The same binary run twice against the same .xcresult produces different HTML.
run A: selectDevice('6CBE1E1A-E712-42AD-BAA8-A179C1823B56')
run B: selectDevice('E2105793-8D30-4944-AF86-2BBC4017C91F')
Byte counts are identical (UUIDs are fixed width); the content differs.
Cause
Five call sites mint a fresh UUID() at parse time, used as DOM element ids and JS handles:
Models/TestSummary.swift:51 — uuid = UUID().uuidStringModels/TargetDevice.swift:21 — uniqueIdentifier = UUID().uuidStringModels/Test.swift:58 and :184 — let uuid = UUID().uuidStringModels/Iteration.swift:12 — let uuid = UUID().uuidString
Why it matters
Rendering is unaffected — these are internal handles, not user-visible. But it means:
- Two reports from the same bundle cannot be diffed to see what actually changed
- Reports cannot be content-addressed or cached by hash
- Any future golden-file testing is impossible without normalising them first
TargetDevice is the clearest case: record.identifier is already read from the bundle on the line above, and then a random uniqueIdentifier is generated alongside it.
Suggested fix
Derive these identifiers deterministically from bundle content rather than randomly — the device identifier, the test's full name, the iteration index. Anything stable across runs for the same input.
Found while verifying that a SwiftFormat reformat had not changed generated output (#410). It hadn't; this had been true all along.
The same binary run twice against the same
.xcresultproduces different HTML.Byte counts are identical (UUIDs are fixed width); the content differs.
Cause
Five call sites mint a fresh
UUID()at parse time, used as DOM element ids and JS handles:Models/TestSummary.swift:51—uuid = UUID().uuidStringModels/TargetDevice.swift:21—uniqueIdentifier = UUID().uuidStringModels/Test.swift:58and:184—let uuid = UUID().uuidStringModels/Iteration.swift:12—let uuid = UUID().uuidStringWhy it matters
Rendering is unaffected — these are internal handles, not user-visible. But it means:
TargetDeviceis the clearest case:record.identifieris already read from the bundle on the line above, and then a randomuniqueIdentifieris generated alongside it.Suggested fix
Derive these identifiers deterministically from bundle content rather than randomly — the device identifier, the test's full name, the iteration index. Anything stable across runs for the same input.
Found while verifying that a SwiftFormat reformat had not changed generated output (#410). It hadn't; this had been true all along.