P1: fix Zig FFI to build and match the Idris2 ABI#33
Merged
Conversation
The Zig FFI bridge is half of the ABI-FFI standard, but nothing installed Zig: .tool-versions only lists it (commented), setup.sh stops at `just`, and the devcontainer's `postCreateCommand: just deps` referenced a `deps` recipe that did not exist. Unlike the other toolchain pieces, Zig is not distributed via GitHub releases, so it must come from ziglang.org. Add scripts/install-zig.sh: an idempotent, fail-soft installer for the pinned Zig 0.14.0 (arch/OS-aware, uses the system CA store the agent proxy populates, never --insecure). If ziglang.org is not on the session's egress allowlist the download 403s and the script exits 0 with an actionable message, so it never blocks setup or a session. Wire it in via the two paths the project already uses: a "Step 1b" in setup.sh (where the template exposes that step), and a new `deps` Justfile recipe backing the devcontainer postCreateCommand. Once ziglang.org is allowlisted, future setups and dev containers install Zig automatically. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019xMKB3T4Vo5FYC7Czx3JSH
`zig test src/main.zig -lc` crashed with "free(): double free detected" on the second test (create supervisor and worker). Root cause: dual ownership of tree nodes. The handle's flat `nodes` ArrayList owns every node created by otpiser_create_supervisor / otpiser_create_worker, while otpiser_add_child also links a child into its parent supervisor's `children` list. TreeNode.deinit then recursively freed each child (child.deinit() + allocator.destroy(child)), so any node that was added as a child was freed twice on otpiser_free: once via the parent's recursive deinit and once via the handle's flat `nodes` sweep. Fix: make `children` a non-owning reference list. TreeNode.deinit now releases only its own `children` container and no longer frees the child nodes. The handle's `nodes` list remains the single owner and frees every node exactly once. This matches the alloyiser reference idiom (single, flat ownership; handles are plain structs behind ?*T). No exported C symbol names or Result-enum integer values were changed. All 17 C:otpiser_* symbols in Foreign.idr keep their matching export fn, and the Result enum still mirrors resultToInt (Ok=0 .. MalformedTree=6). Verification: - src/interface/ffi: `zig test src/main.zig -lc` -> all 7 tests pass, 0 errors/warnings - src/interface/abi: `idris2 --build otpiser-abi.ipkg` -> exit 0 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019xMKB3T4Vo5FYC7Czx3JSH
…claude/new-session-1fphit
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
2 participants
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.
Problem
zig test src/main.zig -lc(Zig 0.14.0) crashed on the second test:Root cause
Dual ownership of supervision-tree nodes:
nodes: ArrayList(*TreeNode)that owns every node created byotpiser_create_supervisor/otpiser_create_worker.otpiser_add_childalso links a child into its parent supervisor'schildrenlist.TreeNode.deinitthen recursively freed each child (child.deinit()+allocator.destroy(child)).So on
otpiser_free, any node that had been added as a child was freed twice: once via the parent supervisor's recursivedeinit, and again via the handle's flatnodessweep — a classic double free.Fix
Make
childrena non-owning reference list.TreeNode.deinitnow releases only its ownchildrencontainer and no longer frees the child nodes. The handle'snodeslist remains the single owner and frees every node exactly once. This mirrors the alloyiser reference idiom (single flat ownership; handles are plain structs behind?*T).Diff is one file, 6 insertions / 5 deletions in
src/interface/ffi/src/main.zig.ABI fidelity (Idris2 is source of truth)
C:otpiser_*symbols inForeign.idrkeep their matchingexport fn.Resultenum integer values changed; the Zig enum still mirrorsresultToInt(Ok=0 … MalformedTree=6).Verification
cd src/interface/ffi && zig test src/main.zig -lc→ all 7 tests pass, 0 errors/warningscd src/interface/abi && idris2 --build otpiser-abi.ipkg→ exit 0 (thenrm -rf build)C:<name>inForeign.idrhas a matchingexport fn <name>; Result values matchresultToInt.Note: any rust-ci / Hypatia / governance red checks are pre-existing estate-infra issues unrelated to this Zig-only change.
🤖 Generated with Claude Code
https://claude.ai/code/session_019xMKB3T4Vo5FYC7Czx3JSH
Generated by Claude Code