C64-dev/newLang/planning/claims/E07-S02.md

---

story: E07-S02 owner: codebuff status: done started: 2026-11-10 completed: 2026-11-10 depends-on: [E07-S01, E04-S01]

E07-S02 Claim

Intended change

Two launchers, one shared implementation: scripts/launch-editor.ps1 and scripts/launch-compiler.ps1. Both take an exclusive lock beside the writable image, wait for the emulator to exit and the image to flush, and refuse a second writable session. D151 moves the editor to the target machine: the launcher copies ../vim/vim-improved.d81, writes ../vim/vim128.prg into the copy as vim128, starts x128 in 80-column mode, and autostarts the editor from that disk. docs/editor-workflow.md states the order, stale-lock recovery and the one-writable-session rule.

File area

• scripts/lib/Nova128Launcher.psm1 (new): session lock, flush wait, Add-Nova128DiskFile, Invoke-Nova128Emulator (80-column and disk-autostart support, VDC/VIC-II screenshot choice)
• scripts/launch-editor.ps1, scripts/launch-compiler.ps1 (new)
• tests/disk/launchers.tests.ps1 (new)
• docs/editor-workflow.md (new), docs/decisions.md (D148, D151), docs/architecture.md (§10)
• planning/claims/E07-S02.md
• generated files under build/

Acceptance evidence

• Editor launcher uses x128 in 80-column mode and autostarts the C128 Vim: scripts/launch-editor.ps1 defaults to ../vim/vim128.prg, copies ../vim/vim-improved.d81 to build/newlang-edit.d81, writes the editor into the copy as vim128, and autostarts work.d81:vim128 in x128 -80col from the image's folder. tests/disk/launchers.tests.ps1 runs it against a copy of the fixture disk, asserts the emulator exits, the session lock is released, the image stays writable, and vim128 is on the session disk.
• Compiler launcher uses x128 and autostarts the compiler: it defaults to build/n128c.prg, prepares build/newlang-dev.d81 (copy plus injection) unless -SkipPrepare, and autostarts it in x128; the test exercises the same sequence with -SkipPrepare on a disposable image.
• Launchers wait for process/image flush before returning: the shared Exit-Nova128ImageSession waits for emulator exit and Wait-Nova128ImageFlush reopens the image exclusively with retries; the test holds the lock and asserts refusal, then asserts the image is reopenable immediately after each launcher returns.
• Documentation and warnings prohibit simultaneous writable attachment: docs/editor-workflow.md and the launcher banners state that only one writable session may exist and that the previous session must be closed before editing or inspecting the disk; the lock enforces it.

Coordination notes

• Depends on E07-S01 (image preparation) and E04-S01 (container/autostart); both are complete. The compiler launcher reuses prepare-disk.ps1.
• D151 changed the editor default from x64sc to the C128 build at the user's request. The C64 editor is still reachable with -Emulator x64sc, but its program lives inside vim-improved.d81, so -EditorPrg must name an extracted vim.
• Autostarting the editor from the session disk is required, not cosmetic: a host-loaded copy has no device number from its load, and vim128 then reports written while the image never receives the file.
• VICE's x64sc plain -exitscreenshot is the VIC-II screen; x128 in 80-column mode needs the same plain option (VDC), while -exitscreenshotvicii captures a blank 40-column screen.

Work log

• The lock test first failed with “cannot access the lock file” instead of a lock message: a live holder opens the file without sharing, so reading it can fail. An unreadable lock now counts as held.
• The screenshot assertion failed for x64sc because -exitscreenshotvicii is the x128 spelling; the option is now chosen by emulator and display mode.
• The compiler test passed no disk, so the launcher correctly refused with “disk image not found”; the test now copies its source image first.
• D151: the C128 editor was validated at the prompt, then its saves failed in three distinct ways (1541 drive type, true-drive emulation, host autostart) before the disk-autostart form worked.

Completion

• Tests run: tests/disk/launchers.tests.ps1, tests/disk/vim-roundtrip.tests.ps1, git diff --check.
• Results: launcher tests passed covering the disk copy with vim128 injected, lock refusal, flush wait, and both x128 sessions; the source disk stayed 0FF83A621B59B312F1BAE9F1D0785421500D22FD00F833A9CA918EBCA7105983.
• Remaining risks: the lock is advisory and PID-based, so a recycled PID in the window before the lock is read is treated as alive; a launcher killed from outside leaves a lock for the next run to replace; and the flush wait assumes VICE exits cleanly, which cannot be forced if the process is killed.