Skip to content

ADR 0005: the test corpus and the published examples are separate artifacts - #49

Merged
DerekCorniello merged 2 commits into
mainfrom
docs/adr-0005-examples-corpus
Aug 14, 2026
Merged

ADR 0005: the test corpus and the published examples are separate artifacts#49
DerekCorniello merged 2 commits into
mainfrom
docs/adr-0005-examples-corpus

Conversation

@DerekCorniello

Copy link
Copy Markdown
Contributor

Decides #44, which blocked #26.

Records that mux-compiler/test_scripts/ and mux-examples are two artifacts and that the corpus does not move.

The three findings that settled it, with the numbers, so it is not re-argued from intuition later:

  • Incompatible pins. A regression corpus must track main; published examples must track a release or they teach syntax the playground rejects. No single corpus satisfies both.
  • Coupling. 36 of 36 commits touching test_scripts/ in 90 days also touched mux-compiler/src/.
  • The fixtures are not examples. They pin behaviour rather than teach - one opens with 30 lines on *mut Value slots and capture cells.

Drift, the main argument for merging, had not materialised: 9 of 224 named website snippets shared a filename with a fixture, and those were deliberately different.

The consequences section is deliberately unflattering about the costs: two corpora that will diverge on purpose, mux-compiler taking a CI dependency on mux-examples, and no release-level guarantee for the examples until the release jobs exist. It also records that authoring the first corpus found nine defects, which is the strongest argument for the repo existing.

Implemented by muxlang/mux-examples#1.

Decides #44, which blocked #26.

Three things settled it, and the numbers are in the record so it does not get
re-argued from intuition: the two artifacts need incompatible compiler pins (a
regression corpus must track main, published examples must track a release),
36 of 36 commits touching test_scripts in 90 days also touched compiler source,
and the fixtures are internals notes rather than teaching material.

The consequences section is explicit about what this costs - two corpora that
will diverge, and no release-level guarantee for the examples until the release
jobs are written - so the follow-up work is visible rather than implied.
@greptile-apps

greptile-apps Bot commented Aug 14, 2026

Copy link
Copy Markdown

Greptile Summary

Records the decision to keep the compiler regression corpus separate from the published examples repository.

  • Documents the distinct purposes, compiler-version requirements, and maintenance models of the two artifacts.
  • Adds ADR 0005 to the decisions index.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename Overview
docs/decisions/0005-test-corpus-and-examples-are-separate.md Adds ADR 0005 with supporting evidence, the selected artifact boundary, and its operational consequences.
docs/decisions/README.md Adds ADR 0005 to the architecture-decisions index.

Reviews (2): Last reviewed commit: "Qualify every cross-repo issue reference..." | Re-trigger Greptile

Comment thread docs/decisions/0005-test-corpus-and-examples-are-separate.md
Comment thread docs/decisions/0005-test-corpus-and-examples-are-separate.md Outdated
A bare '#51' after 'muxlang/mux-runtime#50' does not inherit the qualifier - it
resolves against mux-context, so most of the defects cited as evidence pointed
at the wrong repo. They are the argument for the examples repo existing, so
they have to be followable.

Raised by Greptile on #49.
@sonarqubecloud

Copy link
Copy Markdown

@DerekCorniello
DerekCorniello merged commit 6c6a89d into main Aug 14, 2026
4 checks passed
@DerekCorniello
DerekCorniello deleted the docs/adr-0005-examples-corpus branch August 14, 2026 14:25
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