Line coverage: 50% → 78% (4,519 of 5,787 lines). Verified via a clean :app:jacocoTestReport run (deleted app/build/reports/jacoco first, reran from scratch) — see the "78% vs. the ~85-90% ballpark" note below for why this lands under the issue's estimate, on purpose.
ViewModels, repos, mappers, DAOs, utils, policies, richtext, mail, and reporting-logic are all still counted — spot-checked the per-package HTML (AccountSettingsViewModel, SettingsViewModel, ReaderViewModel, MailboxViewModel, AppLockViewModel, MailRepositoryImpl, AccountRepositoryImpl, ReportStore, ReportSubmitter, DiagnosticsCollector, etc. all present) and confirmed org.libremail.di / org.libremail.data.local.coldopen no longer appear as packages in the report at all.
Exclusion globs added
valnonJvmTestableSurface=listOf(// --- Compose UI render code: one glob per screen/component file (see the exceptions below) ---
"**/LibreMailApp*","**/AccountPickerScreen*","**/AppPasswordSetupScreen*","**/ManualSetupScreen*","**/ComposeScreen*","**/ColorSwatch*","**/FontPicker*","**/FontSizePicker*","**/ParagraphAlignmentControl*","**/DraftsScreen*","**/LockScreen*","**/AppLockGateHost*","**/FolderDrawer*","**/MailboxScreen*","**/AddAnotherAccountScreen*","**/BatteryOptimizationScreen*","**/ContactsAccessScreen*","**/LicenseScreen*","**/OnboardingWelcomeScreen*","**/OutboxScreen*","**/ReaderScreen*","**/ProblemReportsScreen*","**/AccountSettingsScreen*","**/SettingsScreen*","**/SettingsComponents*","**/SignatureEditScreen*","**/SignaturesScreen*",// --- Android framework entry points ---
"**/*Activity*","**/*Service*","**/*Worker*","**/LibreMailApplication*","**/*BackupAgent*",// --- Hilt DI wiring ---
"**/di/**",// --- src/debug cold-open probe (issue #221) ---
"**/data/local/coldopen/**",)
Two things worth your review
1. Four Screen/component files were deliberately left OUT of the exclusion list, even though they live in packages the issue calls out (ui/compose, ui/settings, ui/reader, ui/reporting):
ui/settings/AccountReorderList.kt — AccountReorderListTest covers commitDrag's reorder maths ("Pure so the index maths is unit-testable without a gesture" per its own kdoc).
ui/reporting/ReportReviewScreen.kt — ReportReviewClipboardTest covers copyReportPayloadToClipboard ("Kept as a plain suspend function... unit-testable... without a Compose UI test or emulator" per its own kdoc).
JaCoCo excludes at class-file granularity, and Kotlin compiles every top-level function in a .kt file — @Composable or not — into the same facade class. Excluding these four files would have zeroed out those five dedicated test files' coverage along with the render code. I verified this concretely: temporarily adding these four to the exclusion list brought line coverage to 87% (matching the issue's ~85-90% estimate almost exactly), which is strong evidence the original audit's estimate assumed a blanket **/ui/**-style exclusion that didn't catch this mixed-file wrinkle. I kept them in scope and am reporting the more conservative 78% instead — happy to go the other way if you'd rather hit the estimate and accept losing that coverage signal (or if you'd prefer a third option: split each file so the pure functions live outside the render file, letting both goals be satisfied at once — that's more invasive than this ticket's scope, so I didn't do it here).
2. **/*Worker* removes coverage that is currently real, not just glue. All six Workers it excludes — SyncWorker, BackfillWorker, PruneWorker, SendWorker, ReportPurgeWorker, ReportUploadWorker — are directly unit-tested today (construct the worker with mocked collaborators, call doWork(), assert the Result): SyncWorkerTest, BackfillWorkerTest, PruneWorkerTest, SendWorkerTest (272 lines testing a 182-line class), ReportPurgeWorkerTest, ReportUploadWorkerTest/ReportUploadWorkerHttpTest. I implemented the exclusion exactly as specified in the issue (framework entry points, by category, regardless of current test investment), but flagging it since it's a real, measurable tradeoff rather than a clear-cut "untestable glue" case like MainActivity/IdleService/LibreMailBackupAgent (all zero JVM tests, no ambiguity there).
Validation
:app:jacocoTestReport — clean, HTML/XML regenerated from scratch.
:app:ktlintCheck :app:detekt — green.
Build-script-only change (no source/test files touched), so per the ticket's own "no emulator" validation scope I did not run the full preflight/emulator gate.
Closes #290
## Result
Line coverage: **50% → 78%** (4,519 of 5,787 lines). Verified via a clean `:app:jacocoTestReport` run (deleted `app/build/reports/jacoco` first, reran from scratch) — see the "78% vs. the ~85-90% ballpark" note below for why this lands under the issue's estimate, on purpose.
ViewModels, repos, mappers, DAOs, utils, policies, richtext, mail, and reporting-logic are all still counted — spot-checked the per-package HTML (`AccountSettingsViewModel`, `SettingsViewModel`, `ReaderViewModel`, `MailboxViewModel`, `AppLockViewModel`, `MailRepositoryImpl`, `AccountRepositoryImpl`, `ReportStore`, `ReportSubmitter`, `DiagnosticsCollector`, etc. all present) and confirmed `org.libremail.di` / `org.libremail.data.local.coldopen` no longer appear as packages in the report at all.
## Exclusion globs added
```kotlin
val nonJvmTestableSurface = listOf(
// --- Compose UI render code: one glob per screen/component file (see the exceptions below) ---
"**/LibreMailApp*", "**/AccountPickerScreen*", "**/AppPasswordSetupScreen*",
"**/ManualSetupScreen*", "**/ComposeScreen*", "**/ColorSwatch*", "**/FontPicker*",
"**/FontSizePicker*", "**/ParagraphAlignmentControl*", "**/DraftsScreen*", "**/LockScreen*",
"**/AppLockGateHost*", "**/FolderDrawer*", "**/MailboxScreen*", "**/AddAnotherAccountScreen*",
"**/BatteryOptimizationScreen*", "**/ContactsAccessScreen*", "**/LicenseScreen*",
"**/OnboardingWelcomeScreen*", "**/OutboxScreen*", "**/ReaderScreen*",
"**/ProblemReportsScreen*", "**/AccountSettingsScreen*", "**/SettingsScreen*",
"**/SettingsComponents*", "**/SignatureEditScreen*", "**/SignaturesScreen*",
// --- Android framework entry points ---
"**/*Activity*", "**/*Service*", "**/*Worker*", "**/LibreMailApplication*", "**/*BackupAgent*",
// --- Hilt DI wiring ---
"**/di/**",
// --- src/debug cold-open probe (issue #221) ---
"**/data/local/coldopen/**",
)
```
## Two things worth your review
**1. Four Screen/component files were deliberately left OUT of the exclusion list**, even though they live in packages the issue calls out (`ui/compose`, `ui/settings`, `ui/reader`, `ui/reporting`):
- `ui/compose/RichTextEditor.kt` — `RichTextEditorTest` (17 tests) covers the `AnnotatedString`<->`RichTextContent` editor-op functions (`applyStyle`/`applyBlock`/`applyLink`/`toRichContent`/`toAnnotatedString`/...).
- `ui/settings/AccountReorderList.kt` — `AccountReorderListTest` covers `commitDrag`'s reorder maths ("Pure so the index maths is unit-testable without a gesture" per its own kdoc).
- `ui/reader/HtmlBody.kt` — `HtmlBodyTest` + `InlineImageResolverTest` cover `cidKey`/`resolveInlineImage`/`wrapHtml`/`toCssHex`.
- `ui/reporting/ReportReviewScreen.kt` — `ReportReviewClipboardTest` covers `copyReportPayloadToClipboard` ("Kept as a plain suspend function... unit-testable... without a Compose UI test or emulator" per its own kdoc).
JaCoCo excludes at class-file granularity, and Kotlin compiles every top-level function in a `.kt` file — `@Composable` or not — into the same facade class. Excluding these four files would have zeroed out those five dedicated test files' coverage along with the render code. I verified this concretely: temporarily adding these four to the exclusion list brought line coverage to **87%** (matching the issue's ~85-90% estimate almost exactly), which is strong evidence the original audit's estimate assumed a blanket `**/ui/**`-style exclusion that didn't catch this mixed-file wrinkle. I kept them in scope and am reporting the more conservative 78% instead — happy to go the other way if you'd rather hit the estimate and accept losing that coverage signal (or if you'd prefer a third option: split each file so the pure functions live outside the render file, letting both goals be satisfied at once — that's more invasive than this ticket's scope, so I didn't do it here).
**2. `**/*Worker*` removes coverage that is currently real, not just glue.** All six Workers it excludes — `SyncWorker`, `BackfillWorker`, `PruneWorker`, `SendWorker`, `ReportPurgeWorker`, `ReportUploadWorker` — are directly unit-tested today (construct the worker with mocked collaborators, call `doWork()`, assert the `Result`): `SyncWorkerTest`, `BackfillWorkerTest`, `PruneWorkerTest`, `SendWorkerTest` (272 lines testing a 182-line class), `ReportPurgeWorkerTest`, `ReportUploadWorkerTest`/`ReportUploadWorkerHttpTest`. I implemented the exclusion exactly as specified in the issue (framework entry points, by category, regardless of current test investment), but flagging it since it's a real, measurable tradeoff rather than a clear-cut "untestable glue" case like `MainActivity`/`IdleService`/`LibreMailBackupAgent` (all zero JVM tests, no ambiguity there).
## Validation
- `:app:jacocoTestReport` — clean, HTML/XML regenerated from scratch.
- `:app:ktlintCheck :app:detekt` — green.
- Build-script-only change (no source/test files touched), so per the ticket's own "no emulator" validation scope I did not run the full preflight/emulator gate.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Superseded by #323, which incorporates this scoping, keeps the unit-tested Workers in-scope (fixing this PR's over-broad Worker exclusion), and adds the no-regression CI gate (#251).
Superseded by #323, which incorporates this scoping, keeps the unit-tested Workers in-scope (fixing this PR's over-broad Worker exclusion), and adds the no-regression CI gate (#251).
Pull request closed
Please reopen this pull request to perform a merge.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Closes #290
Result
Line coverage: 50% → 78% (4,519 of 5,787 lines). Verified via a clean
:app:jacocoTestReportrun (deletedapp/build/reports/jacocofirst, reran from scratch) — see the "78% vs. the ~85-90% ballpark" note below for why this lands under the issue's estimate, on purpose.ViewModels, repos, mappers, DAOs, utils, policies, richtext, mail, and reporting-logic are all still counted — spot-checked the per-package HTML (
AccountSettingsViewModel,SettingsViewModel,ReaderViewModel,MailboxViewModel,AppLockViewModel,MailRepositoryImpl,AccountRepositoryImpl,ReportStore,ReportSubmitter,DiagnosticsCollector, etc. all present) and confirmedorg.libremail.di/org.libremail.data.local.coldopenno longer appear as packages in the report at all.Exclusion globs added
Two things worth your review
1. Four Screen/component files were deliberately left OUT of the exclusion list, even though they live in packages the issue calls out (
ui/compose,ui/settings,ui/reader,ui/reporting):ui/compose/RichTextEditor.kt—RichTextEditorTest(17 tests) covers theAnnotatedString<->RichTextContenteditor-op functions (applyStyle/applyBlock/applyLink/toRichContent/toAnnotatedString/...).ui/settings/AccountReorderList.kt—AccountReorderListTestcoverscommitDrag's reorder maths ("Pure so the index maths is unit-testable without a gesture" per its own kdoc).ui/reader/HtmlBody.kt—HtmlBodyTest+InlineImageResolverTestcovercidKey/resolveInlineImage/wrapHtml/toCssHex.ui/reporting/ReportReviewScreen.kt—ReportReviewClipboardTestcoverscopyReportPayloadToClipboard("Kept as a plain suspend function... unit-testable... without a Compose UI test or emulator" per its own kdoc).JaCoCo excludes at class-file granularity, and Kotlin compiles every top-level function in a
.ktfile —@Composableor not — into the same facade class. Excluding these four files would have zeroed out those five dedicated test files' coverage along with the render code. I verified this concretely: temporarily adding these four to the exclusion list brought line coverage to 87% (matching the issue's ~85-90% estimate almost exactly), which is strong evidence the original audit's estimate assumed a blanket**/ui/**-style exclusion that didn't catch this mixed-file wrinkle. I kept them in scope and am reporting the more conservative 78% instead — happy to go the other way if you'd rather hit the estimate and accept losing that coverage signal (or if you'd prefer a third option: split each file so the pure functions live outside the render file, letting both goals be satisfied at once — that's more invasive than this ticket's scope, so I didn't do it here).2.
**/*Worker*removes coverage that is currently real, not just glue. All six Workers it excludes —SyncWorker,BackfillWorker,PruneWorker,SendWorker,ReportPurgeWorker,ReportUploadWorker— are directly unit-tested today (construct the worker with mocked collaborators, calldoWork(), assert theResult):SyncWorkerTest,BackfillWorkerTest,PruneWorkerTest,SendWorkerTest(272 lines testing a 182-line class),ReportPurgeWorkerTest,ReportUploadWorkerTest/ReportUploadWorkerHttpTest. I implemented the exclusion exactly as specified in the issue (framework entry points, by category, regardless of current test investment), but flagging it since it's a real, measurable tradeoff rather than a clear-cut "untestable glue" case likeMainActivity/IdleService/LibreMailBackupAgent(all zero JVM tests, no ambiguity there).Validation
:app:jacocoTestReport— clean, HTML/XML regenerated from scratch.:app:ktlintCheck :app:detekt— green.🤖 Generated with Claude Code
Superseded by #323, which incorporates this scoping, keeps the unit-tested Workers in-scope (fixing this PR's over-broad Worker exclusion), and adds the no-regression CI gate (#251).
Pull request closed