Introduce tiered CI depth and compute profiles (#85) #91

Merged
alexander merged 12 commits from mcp/issue-85-ci-tiers into main 2026-09-24 15:35:52 +02:00
Owner

Summary

Implements the first staged slice of #85 from current main.

  • Adds independent verification depth: Quick / Standard / Intensive.
  • Adds Quiet / Balanced / Fast Gradle worker profiles (2 / 4 / 6) without changing evidence for a given depth.
  • Makes Quick a true cheap KtLint gate that blocks the expensive graph on failure.
  • Keeps Standard on the canonical full verifyDebug --continue graph, excluding only assembleDebug.
  • Separates build/classpath changes from CI-infrastructure changes for --rerun-tasks.
  • Scopes the standalone dependency report to dependency-input changes / Intensive.
  • Scans PR-introduced commits with Gitleaks after ancestry validation.
  • Adds lightweight structured CI telemetry and classifier/telemetry tests.
  • Publishes separate Quick / Standard / Intensive status contexts.

Requested correction

Manual Standard/Intensive runs now have Build APK after checks pass.

The APK step:

  • defaults off;
  • cannot be selected for Quick;
  • executes only after the selected verification gate succeeds;
  • runs :app:assembleDebug after verification;
  • uploads the resulting debug APK;
  • does not expose release/upload signing secrets.

Thus ordinary Standard verification still pays no APK-packaging cost.

verifyDebug coverage check

The latest successful baseline run (#274) ran verifyDebug in 5m13s and still reported:

  • 2,621 JVM/Robolectric tests, 0 failed;
  • KtLint clean;
  • type-resolved Detekt clean;
  • Android Lint 0 errors / 0 warnings;
  • Room schemas current;
  • dependency resolution clean;
  • existing Kover coverage gate.

Gradle reported 1,121 actionable tasks: 506 executed, 157 from cache, 458 up-to-date. The improvement is cache/incremental reuse, not a reduced regression suite.

This PR deliberately does not yet add the later #85 Kover/coverage ratchet slice: debug-only Kover variant scoping, branch-coverage measurement, playback/download non-regression floors and mutation work remain follow-up stages in #85.

Base

Created directly from current main 599962e990b3c4a40d9bc2ac7a530358bb6922d2; branch was 0 commits behind main when opened.

Closes no issue yet; #85 remains the staging issue for later slices.

## Summary Implements the first staged slice of #85 from current main. - Adds independent verification depth: Quick / Standard / Intensive. - Adds Quiet / Balanced / Fast Gradle worker profiles (2 / 4 / 6) without changing evidence for a given depth. - Makes Quick a true cheap KtLint gate that blocks the expensive graph on failure. - Keeps Standard on the canonical full `verifyDebug --continue` graph, excluding only `assembleDebug`. - Separates build/classpath changes from CI-infrastructure changes for `--rerun-tasks`. - Scopes the standalone dependency report to dependency-input changes / Intensive. - Scans PR-introduced commits with Gitleaks after ancestry validation. - Adds lightweight structured CI telemetry and classifier/telemetry tests. - Publishes separate Quick / Standard / Intensive status contexts. ### Requested correction Manual Standard/Intensive runs now have **Build APK after checks pass**. The APK step: - defaults off; - cannot be selected for Quick; - executes only after the selected verification gate succeeds; - runs `:app:assembleDebug` after verification; - uploads the resulting debug APK; - does not expose release/upload signing secrets. Thus ordinary Standard verification still pays no APK-packaging cost. ## verifyDebug coverage check The latest successful baseline run (#274) ran `verifyDebug` in 5m13s and still reported: - 2,621 JVM/Robolectric tests, 0 failed; - KtLint clean; - type-resolved Detekt clean; - Android Lint 0 errors / 0 warnings; - Room schemas current; - dependency resolution clean; - existing Kover coverage gate. Gradle reported 1,121 actionable tasks: 506 executed, 157 from cache, 458 up-to-date. The improvement is cache/incremental reuse, not a reduced regression suite. This PR deliberately does **not** yet add the later #85 Kover/coverage ratchet slice: debug-only Kover variant scoping, branch-coverage measurement, playback/download non-regression floors and mutation work remain follow-up stages in #85. ## Base Created directly from current main `599962e990b3c4a40d9bc2ac7a530358bb6922d2`; branch was 0 commits behind main when opened. Closes no issue yet; #85 remains the staging issue for later slices.
Document tiered Forgejo verification
All checks were successful
CI · PR verification / CI · 91 · resolve target (pull_request) Successful in 1s
CI · PR verification / CI · · Preflight · wf:91 (pull_request) Successful in 8s
CI · PR verification / CI · · verification · wf:91 (pull_request) Successful in 1m24s
BookWave / PR quick BookWave quick verification passed
CI · PR verification / CI · · publish check · wf:91 (pull_request) Successful in 1s
BookWave / PR verification BookWave standard verification passed
9d9f1616aa
alexander deleted branch mcp/issue-85-ci-tiers 2026-09-24 15:35:52 +02:00
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
alexander/Audiobookshelf-Manager-Claude!91
No description provided.