[lawrencecchen]

Summary

Follow-up to #32616 (REOPEN-F7, empty labels need no TextKit layout), which cmux-bot merged at 0b92974 before its review P2 was fixed. This PR is that fix, with no product change:

• Review P2 (REVIEW.md section 8): LabelText.lastLaidOutString, a test-only read in production source, is removed.
• LabelText.measure takes its TextKit layout as a parameter (layOut, type LabelText.Layout, default laidOutMetrics). anEmptyLabelLaysNoTextOut counts layouts through it.
• Review P3s: one emptyLine(font:) builder instead of two copies; the test covers a font with nonzero leading (Times New Roman 14) and no font.

Testing

• EmptyLabelMeasureTests (CmuxNextDesignTests):
  • anEmptyLabelLaysNoTextOut: 0 layouts for an empty text at an unbounded width over 8 fonts and 4 shapes; then 1 layout each for a text and for an empty text at a set width.
  • anEmptyLabelLaysOutAsTextKitWould: the fast result equals TextKit at scales 1-3.
• Red: scratch commit 33e7f76 (the seam and this test, no fast path), compile job 30eb160d passed. Fleet step 06fe0512040c855ed90c899f failed as expected: layouts → 32 == 0 and layouts → 34 == 2. The parity test passed.
• Green: the same change at c1e340c (the REOPEN-F7: an empty label's size needs no TextKit layout (sidebar rows at window build) #32616 branch). Fleet step 748d04ad5fa647e33d935cf5 passed both tests. Compile job 8c0726cbe2c0466855a15cd2 passed (CmuxNext, app, cmux-tui). This head is c1e340c cherry-picked onto the feat tip, with the same diff in these two files. The compile and green at this head are below.
• At this head 8fedf1a: compile job c6925fa2300771e49145a460 passed (CmuxNext, app, cmux-tui). Both tests pass in steps bee43816 (aws-m4pro-9) and 3554f3fe (aws-m4pro-6). Then the test process ends with a failure after about 665 s (xctest signal 6, after Contacts CoreData XPC errors), and the lane marks the suite failed.
• Base finding, not from this PR: the same suite at the feat tip 73ba3f9, without this change, shows the same pattern (step d823cd6b, Blacksmith run 41565867427): both tests pass, then exit 1 after 665 s. At c1e340c on the older base a14bdc2, the lane passed (step 748d04ad).

Review

A review subagent gave SHIP at c1e340c (comment 6592935500 on #32616). It noted that layOut is a parameter whose default is the production function, and that the parameter adds no stored or debug-only state (P3, no change needed).

Observed / Not run

• Observed: the fleet runs above.
• Not run: no pixel check (the measured values do not change; the parity test checks them).

Changelog

none

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review 8fedf1a: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr32799 -f ref=fix/reopen-f7-label-measure-seam -f platform=macos -f backend_mode=local

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4157787538001 (cc7cdf2066)
• cmux NEXT nightly 1.0.0-nightly.4158055269401 (e5e182cf5b)

[lawrencecchen]

Base finding (not from this PR): the Swift test lane on feat-cmux-next 73ba3f9 ends with a failure about 665 s after the selected tests pass. Base step: d823cd6b71adc026df5cf8ce, https://github.com/manaflow-ai/cmuxterm-hq/actions/runs/41565867427. On the owned Macs it is xctest signal 6 after Contacts CoreData XPC errors (steps bee43816, 3554f3fe). The same tests pass in all of these runs.

[lawrencecchen]

Review (subagent opus): SHIP at 8fedf1a

• Same change as reviewed: at this head, both files have the same blobs as c1e340c (LabelText.swift 30ee426, EmptyLabelMeasureTests.swift 0da779d). The base blobs equal REOPEN-F7: an empty label's size needs no TextKit layout (sidebar rows at window build) #32616's merged state (0b92974 = ecce1e0 = 73ba3f9). Feat has not changed these files since 73ba3f9.
• Callers: git grep at 8fedf1a and at the feat tip finds one production caller, ChromeLabel.metrics(width:) (ChromeLabel.swift:194). It uses the default layOut, so its behavior is the same. lastLaidOutString has no remaining reference.
• Correctness: the fast path is unchanged. emptyLine(font:) builds the same string as the two copies it replaces. The @MainActor Layout type agrees with the module's .defaultIsolation(MainActor.self). The count test now checks the negative cases too (text, and empty text at a set width, give 2 layouts). The parity test adds Times New Roman 14 (leading) and nil.
• P3, no change: layOut is an injected function parameter. It adds no stored or debug-only state, so REVIEW.md section 8 is met.
• Lane red: I agree it is a base finding. Both tests pass at this head (bee43816, 3554f3fe). The same 665 s xctest exit shows at the feat tip without this change (d823cd6b). A change to label measurement cannot cause a Contacts CoreData XPC abort.

[lawrencecchen]

Correction to my base-finding comment above: the failure about 665 s after the tests passed was not a worker fault. It was the test-bundle run-loop guard holding xctest's main thread, fixed on feat-cmux-next by #33023 / #33013 (ecfbb44) and #33276.