[devin-ai-integration]

What and why

The 0.11.0 performance record against v0.10.1 (#1337) puts the StateLoop and ExecuteState rows +12 % in allocations, bisected by exact allocation counts to #1136 (transitive order dependence). Half of that is the statement order rebuilt per entry, fixed in #1355; the other half is here: taking a transition made two closures on every transition taken.

StateExecutor.taking put the executor under the choice points the guard stood on through Context.underControl, which returned a closure releasing the control, and itself returned a closure calling that and putting the firing transition and notes back — two heap allocations per transition taken, on every route resolved, every firing, every resumed cascade; a machine looping on a transition ten thousand times made twenty thousand of them. underControl now returns the position to release (releaseControl(n)), and taking returns a taken value whose restore method puts the executor back, so taking a transition allocates nothing of its own. The seven underControl and five taking call sites change form only (defer ctx.releaseControl(ctx.underControl(by)), defer e.taking(trans, notes).restore()).

No behaviour changes: the same control is held for the same extent and the same state restored.

How it was verified

Failing-before test TestTakingATransitionAllocatesNothing (internal/exec/runtime/taking_allocs_test.go): taking and restoring each of a machine's three transitions allocates 6.00 objects on develop (two per transition, as exec.taking(tr, nil)()), 0 on this branch. The per-event differential test of #1355 (one more event of a counting state machine) goes 133.50 → 127.50 allocations on this branch alone.

A/B over six interleaved rounds (go test -bench -count 1 per round, nice; pre-#1136 d5e2317 | develop a662c2b | this branch | this branch + #1355). Allocations are exact; the machine was loaded, so only the allocation column is stated:

row	allocs/op
ExecuteState	12.28k | 13.75k | 13.14k | 12.39k
StateLoop/count50	12.26k | 13.72k | 13.12k | 12.37k
StateLoop/count500	119.4k | 134.0k | 128.0k | 120.5k
StateLoop/count5000	1.191M | 1.336M | 1.276M | 1.201M
ActionLoop/for1000	4 239 | 6 256 | 6 256 | 4 241
ActionChain/chain1000	66.01k | 80.90k | 80.97k | 66.38k
REPL RunStateMachine/elements=*	226 | 228 | 228 | 228

This branch takes the state rows from +12 % to +7 %; with #1355 they are within +0.9 % of pre-#1136 (the last six objects per thousand events are #1276's declared-types lookup). The action rows are #1355's and untouched here; the REPL's RunStateMachine (+2 objects, #1315's) is not this cost.

Gates: gofmt -l . empty, go vet ./..., go build ./..., go test ./internal/exec/runtime, python3 scripts/changelog.py check.

Checklist

• make test and make lint pass locally (the package above; the full suite runs in CI)
• Tests added or updated for the change
• Documentation extended where it already covers the surface (none covers this)
• Changelog entry added as changes/unreleased/take-transition-without-closures.fixed.md, not as an edit to CHANGELOG.md
• baselines regenerated and make docs-counts run if a gate count moved (none moved)
• No internal work-item labels (waves, slices, F4, K5) in the body, docs, or changelog

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/cf22d7a746814d20b2f0eaf11682fef9
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/cf22d7a746814d20b2f0eaf11682fef9?variant=devin
Requested by: @HuiJun

---

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.