Skip to content

docs: record the promotion as finished, and the rough edges it left - #86

Merged
hanthor merged 1 commit into
ptyxis-portfrom
docs/promotion-complete
Sep 6, 2026
Merged

docs: record the promotion as finished, and the rough edges it left#86
hanthor merged 1 commit into
ptyxis-portfrom
docs/promotion-complete

Conversation

@hanthor

@hanthor hanthor commented Sep 6, 2026

Copy link
Copy Markdown
Member

Everything docs/TUNA_OS_PROMOTION.md was tracking is now real:

One prediction in the doc was wrong

It warned that ghcr.io/tuna-os/ghostty would be created private and need a manual visibility flip. It came out public on first push — verified by an anonymous GHCR token pull returning 200 for latest, latest-x86_64 and latest-aarch64.

Corrected as "re-check rather than assume" rather than deleted: a private package and a missing one fail identically, so the next app is still worth checking.

New: rough edges found while finishing this

Three real problems, all out of scope, recorded so they aren't rediscovered:

  • Test can never pass in this repo. Upstream's test.yml runs on namespace-profile-ghostty-* runners — Namespace.so machines the tuna-os org doesn't have. Those jobs queue indefinitely, so "Required Checks: Test" is permanently pending on every PR and cannot function as a merge gate. ptyxis-tests is the suite that actually runs.
  • publish-flatpak.yml is still a vendored copy of the org pipeline, outside tuna-os/.github's drift check. Tracked in Migrate publish-flatpak.yml onto the shared tuna-os pipeline #85.
  • ghcr.io/tuna-os/blueshell has no plain :latest tag — the vendored workflow pushes only latest-<arch>. Harmless for flatpak, but a plain pull of that path 404s while ghcr.io/tuna-os/ghostty:latest works.

🤖 Generated with Claude Code

https://claude.ai/code/session_01EMBRswEeeX2kzVSwpbxCoy

Everything this runbook was tracking is now real: Ghostty publishes to
the remote, both apps install from it, the site lists them, and
expected-apps.json turns the sanity check from a warning into a gate.
Checks off what happened and links where.

One prediction in here was wrong and is corrected rather than quietly
dropped: the GHCR package came out public on first push, so the manual
visibility flip this document warned about was not needed. Recorded as
"re-check rather than assume" instead of "not a problem", since a
private package and a missing one fail identically and the next app may
land the other way.

Adds a rough-edges section for three things found while finishing this
that are real but out of scope: Test can never pass here (upstream's
namespace-profile-ghostty-* runners don't exist in this org, so those
jobs queue forever and cannot gate a merge), publish-flatpak.yml is
still a vendored copy outside the org's drift check, and
ghcr.io/tuna-os/blueshell has no plain :latest tag.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EMBRswEeeX2kzVSwpbxCoy
@hanthor
hanthor merged commit 393a2ea into ptyxis-port Sep 6, 2026
97 of 100 checks passed
@hanthor
hanthor deleted the docs/promotion-complete branch September 6, 2026 08:58
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