparation.

.changes/remove-unbounded-pbkdf2-verification.md

@@ -0,0 +1,7 @@
+---
+"rscrypto" = "minor"
+---
+
+Remove `verify_with_policy` and `verify_password_with_policy` from PBKDF2-SHA256 and PBKDF2-SHA512.
+Explicit password policies use `verify_with_policy_bounded` and `verify_password_with_policy_bounded`,
+which require an upper iteration limit. Default password verification and explicit primitive operations remain.

.config/rail.toml

@@ -32,3 +32,48 @@ msrv_policy = { mode = "disabled" } # Cargo.toml owns the declared package MSRV.
 # Cargo Rail 0.25 cannot distinguish expected `compile_fail` doctest invocations from compiler failures.
 enabled = false
 consumer_scope = "open" # Published library: external consumers remain part of the visibility boundary.
+
+[release]
+# Keep standalone fuzz and evidence-tool lockfiles on the released package version.
+auxiliary_cargo_manifests = [
+  "fuzz-packages/aead-aegis256/Cargo.toml",
+  "fuzz-packages/aead-aes-siv-cmac256/Cargo.toml",
+  "fuzz-packages/aead-aes128gcm/Cargo.toml",
+  "fuzz-packages/aead-aes128gcmsiv/Cargo.toml",
+  "fuzz-packages/aead-aes256gcm/Cargo.toml",
+  "fuzz-packages/aead-aes256gcmsiv/Cargo.toml",
+  "fuzz-packages/aead-ascon128/Cargo.toml",
+  "fuzz-packages/aead-chacha20poly1305/Cargo.toml",
+  "fuzz-packages/aead-header-protection/Cargo.toml",
+  "fuzz-packages/aead-nonce-counter/Cargo.toml",
+  "fuzz-packages/aead-xchacha20poly1305/Cargo.toml",
+  "fuzz-packages/auth-argon2/Cargo.toml",
+  "fuzz-packages/auth-ed25519/Cargo.toml",
+  "fuzz-packages/auth-hkdf/Cargo.toml",
+  "fuzz-packages/auth-hmac/Cargo.toml",
+  "fuzz-packages/auth-kmac256/Cargo.toml",
+  "fuzz-packages/auth-p256-ecdh/Cargo.toml",
+  "fuzz-packages/auth-pbkdf2/Cargo.toml",
+  "fuzz-packages/auth-phc/Cargo.toml",
+  "fuzz-packages/auth-scrypt/Cargo.toml",
+  "fuzz-packages/auth-x25519/Cargo.toml",
+  "fuzz-packages/checksums-crc16/Cargo.toml",
+  "fuzz-packages/checksums-crc24/Cargo.toml",
+  "fuzz-packages/checksums-crc32/Cargo.toml",
+  "fuzz-packages/checksums-crc64/Cargo.toml",
+  "fuzz-packages/fast-rapidhash/Cargo.toml",
+  "fuzz-packages/fast-xxh3/Cargo.toml",
+  "fuzz-packages/hash-ascon/Cargo.toml",
+  "fuzz-packages/hash-blake2/Cargo.toml",
+  "fuzz-packages/hash-blake3/Cargo.toml",
+  "fuzz-packages/hash-sha2/Cargo.toml",
+  "fuzz-packages/hash-sha3/Cargo.toml",
+  "fuzz-packages/hash-websocket-sha1/Cargo.toml",
+  "fuzz-packages/surface-hex-parse/Cargo.toml",
+  "fuzz-packages/traits-io/Cargo.toml",
+  "fuzz/Cargo.toml",
+  "tools/ct-binsec-harness/Cargo.toml",
+  "tools/ct-dudect/Cargo.toml",
+  "tools/ct-harness/Cargo.toml",
+  "tools/wasm-runtime-vectors/Cargo.toml",
+]

.config/tooling.toml

@@ -73,7 +73,7 @@ nightly-components = ["miri", "rust-src"]
 packages = ["build-essential", "ca-certificates", "curl", "git", "python3"]
 
 [ci-ct]
-cargo = ["just"]
+cargo = ["just", "cargo-nextest"]
 components = ["llvm-tools"]
 packages = ["build-essential", "ca-certificates", "curl", "git", "python3"]
 

.github/runs-on.yml

@@ -1,5 +1,5 @@
-# AWS runner profiles. Public repositories resolve this catalog from the default branch.
-# Workflows reference these names; other providers keep their own labels.
+# AWS runner profiles. Keep sizing here; workflows select names without shape overrides.
+# Public repositories resolve this catalog from the default branch.
 runners:
   plan:
     family: [c7a.large, c7i.large, c8a.large, c8i.large]
@@ -8,14 +8,14 @@ runners:
     spot: price-capacity-optimized
     volume: 40gb:gp3
   miri:
-    family: [c7a.xlarge, c7i.xlarge, c8a.xlarge, c8i.xlarge]
-    cpu: 4
+    family: [c7a.large, c8a.large]
+    cpu: 2
     image: ubuntu24-minimal-x64
     spot: price-capacity-optimized
     volume: 150gb:gp3
   package:
-    family: [c7a.2xlarge, c7i.2xlarge, c8a.2xlarge, c8i.2xlarge]
-    cpu: 8
+    family: [c7a.xlarge, c7i.xlarge, c8a.xlarge, c8i.xlarge]
+    cpu: 4
     image: ubuntu24-minimal-x64
     spot: price-capacity-optimized
     volume: 100gb:gp3
@@ -25,12 +25,6 @@ runners:
     image: ubuntu24-minimal-x64
     spot: price-capacity-optimized
     volume: 150gb:gp3
-  cross-build:
-    family: [c7a.8xlarge, c7i.8xlarge, c8a.8xlarge, c8i.8xlarge]
-    cpu: 32
-    image: ubuntu24-minimal-x64
-    spot: price-capacity-optimized
-    volume: 150gb:gp3
   ci-cross-build:
     family: [c7a.4xlarge, c7i.4xlarge, c8a.4xlarge, c8i.4xlarge]
     cpu: 16
@@ -62,11 +56,12 @@ runners:
     spot: price-capacity-optimized
     volume: 150gb:gp3
   x86_64-win:
-    family: [c7a.4xlarge, c7i.4xlarge, c8a.4xlarge, c8i.4xlarge]
-    cpu: 16
+    family: [c7a.2xlarge, c8a.2xlarge]
+    cpu: 8
     image: windows25-full-x64
     spot: price-capacity-optimized
     volume: 200gb:gp3
+  # Eight concurrent fuzz targets need eight physical cores, including on x86.
   fuzz-x86_64-linux:
     family: [c7a.2xlarge, c8a.2xlarge]
     cpu: 8
@@ -79,6 +74,7 @@ runners:
     image: ubuntu24-minimal-arm64
     spot: price-capacity-optimized
     volume: 150gb:gp3
+  # Fixed, non-Flex, On-Demand CT hardware; sized independently of benchmarks.
   ct-x86_64-linux-intel:
     family: [c8i.2xlarge]
     cpu: 8
@@ -97,7 +93,7 @@ runners:
     image: ubuntu24-minimal-arm64
     spot: false
     volume: 100gb:gp3
-  # Fixed, non-Flex, On-Demand hardware for timing and benchmark evidence.
+  # Fixed benchmark hardware preserves the existing measurement baseline.
   measure-x86_64-linux-intel:
     family: [c8i.2xlarge]
     cpu: 8

.github/workflows/bench.yml

@@ -66,7 +66,7 @@ jobs:
     needs: plan
     if: needs.plan.outputs.cross == 'true'
     name: Prepare ${{ matrix.target }} benchmarks
-    runs-on: runs-on=${{ github.run_id }}/runner=cross-build/env=production
+    runs-on: runs-on=${{ github.run_id }}-${{ matrix.target }}/runner=bench-cross-build/env=production
     timeout-minutes: 120
     strategy:
       fail-fast: true

.github/workflows/ci.yml

@@ -24,7 +24,7 @@ defaults:
 jobs:
   cross-build:
     name: Prepare ${{ matrix.target }} tests
-    runs-on: runs-on=${{ github.run_id }}/runner=cross-build/env=production
+    runs-on: runs-on=${{ github.run_id }}-${{ matrix.target }}/runner=ci-cross-build/env=production
     timeout-minutes: 45
     strategy:
       fail-fast: true
@@ -59,7 +59,7 @@ jobs:
       matrix:
         include:
           - platform: compatibility
-            runner: runs-on=${{ github.run_id }}/runner=cross-build/env=production
+            runner: runs-on=${{ github.run_id }}/runner=compatibility/env=production
             timeout: 60
           - platform: package
             runner: runs-on=${{ github.run_id }}/runner=package/env=production
@@ -109,6 +109,7 @@ jobs:
           just ci-check
           just test --all --release
           just test --all --release --portable
+          if [[ "$PLATFORM" != x86_64-linux ]]; then just test-evidence; fi
           if [[ "$PLATFORM" == x86_64-linux || "$PLATFORM" == aarch64-linux ]]; then just test-musl; fi
           git diff --exit-code HEAD
 
@@ -169,5 +170,7 @@ jobs:
           if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
           just test --all --release --portable
           if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
+          just test-evidence
+          if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
           git diff --exit-code HEAD
           if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

.github/workflows/ct.yml

@@ -84,8 +84,7 @@ jobs:
 
   select:
     if: ${{ !inputs.replay_p384 }}
-    # Public-repository named runners resolve from main, not this workflow's ref.
-    runs-on: runs-on=${{ github.run_id }}/cpu=2/family=c7a.large+c7i.large+c8a.large+c8i.large/image=ubuntu24-minimal-x64/spot=price-capacity-optimized/volume=40gb:gp3/env=production
+    runs-on: runs-on=${{ github.run_id }}/runner=plan/env=production
     timeout-minutes: 5
     outputs:
       matrix: ${{ steps.select.outputs.matrix }}
@@ -101,7 +100,7 @@ jobs:
   cross-build:
     needs: select
     if: needs.select.outputs.cross == 'true'
-    runs-on: runs-on=${{ github.run_id }}-${{ matrix.target }}/cpu=32/family=c7a.8xlarge+c7i.8xlarge+c8a.8xlarge+c8i.8xlarge/image=ubuntu24-minimal-x64/spot=price-capacity-optimized/volume=150gb:gp3/env=production
+    runs-on: runs-on=${{ github.run_id }}-${{ matrix.target }}/runner=ct-cross-build/env=production
     timeout-minutes: 45
     name: Prepare ${{ matrix.target }} CT
     strategy:
@@ -185,6 +184,5 @@ jobs:
         uses: "actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a" # v7.0.1
         with:
           path: target/ct-${{ matrix.platform }}-full.tar.gz
-          archive: false
           if-no-files-found: warn
           retention-days: 7

.github/workflows/fuzz.yml

@@ -58,7 +58,7 @@ jobs:
         include:
           - platform: x86_64-linux
             task: miri
-    runs-on: runs-on=${{ github.run_id }}/runner=${{ matrix.task == 'miri' && 'miri' || matrix.platform }}/env=production
+    runs-on: runs-on=${{ github.run_id }}/runner=${{ matrix.task == 'miri' && 'miri' || format('fuzz-{0}', matrix.platform) }}/env=production
     timeout-minutes: ${{ (matrix.task == 'miri' || github.event_name == 'pull_request') && 60 || 90 }}
     steps:
       - uses: "actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1" # v7.0.1