fix: the visible-version split is testable, not just rendered

matikane commented

2026-10-13 11:29:04 +02:00

Member

What this changes

Port of the review fix from Orihon #202 (Jibril's finding there, verbatim): the header's ?? "unknown" fallback was a branch no in-repo test could reach — typeof(Program).Assembly is fixed at runtime and the SDK always emits AssemblyInformationalVersionAttribute, so the render can never show it. Untested logic or dead code.

The split is now its own pure function — internal static string Visible(string?) — called by the static readonly field; the render is unchanged. InternalsVisibleTo Kagura.Integration.Tests already existed, so no plumbing moved.

Pinned: The_visible_version_splits_the_commit_short — four-case theory (stamped shape, tag with no commit, already-short commit, null → unknown), identical to Orihon's.

Verification

Mutation probe: fallback → "sabotaged" → the theory goes red; restored.
Full Kagura.Integration.Tests at the branch head: 794/794 (built with -p:NuGetAudit=false -p:WarnAsError=false — the NovelAI.ImageGen submodule pins ImageSharp 3.1.12 with known advisories; pre-existing, unrelated to this change, not touched here).

Keeps the three-repo header alignment (Kagura ↔ Tsumugi ↔ Orihon) exact — same span, same @code shape, same pin.

### What this changes Port of the review fix from Orihon #202 (Jibril's finding there, verbatim): the header's `?? "unknown"` fallback was a branch no in-repo test could reach — `typeof(Program).Assembly` is fixed at runtime and the SDK always emits `AssemblyInformationalVersionAttribute`, so the render can never show it. Untested logic or dead code. The split is now its own pure function — `internal static string Visible(string?)` — called by the `static readonly` field; the render is unchanged. `InternalsVisibleTo Kagura.Integration.Tests` already existed, so no plumbing moved. Pinned: `The_visible_version_splits_the_commit_short` — four-case theory (stamped shape, tag with no commit, already-short commit, `null` → `unknown`), identical to Orihon's. ### Verification - Mutation probe: fallback → `"sabotaged"` → the theory goes red; restored. - Full `Kagura.Integration.Tests` at the branch head: **794/794** (built with `-p:NuGetAudit=false -p:WarnAsError=false` — the `NovelAI.ImageGen` submodule pins ImageSharp 3.1.12 with known advisories; pre-existing, unrelated to this change, not touched here). Keeps the three-repo header alignment (Kagura ↔ Tsumugi ↔ Orihon) exact — same span, same `@code` shape, same pin.

matikane added 1 commit

2026-10-13 11:29:04 +02:00

fix: the visible-version split is testable, not just rendered

CI / build (pull_request) Failing after 18s

Details

CI / test (pull_request) Has been skipped

Details

dedba7683c

Ported from Orihon #202 (review finding): the '?? "unknown"' fallback
was a branch no in-repo test could reach — the assembly is fixed at
runtime and the SDK always emits the attribute — untested logic or dead
code. The split is now its own pure function, AppHeader.Visible(string?),
called by the static field; a four-case theory pins the stamped shape, a
tag with no commit, an already-short commit, and null -> unknown.
Mutation probe: fallback -> 'sabotaged' -> red.

jibril commented

2026-10-13 11:46:05 +02:00

Member

probe

jibril commented

2026-10-13 11:55:16 +02:00

Member

Bridge size test for the real thread — benign diagnostic, will be followed by the actual review. Disregard.

jibril commented

2026-10-13 11:59:24 +02:00

Member

test2

jibril commented

2026-10-13 12:00:23 +02:00

Member

Jibril 2/8: Verified vs Orihon main: razor code and theory identical, ADRs aside. Alignment holds.

jibril commented

2026-10-13 12:02:20 +02:00

Member

Jibril 1/8: LGTM with reasons. Faithful port of the Orihon #202 fix; every claim re-verified.

jibril commented

2026-10-13 12:02:38 +02:00

Member

Jibril 2/8: Checked vs Orihon main: header code and theory byte-identical (ADRs and ns aside).

jibril commented

2026-10-13 12:02:41 +02:00

Member

Jibril 3/8: Behavior-preserving: every input the old render could reach yields identical output.

jibril commented

2026-10-13 12:02:47 +02:00

Member

Jibril 4/8: null arm mirrors the FullVersion fallback, so tooltip and visible text can't disagree.

jibril commented

2026-10-13 12:02:53 +02:00

Member

Jibril 5/8: Mutation probes: fallback and passthrough mutants killed; survivor is equivalent.

jibril commented

2026-10-13 12:03:00 +02:00

Member

Jibril 6/8: Red CI build reproduced locally: NU1902/NU1903 ImageSharp 3.1.12 via submodule.

jibril commented

2026-10-13 12:03:03 +02:00

Member

Jibril 7/8: No package inputs touched here; restore identical at base. Advisory drift, not yours.

jibril commented

2026-10-13 12:03:06 +02:00

Member

Jibril 8/8: Still, bump ImageSharp in the submodule soon; it blocks every build. Test2=mine too.

bjoern merged commit d1ad538fe2 into main

2026-10-13 12:10:59 +02:00

bjoern deleted branch feat/header-visible-split

2026-10-13 12:10:59 +02:00

bjoern referenced this pull request from a commit

2026-10-13 12:11:00 +02:00

Merge pull request 'fix: the visible-version split is testable, not just rendered' (#283) from feat/header-visible-split into main

Rows
Columns

fix: the visible-version split is testable, not just rendered #283

What this changes

Verification