Skip to content

Bring the documentation up to the first release, and measure the update path - #4

Merged
geeooff merged 1 commit into
mainfrom
docs-after-first-release
Sep 18, 2026
Merged

geeooff merged 1 commit into
mainfrom
docs-after-first-release

Conversation

@geeooff

@geeooff geeooff commented Sep 18, 2026

Copy link
Copy Markdown
Owner

Summary

A pass over every Markdown page and the comments touched since 2026-09-17, after v0.1.0 went out. No code change; three comments, one AGENTS.md paragraph, the design record and the user pages.

  • Lot 8 closed in the design index and its record, with the second machine's first install as the one field check still outstanding.
  • A correction in place: the record claimed the summary stream's "no elevation" bit alone makes the package per-user and that ALLUSERS / MSIINSTALLPERUSER were unnecessary. That was the probe's reading; the real package measured the opposite (bit alone → C:\Program Files (x86), ALLUSERS=2 alone → per-machine, 1603) and the measurement had only lived in a comment in msi.ps1. Both readings now stand in the record, dated.
  • The recipes ended with install-task after copying a configuration. For an installed copy the task exists and the watcher is already running, so it kept the old file until the next logon. They now say stop then install-task, as Getting started does.
  • AGENTS.md says how a change is verified on a machine the installer owns (stop, copy, install-task; uninstall/reinstall for a package change, since the product code is fixed per version).
  • Lot 13's page records what v0.1.0 actually answers — the 302 carrying the tag, SHA256SUMS.txt by tag, the asset chain, the API's digest agreeing with the workflow's hash — and what the package now does for the updater: it stops and restarts the watcher itself, so the earlier quit-then-relaunch plan is superseded and the updater keeps only the failure path. Still a proposal; nothing built.
  • Smaller: Turning it off points at removal, How it works tells the installer's location from the zip's, stop joins the list of commands that log under setup, and three comments no longer describe the console binary as the only one running the setup commands.

Test plan

  • .\scripts\build.ps1 test — 112 + 3 tests, 24 pages, every link resolves
  • CI

🤖 Generated with Claude Code

…te path

A pass over every page and the comments the last two days touched. Lot 8
is closed in the index and its record, with the second machine's install
as the one field check still outstanding; the record's claim that the
summary bit alone makes a package per-user is corrected in place with
the real package's measurement. The recipes told an installer user to
run install-task after copying a configuration, which keeps a running
watcher on the old file: they say stop then install-task now, as
getting-started does. AGENTS.md says how a change is verified on a
machine the installer owns. Comments in task.rs, the twin and msi.ps1
no longer describe the console binary as the only one that runs the
setup commands.

Lot 13's page records what v0.1.0 actually answers -- the redirect that
carries the tag, the checksum file, the asset chain, the API's digest --
and what the package now does for the updater: it stops and restarts
the watcher itself, so the earlier quit-then-relaunch plan is
superseded, and the updater keeps only the failure path.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@geeooff
geeooff merged commit 3a32c62 into main Sep 18, 2026
1 check passed
@geeooff
geeooff deleted the docs-after-first-release branch September 18, 2026 09:39
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant