test(tri_sha256): assert two-block chaining in-spec (closes an out-of-CI proof) - #371
Merged
Conversation
…-CI proof) The pad2 comment admitted the gap outright: 'full two-block bit-exactness vs hashlib is proven by the trinet_ledger_head binary' -- i.e. the multi-block property was proven OUTSIDE the gates, so no PR ever re-checked it. The specs only ever asserted single-block vectors. Add sha256_two_block_64a: sha256(64 x 'a'), digest ffe054fe...154668eb from the OS SHA-256. That 64-byte message fills block 1 exactly, so all padding lands in block 2 (0x80 marker in w0, the 512-bit length in w15) -- exactly the sha256_pad2_word layout. Block 1 compresses from the standard IV via sha256_word; block 2 compresses from THAT intermediate state via sha256_compress, asserting all 8 final words. This is the first test that exercises state chaining at all. Written as ONE test block (8 intermediates reused across 8 assertions = 16 compress calls) rather than 8 separate tests (which would have cost 72), so the icarus run stays sub-second. Verified: 19/19 under icarus and zig. gen/rust is unchanged (gen-rust does not emit test blocks, so the dedicated sha256-gen-check stays green); tri_sha256 has no C gen. The stale pad2 comment now points at the in-spec proof. gen/zig regeneration via the no-gen-edits escape (LEFTHOOK=0). 🤖 Generated with [Claude Code](https://claude.com/claude-code)
gHashTag
pushed a commit
that referenced
this pull request
Aug 9, 2026
…stop the bleed Same trap as the dead test target (#376), one directory over. Cargo.toml sets 'autobins = false', so a file in src/bin/ is compiled ONLY with an explicit [[bin]] block. There is exactly ONE such block for 29 files: 28 binaries are never built, never type-checked, free to rot silently. Among the dead: trinet_compute_lifecycle.rs -- the end-to-end proof that tests/compute_ring_invariants.rs cites as living outside cargo -- and trinet_ledger_head.rs, the binary the tri_sha256 spec credited with proving two-block bit-exactness before #371 moved that proof in-spec. Triaged all 28 with cargo check, one target at a time: 24 compile once the gen/rust modules they include by #[path] exist. Ten of those gens were simply never committed even though every spec is present (the same omission #376 hit). NOT revived here: those #[path] mods lack #[allow(clippy::all, unused)], so 'clippy --all-targets -D warnings' reports 349 errors from the GENERATED code and would redden the required build+test gate. Reviving them is attribute surgery across 24 files -- its own change. 3 are genuinely rotted, and this records why: tri_rti unresolved import trios_mesh::rti trinet_a2a_node needs gen/rust/router_ttl.rs -- no such spec exists trios_meshd method 'send' is not a member of trait 'Transport' The guard makes the debt visible and bounded: every src/bin/*.rs must be declared or named in KNOWN_DEAD, so a NEW binary can never join them silently. Shrink the list; never grow it. Verified: clean tree exits 0 and prints '29 files, 1 declared, 28 known-dead'; a new undeclared bin is caught; the test-side guard still catches an undeclared test file. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
gHashTag
added a commit
that referenced
this pull request
Aug 9, 2026
…stop the bleed (#377) Same trap as the dead test target (#376), one directory over. Cargo.toml sets 'autobins = false', so a file in src/bin/ is compiled ONLY with an explicit [[bin]] block. There is exactly ONE such block for 29 files: 28 binaries are never built, never type-checked, free to rot silently. Among the dead: trinet_compute_lifecycle.rs -- the end-to-end proof that tests/compute_ring_invariants.rs cites as living outside cargo -- and trinet_ledger_head.rs, the binary the tri_sha256 spec credited with proving two-block bit-exactness before #371 moved that proof in-spec. Triaged all 28 with cargo check, one target at a time: 24 compile once the gen/rust modules they include by #[path] exist. Ten of those gens were simply never committed even though every spec is present (the same omission #376 hit). NOT revived here: those #[path] mods lack #[allow(clippy::all, unused)], so 'clippy --all-targets -D warnings' reports 349 errors from the GENERATED code and would redden the required build+test gate. Reviving them is attribute surgery across 24 files -- its own change. 3 are genuinely rotted, and this records why: tri_rti unresolved import trios_mesh::rti trinet_a2a_node needs gen/rust/router_ttl.rs -- no such spec exists trios_meshd method 'send' is not a member of trait 'Transport' The guard makes the debt visible and bounded: every src/bin/*.rs must be declared or named in KNOWN_DEAD, so a NEW binary can never join them silently. Shrink the list; never grow it. Verified: clean tree exits 0 and prints '29 files, 1 declared, 28 known-dead'; a new undeclared bin is caught; the test-side guard still catches an undeclared test file. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: SSD DDD <ssdm4@MacBook-Pro.local>
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.
The pad2 comment admitted the gap outright: "full two-block bit-exactness vs hashlib is proven by the trinet_ledger_head binary" — i.e. the multi-block property was proven outside the gates, so no PR ever re-checked it. The specs only ever asserted single-block vectors.
Added
sha256_two_block_64a:sha256(64 × 'a'), digestffe054fe…154668ebfrom the OS SHA-256. That 64-byte message fills block 1 exactly, so all padding lands in block 2 (0x80 marker in w0, the 512-bit length in w15) — exactly thesha256_pad2_wordlayout. Block 1 compresses from the standard IV viasha256_word; block 2 compresses from that intermediate state viasha256_compress, asserting all 8 final words. This is the first test that exercises state chaining at all.Written as one test block (8 intermediates reused across 8 assertions = 16 compress calls) rather than 8 separate tests (which would have cost 72), so the icarus run stays sub-second.
Verified: 19/19 under icarus and zig.
gen/rustis unchanged (gen-rust doesn't emit test blocks, so the dedicatedsha256-gen-checkstays green); tri_sha256 has no C gen. The stale pad2 comment now points at the in-spec proof.Continues the semantic-anchor series (#370 added the empty-string NIST vector).
gen/zig regeneration via the
no-gen-editsescape (LEFTHOOK=0).🤖 Generated with Claude Code