Add DriverExtension hook for embedding paralegal as a library#196
Draft
JustusAdam wants to merge 4 commits into
Draft
Add DriverExtension hook for embedding paralegal as a library#196JustusAdam wants to merge 4 commits into
JustusAdam wants to merge 4 commits into
Conversation
Lets external rustc drivers (haven's planned MIR rewriter being the motivating case) plug into paralegal between borrow-check and PDG construction without forking. The pre_pdg hook receives the shared Pctx/MarkerCtx so any MIR mutation happens before paralegal's analyzer consumes MIR and any marker lookup the extension does hits the same cache the downstream analyzer will use. NoopExtension preserves the existing standalone paralegal binary's cost profile; DumpOnlyCallbacks skips the (expensive) MarkerCtx build for dependency crates unless an extension opts in via requires_dependency_markers. The Pctx lift incidentally collapses the two MarkerDatabase::init calls the previous flow did per analyzed crate (one in Pctx::new inside into_generator that's no longer needed) into one. Ordering matters: MarkerDatabase::init reads each included crate's .pmeta from disk including the local one, so dump_markers must run before Pctx::new — the cross-crate tests caught this when the dumps and the init were reordered. Both after_expansion paths now document the constraint. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Extensions like haven need to install their own query providers (notably `mir_borrowck` and `fn_sig` for synthetic DefIds) alongside paralegal's own. Since `Config::override_queries` is `Option<fn(..)>` and can't be layered externally, expose a hook on the trait and compose inside `configure()` via a process-wide OnceLock. No behavior change when no extension is registered. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Lets embedders (haven, future paralegal consumers) swap the rustc_driver-linked analyzer binary without forking the cargo orchestrator. The wrapper-shim execs the overridden binary, and the build hash mixes in its identity so cached artifacts from a different analyzer aren't reused silently. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`prepare_compiler_args` in cargo_paralegal_flow owns the boilerplate (VersionArgs parsing, --version/-vV print + nightly-version spoof, EXEC_HASH_ARG strip, RUSTC_WRAPPER mode detection, sysroot append) so both paralegal-flow-impl and embedders linking the analyzer as a library (haven-driver) share one source-of-truth instead of keeping ~100 lines of mirrored argv-prep in sync. Co-located orchestrator change: new --build flag flips the wrapped cargo invocation from `check` to `build` so MIR-rewriting embedders can codegen the rewritten bodies into real binaries. Both paralegal Callbacks already return Compilation::Continue, so codegen runs unconditionally past the analyzer hooks; the flag just unlocks it at the cargo layer. Mixed into the build hash to keep check/build caches disjoint. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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
Adds a
DriverExtensiontrait and arun_with_extensionentry point so external rustc drivers can plug into paralegal between borrow-check and PDG construction without forking. The motivating consumer is haven's planned MIR rewrite pass, which needs to mutate MIR before paralegal walks it.DriverExtension::pre_pdg(tcx, &MarkerCtx)fires after MIR borrow-check, before paralegal consumes MIR.run_with_extension(args, plugin_args, Box<dyn DriverExtension>); existingrunbecomes sugar that passes aNoopExtension.Pctxis now built once at the top ofCallbacks::after_expansionand shared with both the hook and the analyzer, collapsing the previous doubleMarkerDatabase::init.DumpOnlyCallbacksopts into the same hook for dependency crates only when the extension setsrequires_dependency_markers, preserving the standalone paralegal cost profile.Notes for reviewers
The non-obvious ordering constraint surfaced by the cross-crate tests:
MarkerDatabase::initreads each included crate's.pmetafrom disk, including the local one.dump_markers(tcx)therefore has to run beforePctx::new. Bothafter_expansionpaths now document that constraint in code, and therun_in_context_with_pctxhelper expects callers to have already dumped.Marked draft: opening this now to get directional feedback on the trait surface and the timing/ordering changes before haven starts depending on the API.
🤖 Generated with Claude Code