brew-cask: AeroSpace tap Ruby metadata evaluation rejects postflight_steps on 2026.9.4
#13053
ProblemOn mise 2026.9.4, a dry-run for Expected: resolve metadata and produce a dry-run plan. Actual: evaluation fails before installation or lifecycle execution. Environment
Tested with isolated HOME and mise config/data/cache/state under ReproductionMinimal [settings]
experimental = true
[tools]
ruby = "4.0.6"
[bootstrap.packages]
"brew-cask:nikitabobko/tap/aerospace" = { version = "latest", os = "macos" }With Ruby 4.0.6 already available on ruby --version
mise --version
mise --verbose bootstrap packages apply --dry-run brew-cask:nikitabobko/tap/aerospaceThe command exited 1 on both runs, with the error above; the first used a fresh cache. The verbose log confirms API JSON returned 404, followed by successful retrieval of AeroSpace Reduced metadata-only testUnder the same sandbox profile, the unmodified v2026.9.4 metadata shim, verified against the embedded shim, fails on this fixture: cask "aerospace" do
version "1.0"
sha256 "0000000000000000000000000000000000000000000000000000000000000000"
url "https://example.invalid/aerospace.zip"
postflight_steps do
end
endRemoving only the empty The shim defines Related work and question
Is representing structured
|
Replies: 1 comment
|
Opened #13060 to address the structured flight-step metadata gap. The tap metadata shim now collects The focused regression tests pass, including sandboxed metadata extraction with Ruby 4.0.6, and One limitation surfaced when checking the complete cask pinned in this report: with this patch, evaluation advances beyond AI-assisted — Tool: Codex; model: openai/unavailable; version: unavailable. |
Opened #13060 to address the structured flight-step metadata gap.
The tap metadata shim now collects
preflight_steps/postflight_stepscontainingruncommands into structured artifacts without executing the commands. It preserves version interpolation and path placeholders. The Rust runner also honors the explicitmust_succeed: falseused by the two AeroSpace commands; ordinary lifecycle evaluation skips structured blocks so they are not run twice.The focused regression tests pass, including sandboxed metadata extraction with Ruby 4.0.6, and
mise run lint-fix/ Cargo check pass.One limitation surfaced when checking the complete cask pinned in this report: with this patch, evaluation ad…