[mache-dcb808] chore(leyline): adopt ley-line-open v0.8.0 — validate 27/28, path inference, node_refs.qualifier#528
Open
jamestexas wants to merge 1 commit into
Open
[mache-dcb808] chore(leyline): adopt ley-line-open v0.8.0 — validate 27/28, path inference, node_refs.qualifier#528jamestexas wants to merge 1 commit into
jamestexas wants to merge 1 commit into
Conversation
…27/28, path inference, qualifier Pin v0.7.8 -> v0.8.0 (SHA-verified, fresh download confirmed). VALIDATE SET (empirically probed against the pinned daemon, not assumed): v0.8.0 validates all 32 probed extensions EXCEPT cue (no tree-sitter-0.26 grammar exists). Replaced the hardcoded 8-entry leylineValidateLangs map + extension-stripping langKeyForPath with registry-driven daemonValidates (lang.ForPath != nil, not cue, not HCL) + PATH INFERENCE: the daemon identifies the language from the path, so mache keeps no per-extension language-id map — the old one missed C++'s .cc/.cxx/.hpp and JS's .mjs, and mache's canonical Names (javascript/python/markdown) don't match the daemon's ids (js/py/md) anyway. Deleted the pass-through disclaimers; only cue and non-source extensions (.json etc.) pass through now. HCL kept in-process (hclsyntax) as the fast, offline path — the daemon covers it too but in-process needs no daemon. QUALIFIER (ley-line-open-4dde42, shipped v0.8.0): v_refs mention arm surfaces node_refs.qualifier (probe-guarded) instead of the hardcoded ''. fan_out_skew becomes more accurate — projection-shape call families (require.NoError, assert.Equal) collapse to one qualifier (88->83). Unblocks fatal_call rung-1 (mache-4dc205). Legacy dbs degrade to '' (tested). BASELINE regenerated (deterministic, two byte-identical regens). Delta fully attributable: dead_code 3->5 (v0.8.0 C grammar now parses the vendored elixir scanner.c/parser.h — real dead code, vanishes at PR-B), untested 8->6 (#527 migration tests detected), duplicate_definitions +10 / fan_out +7 net (tier-3 extraction fidelity minus qualifier collapse). Coverage-guard test switched sql->cue (sql now parses; cue is the sole hollow-projection language). PR-B language-coverage decision shrinks to cue-only. Closes ley-line-open-e5addb + ley-line-open-4dde42 (delivered).
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.
Summary
Adopt ley-line-open v0.8.0 (exact pin + SHA map, fresh SHA-verified download confirmed). Everything additive; wire unchanged (
wire_format_major=1, floor 0.6.0).Validate coverage: 8 → 27/28 languages
Empirically probed the pinned daemon (not assumed): it validates every registry language except cue (no tree-sitter-0.26 cue grammar exists). Replaced the hardcoded 8-entry
leylineValidateLangsmap + extension-strippinglangKeyForPathwith registry-drivendaemonValidates(lang.ForPath ≠ nil, not cue, not HCL) + path inference: the daemon identifies the language from the path, so mache keeps no per-extension language-id map — the old one missed C++'s.cc/.cxx/.hpp, JS's.mjs, and mache's canonical names (javascript/python/markdown) don't match the daemon's ids (js/py/md) anyway. Pass-through disclaimers deleted; only cue + non-tree-sitter extensions (.json) pass through. HCL kept in-process (fast/offline).node_refs.qualifier (ley-line-open-4dde42)
v_refsmention arm surfacesnode_refs.qualifier(probe-guarded) instead of the hardcoded''. fan_out_skew becomes more accurate — projection-shape call families (require.NoError,assert.Equal) collapse to one qualifier (88→83). Unblocks fatal_call rung-1 (mache-4dc205). Legacy dbs degrade to''(tested).Version-parity gate fix (latent bug surfaced by the bump)
The gate asserted
binary-minor == schema-client-minor— true only by coincidence while everything was in the 0.7.x line. LLO tags the binary every release but the schema Go module only on wire changes (correctly stuck at v0.7.1 since the wire is unchanged). Rewrote the gate to the real contract: schema pin ∈ [leylineSchemaCompatFloor(0.6.0), binary]. Added a synthetic boundary test. Without this fix the gate would mis-fire on every future minor bump with a stable wire. Verified mache only consumes thebinding/cachecapnp types (unchanged).Baseline regen (deterministic)
Delta fully attributable: dead_code 3→5 (v0.8.0's C grammar now parses the vendored
elixir/scanner.c+parser.h— real dead code, vanishes at PR-B), untested 8→6 (#527 migration tests detected), duplicate_definitions +10 / fan_out +7 net (tier-3 fidelity − qualifier collapse). Coverage-guard test switched sql→cue (sql now parses; cue is the sole hollow-projection language).Strategic consequence
PR-B (mache-37ae8b) language-coverage decision is now zero net loss: deleting the CGO fallback loses construct-parsing for only cue, which no tree-sitter-0.26 grammar can parse anyway.
Closes / unblocks
Closes ley-line-open-e5addb + ley-line-open-4dde42 (both delivered). Unblocks fatal_call rung-1 (mache-4dc205). Shrinks PR-B risk to near-zero.
Verification
Full
task checkgreen; fresh SHA-verified v0.8.0 download exercised; baseline determinism confirmed (two byte-identical regens).🤖 Generated with Claude Code