Repository navigation
Cross-crate call graphs stop at boundaries #56
Description
Root cause:
httplink only matches REST URL paths between services
ApproachL
New pipeline pass that parses .proto services, finds generated XxxClient::method() calls, and links them to server-side impls via proto naming conventions (prost/tonic codegen is predictable)
Activity
fix(security): close open scanner alerts
fix(security): close open scanner alerts
fix(security): close open scanner alerts
fix(security): close open scanner alerts
fix(pkgmap+routes): resolve workspace imports + extract SvelteKit routes
DeusData commentedon Jun 12, 2026
Late update on this one — two separate parts:
In-workspace cross-crate calls: v0.8.0 added a Rust hybrid-LSP extraction path (current is v0.8.1), which materially improves call resolution across crates inside one indexed workspace. If your boundary cases were within a Cargo workspace, a retest on v0.8.1 would be very useful — please share what (if anything) still stops at the crate boundary.
gRPC client→server linking (parsing
.protoservices and linking generatedXxxClient::method()calls to server-side impls via tonic/prost naming conventions): not shipped yet — keeping this issue open to track it. The existing cross-service linker only matches REST URL paths today, as you diagnosed.A v0.8.1 result on part 1 would help sharpen exactly how much of this remains a proto-linker feature vs. already covered. Thanks!
test(repro): cross-file + trace reproductions for #408, #56, #480
test(repro): #548 lighter harness, strengthen #56, + 3 NEW bugs
fix: bulk parallel — Rust cross-crate, Python builtins, #607 reinstal…
DeusData commentedon Jun 28, 2026
This issue now has a first-class, durable reproduction in the cumulative bug-reproduction suite landing in #667 — a test that pins the exact failing behaviour, so it's tracked and guarded against silent regression. Where #667 also lands the root-cause fix, that's called out in the PR; otherwise the reproduction makes the eventual fix straightforward and verifiable.
The reproduction goes live once #667 merges. (Posted from the release branch's QA pass.)
fix: 17 reproduce-first issue + new-bug fixes (per-case verified)
DeusData commentedon Jul 26, 2026
Correcting our own bookkeeping here: this issue was labelled
awaiting-reporter, which arms an automated stale/close cycle. That was wrong, and I have removed it — you do not owe us anything on this one.Where it actually stands: the reproduction promised in the last update did land. #667 merged on 2026-06-28 (
36d83280) and the case now lives in-tree astests/repro/repro_issue56.c, pinning cross-crateCALLSresolution against a real two-crate Cargo workspace fixture rather than the single-file LSP layer that the earlierrustlsp_extra_cargo_wires_workspace_membertest covered.So the bug is confirmed, reproduced, guarded against silent regression — and still unfixed. The root cause documented in that test is that the tree-sitter-only Rust extractor has no Cargo metadata at extraction time, so a
crate_a::helper()call site incrate_bhas no registry entry to resolve against and noCALLSedge is minted.The next step is ours, not yours. Apologies for a label that suggested otherwise — and for the fact that it would have auto-closed this as
not plannedin three days. The stale workflow has been changed so it can no longer close issues at all; it now only flags them for a human to review.Thanks for the original report, and for your patience with it.
fix(stale): mark awaiting-reporter issues stale, never auto-close
DeusData commentedon Aug 29, 2026
Partial progress landed on main in
56dc82ec(PR #1850, "fix: preserve Cargo context in parallel Rust resolution").Deliberately leaving this open. That PR references this issue rather than claiming to close it, and the distinction is real: it fixes one mechanism by which cross-crate edges were lost — Cargo context not surviving parallel Rust resolution — but "call graphs stop at crate boundaries" is a broader symptom that may have other contributing causes.
If you are in a position to re-test against main, it would be genuinely useful to know whether your case improves, stays the same, or improves partially. That distinguishes "this was the cause" from "this was a cause" better than anything I can determine from the code alone.
test(rust): rebase cargo-route mutants and port the workspace route t…
feat(rust): admit cross-crate workspace CALLS behind the fail-closed …