Repository navigation
fix: small misc fixes (substrate comment, pred_stats heap accounting, docs nav, virtuoso isql port) - #6664
Conversation
… nav, virtuoso isql port) - sparq-substrate/Cargo.toml: the scaffold comment claimed nothing depends on the crate; sparq-engine and sparq-reason do, so it is in the feature-OFF wasm graph (#3806). - sparq-core: TripleStore::heap_bytes now counts the pred_stats planner map, with a direct unit test (#3115). - site: the Docs nav item points at the mdBook guide overlaid at /guide/ instead of the agent-skills router; e2e witness updated (#2549). - scripts/virtuoso-same-box.sh: docker exec isql always targets the container-internal port 1111; VIRTUOSO_ISQL_PORT is only the host mapping (#3361). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ScyGGohDhirnbLbrUSnA5n
…3115) Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ScyGGohDhirnbLbrUSnA5n
🔎 Codex reviewer —
|
The Docs nav link now points at /guide/, but pages.yml ran check-site-links.sh on site/out before the guide was built and overlaid, so lychee could not resolve /guide and failed the deploy. Build and overlay the guide first, and have check-site-links.sh exclude guide/ as an input (its own gate is build-guide.sh) while still resolving the site's links into it. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ScyGGohDhirnbLbrUSnA5n
🔎 Codex reviewer —
|
1 similar comment
🔎 Codex reviewer —
|
|
Local ci-fast gate (GitHub Actions outage; Jesse approved local-gate merges). PR head
Main moved to Squash-merging under the local-gate rule. Generated by Claude Code |
Requested by Jesse · project thread
Summary
Four small, independent fixes:
crates/sparq-substrate/Cargo.tomlscaffold comment said nothing in the workspace depends on the crate, so the wasm build never compiles it. That stopped being true:sparq-enginedepends on it unconditionally (numeric + join + compare) and so doessparq-reason(numeric), so the crate is part of the feature-OFF wasm graph. The comment now says so and points atbench/feature-off-declarations/. No other Cargo.toml has the same "nothing depends on it" wording.TripleStore::heap_bytesnow counts thepred_statsplanner map, as capacity × (entry + 1 control byte), the same modelDictuses for its tables. The new unit testheap_bytes_counts_pred_statsfailed before the fix and passes after it. Any memstat or B/triple metric built onheap_byteswill rise a little, by a fixed amount: about 1 KiB for ~90 predicates.skills/SKILL.mdon GitHub, which is the agent-skills router. It now points at the human-facing mdBook guide, whichpages.ymloverlays at/guide/(Decide + implement the Pages mount for the mdBook guide (the actual blocker on publishing docs) #5022). It is anexternalhard-navigation slot like/app, with the basePath prefix and a trailing slash. The e2e witness insite/e2e/site-nav.spec.tsnow asserts the/guide/href.scripts/virtuoso-same-box.shrandocker exec … isql $VIRTUOSO_ISQL_PORT, but inside the container isql always listens on 1111.VIRTUOSO_ISQL_PORTis only the host-side-pmapping. Bothdocker execcalls (the readiness poll and the bulk load) now use 1111.Not changed: #3090 (glib >= 0.20 in
gui/src-tauri/Cargo.lock). Runningcargo update -p glib --precise 0.20.0fails withfailed to select a version for the requirement glib = "^0.18", required bygtk v0.18.2←tauri v2.11.3. On Linux, Tauri 2 is tied to the gtk-rs 0.18 / GTK3 stack (see #2818), so the bump needs a Tauri major migration and is not possible as a lockfile change.Closes #3806
Closes #3115
Closes #2549
Closes #3361
Base gate (always required)
cargo build --workspacesucceeds. (Left to CI. I ran checks scoped to the crate only.)cargo clippy --workspace --exclude sparq-py --all-targets -- -D warningsis clean. (Left to CI.cargo clippy -p sparq-core --all-targets --all-features -- -D warningsis clean.)cargo testpasses for every crate this PR touches (cargo test -p sparq-core). I did not run the site e2e locally because there is no node_modules; it is left to CI.Targeted re-evaluation (check the rows that apply to your change)
heap_bytesaccounting changed; storage and encoding are untouched. Expect a small, fixed rise in heap_bytes-derived metrics.heap_bytesis a benchmarking accessor in always-compiled sparq-core. Its effect on the bundle size is left to thewasm_bundle_bytesgate.Ratchets and conventions
TODO/FIXMEmarkers added.Security
🤖 Generated with Claude Code
https://claude.ai/code/session_01ScyGGohDhirnbLbrUSnA5n
Generated by Claude Code