[teamleaderleo]

Leo (2026-11-04): "Couldn't we architect stuff to make everything more amenable to hot swappage? Bridge better api better thin the layer nicer / Idk how well we can pre prepare stored properties/function signatures/deal with files"

This is the "new files" piece of that: a new Swift file no longer forces a rebuild of the view shell. In the 8,315-PR sample, 41% of the PRs that needed a rebuild added a file.

What changes

• Prototypes/ViewShell/inject.sh: besides the files named, the patch now links two kinds of file:

  • every Swift file of the module that the running Debug build never compiled (it has no object in ViewShell.build): a new file;
  • every file newer than that build's executable.

  It prints also X.swift (new since the build | changed since the build) for each one. So a new view in a new file, used from an edited file, links inside the patch. Each repeat patch holds all of the new code, so callers never keep an earlier patch's copy of the new file. main.swift and LiveCode.swift still refuse. With no Debug build at .build/debug (or CMUX_VIEW_SHELL_BUILD), the patch holds only the files named, as before, and the script says so.

• scripts/cmux-next/live_apply.py (try-pr's classifier): an added Swift file no longer counts as a rebuild. An added main.swift or @main is the exception and counts as new entry point. New types in existing files keep their rebuild reason.

• scripts/measure/live-code-shell.sh: a new step 3 writes LiveCodeProbeDot.swift, has Badge.swift use it, and injects naming only Badge.swift. The stored-property step becomes step 4. result.json gains new_file_edit and new_file_edit_to_pixels_s.

• Docs: the view shell README ("Live code") and docs/cmux-next/hot-reload.md.

Verification

• python3 -m unittest scripts/cmux-next/tests/test_live_apply.py: 12 tests pass, including the new-file test and the @main one.
• On a fleet Mac, scripts/measure/live-code-shell.sh was run at this head and at its base (4546c99a, which has the view shell live code: repoint already-bound symbol pointers, so a patch changes what draws #35850 pointer fix but the old inject.sh). Both ran on the same machine. Rendered badge widths at 2×:

step	base	this PR
0, the build	135 px	135 px
1, padding 24, red	199 px	199 px
2, padding 40	263 px	263 px
3, a new file LiveCodeProbeDot.swift used by Badge.swift; only Badge.swift named	refused: dlopen … symbol not found … LiveCodeProbeDot…	also LiveCodeProbeDot.swift (new since the build), 283 px with the dot, 1.26 s from edit to pixels
4, a new @State	refused	refused by the layout check

• A file counts as new when the Debug build has no object for it (ViewShell.build/<name>.swift.o). An earlier version read SwiftPM's sources list, but that list quotes a path with a space, which made every file look new.

Part of manaflow-ai/cmuxterm-hq#7066.

🤖 Generated with Claude Code
Agent session: agent session

[teamleaderleo]

Review (codex, router-review): SHIP at 31046ce

Reviewer: codex gpt-6-astra high, fast tier.

Follow-ups: 1. They land as checklist items in this PR's one follow-up tracking issue in this repo (Follow-ups: #N ..., label follow-up; add to an open one before filing, never one issue per finding), linking this receipt; the lane that opened the PR fixes them forward.

SHIP at 31046ce

Visual verdict: no images; Leo tries the build

Checked the focused unit tests: 12 pass.

Follow-ups:

• [P2] [follow-up] [category:correctness] [finding:F1] [outcome:pending] [Prototypes/ViewShell/inject.sh:41] The source glob only includes direct children of Sources/ViewShell. SwiftPM targets recurse into subdirectories, so a new Sources/ViewShell/Subdir/NewView.swift is neither compiled nor auto-added; an edited file referring to it fails injection. Use a recursive source enumeration.

[blacksmith-sh]

Found 1 test failure on Blacksmith runners:

Failure

Test	View Logs
test_every_listed_hook_is_one_glaeda_released_with_the_check (main.SigningHooksInSt epWithGlaedaTests.test_every_listed_hook_is_one_glaeda_released_with_the_check)/ test_every_listed_hook_is_one_glaeda_released_with_the_check (main.SigningHooksInSt epWithGlaedaTests.test_every_listed_hook_is_one_glaeda_released_with_the_check)	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[teamleaderleo]

Review (codex, router-review): SHIP at 476a2ad

Reviewer: codex gpt-6-astra high, fast tier.

Follow-ups: 2. They land as checklist items in this PR's one follow-up tracking issue in this repo (Follow-ups: #N ..., label follow-up; add to an open one before filing, never one issue per finding), linking this receipt; the lane that opened the PR fixes them forward.

SHIP at 476a2ad

Visual verdict: no images; Leo tries the build

• [P2] [follow-up] [category:correctness] [finding:F1] [outcome:pending] live_apply.py:42,56 only detects @main at the start of a line. A new file using valid combined attributes such as @MainActor @main is classified swift-live instead of rebuild.

• [P2] [follow-up] [category:ux] [finding:F2] [outcome:pending] No after screenshots were supplied, so the visual result remains unverified.

Focused live_apply tests pass.

Follow-ups:

• F1: broaden entry-point detection to recognize @main anywhere in the attribute sequence.
• F2: provide after-build screenshots.

[teamleaderleo]

Pictures of the new-file step (generation 3, the green dot from LiveCodeProbeDot.swift, 283 px) and the steps before it are on #35850: #35850 (comment). That run used this PR plus #35850.

Agent session: agent session

[github-actions]

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

Review 476a2ad: 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=pr35705 -f ref=lane/live-code-new-files -f platform=macos -f backend_mode=local

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4191131679101 (e01bbae5c9)

[cmux-bot]

cmux-bot approves 476a2ad: an independent review passed this exact head (router-review receipt by teamleaderleo). A push needs a new review; a later BLOCK at this head dismisses this approval.