[repo-assist] Monthly Activity 2026-08

This is an automated monthly activity report rather than a single change. Review the linked PRs #462 and the two PRs created in the August 28 run, then read issues #452 and #62 for context; there is no single entry point or defined done condition for a first-time contribution.

🤖 Repo Assist here — I'm an automated AI assistant for this repository.

Activity for August 2026

Suggested Actions for Maintainer

Comprehensive list of all pending actions requiring maintainer attention:

• Review PR #462: Dependabot bump of github/gh-aw-actions/setup 0.86.2→0.87.5 — View
• Review PR: eng: bump GitHub Actions to latest versions (checkout v7, setup-dotnet v6, cache v6, upload-artifact v7, dorny/test-reporter v3) — created this run
• Review PR: tests: add coverage for side-effect re-execution with independent cancellation tokens — created this run
• Investigate #452: reporter published a standalone minimal repro (Neftedollar/taskseq-orleans-repro) confirming TaskSeq.map/taskSeq { for .. yield .. } duplicate the final item when wrapping an externally-produced IAsyncEnumerable driven by Orleans' custom per-activation scheduler. Root cause narrowed to the shared CE/resumable-code machinery (TaskSeqBuilder.fs), but cannot be reproduced offline without real Orleans infra — needs a maintainer or contributor with an Orleans dev environment to drive further — View
• Check comment on #62: tail recursion discussion — awaiting maintainer direction on whether to pursue stub overloads for TaskSeqBuilder now that upstream YieldFromFinal fix has landed — View

Future Work for Repo Assist

• Do NOT re-attempt: rev/sort, allPairs, zip4, map3, iteri2, splitInto, exists2, map2, findBack (all rejected/closed)
• Do NOT re-engage on issue #139 (TaskEx index) — maintainer feedback was pointed; only re-engage if new human activity appears
• Issue #452: bug is confirmed to live in the shared for-loop-over-external-IAsyncEnumerable CE machinery, not TaskSeq.map-specific code; not safe to attempt a blind fix without real Orleans repro access
• Test files still without a SideEffects module: TaskSeq.Empty.Tests.fs, TaskSeq.Let.Tests.fs, TaskSeq.ReplicateInfinite.Tests.fs, TaskSeq.ThreadState.Tests.fs (CancellationToken.Tests.fs now has one, added this run)
• Consider evaluating fsdocs-tool 21.0.0 → 22.1.0 bump in a future engineering run (deferred this run, no breaking-change review done yet)

Run History

2026-08-28 — Run

• 🔧 Created PR: eng: bump GitHub Actions to latest versions across all workflows (checkout v7, setup-dotnet v6, cache v6, upload-artifact v7, dorny/test-reporter v3) — build succeeds, 5432 tests pass, 2 skipped
• 🔧 Created PR: tests: add coverage for side-effect re-execution with independent cancellation tokens (test-only, TaskSeq.CancellationToken.Tests.fs) — 19/19 targeted tests pass
• 💬 Commented on #452: summarized the characterization matrix from the reporter's standalone repro, confirmed the shared-CE-machinery root cause connecting TaskSeq.map and taskSeq { for .. yield .. }, and suggested next steps (instrumented build run against the public repro rather than further offline approximation)
• 📁 Closed #443 (July 2026 monthly activity) and opened this issue for August

Generated by 🌈 Repo Assist, see workflow run.
_{Comment /repo-assist to run again}

Generated by 🌈 Repo Assist, see workflow run. Learn more.
_{Comment /repo-assist to run again}

gh aw add githubnext/agentics/workflows/repo-assist.md@ae8d551f07c7ed7619f8c58c7bb4c3ac89395d38