[Dont merge] Add R8 regression check workflow by AjeshRPai · Pull Request #219 · android/androidify

AjeshRPai · 2026-10-27T09:28:35Z

Adds Automated presubmit (Git hooks) and CI (GitHub Actions) regression gates powered by android analyze to catch R8 configuration regressions, unshrunk libraries, and DEX size bloat before merge.

1. Prerequisites & Setup Requirements

To adopt this workflow in a repository, the following baseline file, secrets, build settings, and tools are required:

A. Initial R8 Baseline File (`r8-baseline.json`)

Why it is needed: check_app_r8_score.py compares every PR's release App Bundle (.aab) against a committed r8-baseline.json file in the repository root. This file stores the reference android analyze output (shrinkingRatio, optimizationRatio, obfuscationRatio, and uncompressedDexSizeBytes).

How to generate it the first time:

# 1. Build the release App Bundle
./gradlew :app:bundleRelease

# 2. Run android analyze and save the output to r8-baseline.json
python3 scripts/check_app_r8_score.py \
  --aab app/build/outputs/bundle/release/app-release.aab \
  --baseline r8-baseline.json \
  --save-baseline

# 3. Commit r8-baseline.json to your default branch
git add r8-baseline.json && git commit -m "Add initial R8 baseline"

Ongoing updates: Once committed, the GitHub Actions workflow automatically refreshes and commits r8-baseline.json whenever a passing PR merges to main.

B. GitHub Actions Secrets & Permissions

Configure the following in Settings → Secrets and variables → Actions:

GEMINI_API_KEY (Required for AI Diagnosis on Failure):
- Used by the google-github-actions/run-gemini-cli step when either the App Bundle gate or Library Shrinking gate fails.
- Allows Gemini CLI (equipped with the r8-analyzer skill) to inspect the PR diff (.gemini/pr_diff.patch) and comment with the exact keep rule or dependency that caused the regression.
Workflow Token Permissions (GITHUB_TOKEN / Optional PAT):
- The workflow declares contents: write, pull-requests: write, and issues: write permissions so it can post scorecard comments on PRs (gh pr comment) and auto-commit updated r8-baseline.json files on main.
- Optional PAT Secret: If your main branch has branch protection rules that block github.token from pushing commits, add a Personal Access Token secret named PAT with repo / contents: write scope.
Project Build Secrets (e.g., DEBUG_GOOGLE_SERVICES_JSON_BASE64):
- Any secrets required to build :app:bundleRelease in CI (in Androidify, if DEBUG_GOOGLE_SERVICES_JSON_BASE64 is not set, the workflow falls back to copying test-google-services.json to app/google-services.json).

C. Android Gradle Plugin (AGP) & R8 Build Configuration

AGP 9.0+ with Release Minification Enabled (isMinifyEnabled = true):
- In app/build.gradle.kts, the release build type must have isMinifyEnabled = true.
- AGP 9.0+ embeds BUNDLE-METADATA/com.android.tools/r8.json inside app-release.aab, which android analyze reads to calculate shrinking, optimization, and obfuscation scores.
Library Module Consumer Rules (consumerProguardFiles):
- For multi-module checks (check_library_shrinking.py --scan-libraries), library modules must build release AARs (./gradlew assembleRelease) so android analyze can inspect their packaged proguard.txt consumer rules and run isolated D8 vs. R8 shrinking comparisons.

D. Android CLI & Agent Skill

Android CLI (android):
- Installed automatically in CI via https://dl.google.com/android/cli/latest/linux_x86_64/install.sh (or locally on macOS via darwin_arm64/install.sh), providing the android analyze --analysis=optimization command.
r8-analyzer Skill:
- Installed in CI via android skills add --agent=gemini --project=. r8-analyzer, providing the domain instructions at .agents/skills/r8-analyzer/SKILL.md for root-cause keep-rule analysis.

2. Modular Check Scripts (`scripts/`)

Teams can adopt all three checks or pick only the ones that fit their workflow:

Script	Target	What It Checks	Typical Stage
`check_staged_r8_diff.py`	Staged `git diff`, `*.pro`, `libs.versions.toml`	Blocks `enableR8.fullMode=false`, newly added `-keep class <pkg>.** { *; }` rules, and audits newly added Maven dependencies in `~/.gradle/caches`	Local `pre-commit` (<1s)
`check_library_shrinking.py`	`.aar` & `.jar` modules/libraries	Runs an isolated R8 dry-run (`(1 - r8DexSize / d8DexSize) * 100`) and flags consumer `proguard.txt` rules with package-wide wildcards	Local `pre-commit` / CI
`check_app_r8_score.py`	Release App Bundle (`.aab`)	Compares `shrinkingRatio`, `optimizationRatio`, `obfuscationRatio` (0–100) and `uncompressedDexSizeBytes` against `r8-baseline.json`	Local `pre-push` / CI

3. GitHub Actions Workflow (`.github/workflows/r8_regression_check.yml`)

Triggers & Inputs

pull_request / push (main): Runs automatically whenever ProGuard/R8 files (**/*.pro), Gradle build scripts (**/build.gradle.kts, gradle/libs.versions.toml, gradle.properties), r8-baseline.json, or scripts/check_*.py change.
workflow_dispatch: Supports manual execution against any PR or commit with configurable thresholds:
- pr_number: Optional PR number to checkout and comment on.
- max_score_drop (default 2.0): Maximum allowed drop in R8 shrinking, optimization, or obfuscation scores (percentage points).
- max_dex_growth_pct (default 5.0): Maximum allowed uncompressed DEX size increase (%).
- min_lib_shrinking_pct (default 90.0): Minimum required isolated DEX shrinking percentage for module AARs (%).

Workflow Pipeline Steps

Prepare Review Context: Resolves the PR or commit diff range and extracts modified .pro, .gradle.kts, libs.versions.toml, and gradle.properties changes into .gemini/pr_diff.patch.
Install Android CLI & Skills: Installs the latest android CLI (https://dl.google.com/android/cli/latest/linux_x86_64/install.sh) and adds the r8-analyzer skill (android skills add --agent=gemini --project=. r8-analyzer).
Build Release Artifacts: Runs ./gradlew :app:bundleRelease assembleRelease to produce the release App Bundle (app-release.aab) and module AARs (*-release.aar).
Check App Bundle R8 Scores & DEX Size (app_r8_gate): Runs check_app_r8_score.py to compare the release .aab against r8-baseline.json and writes the scorecard to .gemini/r8_summary.md and $GITHUB_STEP_SUMMARY.
Check Module & Library Shrinking Percentage (lib_r8_gate): Runs check_library_shrinking.py (--scan-libraries --append-summary) across built module AARs to verify isolated R8 shrinking ratios and flag package-wide consumer keep rules.
Auto-Ratify Baseline on main: On push to main when both gates pass, runs check_app_r8_score.py --save-baseline and auto-commits updated metrics to r8-baseline.json.
AI-Assisted Diagnosis on Failure (gemini_review): If either gate fails, invokes google-github-actions/run-gemini-cli with the r8-analyzer skill to inspect .gemini/pr_diff.patch and pinpoint the exact keep rule or dependency causing the regression.
PR Comment & HTML Report Upload: Posts the combined R8 scorecard and r8-analyzer remediation recommendations directly to the PR via gh pr comment, uploads r8-config-analyzer-report (r8-config-analyzer-release.html), and fails the check if any regression threshold was breached.

4. Presubmit Git Hooks (`scripts/git-hooks/` & `.pre-commit-config.yaml`)

scripts/git-hooks/pre-commit: Runs check_staged_r8_diff.py on staged changes and check_library_shrinking.py on built module AARs.
scripts/git-hooks/pre-push: Runs check_app_r8_score.py against r8-baseline.json when a release .aab is present.
.pre-commit-config.yaml: Exposes three independent hook IDs (r8-staged-diff-and-deps, r8-library-shrinking-gate, r8-app-bundle-score-gate) so teams using pre-commit can enable individual checks.

…GP 9.2.1 baseline

…cripts

cwoodwar6 · 2026-10-28T15:55:52Z

+from typing import Any
+
+DEFAULT_AAB = "app/build/outputs/bundle/release/app-release.aab"
+FALLBACK_AAB = "app/app-release.aab"


I don't understand the reason to have both a default and a fallback.

Agreed, there wasn't a good reason. The fallback was left over from a local demo setup. Removed it in be01dc1; there's now one default path.

cwoodwar6 · 2026-10-28T15:56:17Z

+    parser.add_argument(
+        "--aab",
+        type=Path,
+        default=None,


You should set default=Path(DEFAULT_AAB).

Done in be01dc1.

cwoodwar6 · 2026-10-28T15:57:43Z

+        subprocess.run(["./gradlew", ":app:bundleRelease"], check=True)
+
+    aab_path = args.aab
+    if aab_path is None:


If you set a default, this can just be if not aab_path.exists(): and you can say "No AAB found at {aab_path}"

Done in be01dc1: it's now a single if not args.aab.exists() check that says "No AAB found at {path}".

cwoodwar6 · 2026-10-28T15:58:43Z

+            )
+            return 2
+
+    print(f"{BOLD}==> Running `ANDROID_CLI_ANALYZE=1 android analyze` on {aab_path}...{RESET}")


ANDROID_CLI_ANALYZE=1 is no longer necessary.

Removed in be01dc1, here and in the other scripts and the workflow.

cwoodwar6 · 2026-10-28T15:58:53Z

+def run_android_analyze(artifact_path: Path) -> dict[str, Any]:
+    """Invokes `android analyze --analysis=optimization --format=json` on an AAR/JAR."""
+    env = os.environ.copy()
+    env["ANDROID_CLI_ANALYZE"] = "1"


No longer necessary.

Removed in be01dc1.

cwoodwar6 · 2026-10-28T16:16:08Z

+
+
+def _unwrap_optimization(data: dict[str, Any]) -> dict[str, Any]:
+    """Supports both `{"results": [{"optimization": ...}]}` and `{"optimization": ...}`."""


No longer necessary

Removed in be01dc1.

cwoodwar6 · 2026-10-28T16:20:48Z

+    }
+
+
+def format_kb(num_bytes: int) -> str:


function name is misleading, it formats bytes to either MB and KB

Renamed to format_size in be01dc1.

cwoodwar6 · 2026-10-28T16:23:07Z

+    r8_bytes = int(r8_compile.get("dexSizeBytes", 0))
+
+    keep_ratio_pct = (r8_bytes / d8_bytes * 100.0) if d8_bytes > 0 else 0.0
+    dex_shrinking_pct = 100.0 - keep_ratio_pct if d8_bytes > 0 else 100.0


Why both the keep % and shrinking %?

No need for both. In be01dc1 I dropped the keep ratio; only the shrinking % is reported.

cwoodwar6 · 2026-10-28T16:25:37Z

+    fld_total, fld_kept, fld_shrunk_pct = _item_stats("fields")
+
+    validate = dex.get("validateConsumerKeepRulesResult", {})
+    used_pkg_wide = int(validate.get("usedPackageWideKeepRules", 0))


I think you want len() here and below instead to get the count of rules.

Yes, thanks. Checking it against the real android analyze output turned up more problems in this function: the package-wide rule lists are under r8CompileResult.configuration, not validateConsumerKeepRulesResult; the kept counts are keptItemCount, not keptCount; and blocked rules use source, not value. All fixed in be01dc1. The watchface AAR now reports its 11 package-wide consumer rules (WARN) instead of 0.

cwoodwar6 · 2026-10-28T16:26:37Z

+def run_android_analyze(artifact_path: Path) -> dict[str, Any]:
+    """Invokes `android analyze --analysis=optimization --format=json` on an AAB/APK/AAR/JAR."""
+    env = os.environ.copy()
+    env["ANDROID_CLI_ANALYZE"] = "1"


Removed the whole file in be01dc1. It was the old all-in-one script, replaced by the three split scripts.

- check_library_shrinking.py: read the keys `android analyze` actually outputs (keptItemCount, configuration.used/unusedPackageWideKeepRules, blockedKeepRules[].source) and count rules with len(). Package-wide consumer rules were always reported as 0 before (watchface has 11). Drop the redundant keep ratio and rename format_kb to format_size. - check_app_r8_score.py: single default AAB path, drop --build and the old JSON shape, read the package name from --analysis=basic. - Remove ANDROID_CLI_ANALYZE (no longer needed) and --no-configuration-cache from the scripts and workflow. - Always run the git hooks, the pre-commit config hooks and the CI workflow instead of filtering by changed files. - Post one PR comment and edit it on later runs. - Remove check_r8_regression.py, superseded by the split scripts.

github-actions · 2026-10-29T08:47:26Z

🛡️ R8 Optimization & Shrinking Gate (`android analyze`)

📦 App Bundle R8 Scores & DEX Size

Package: com.android.developers.androidify
Compiler: AGP 9.3.2 / R8 9.3.16 (Full Mode: True)
Baseline: r8-baseline.json

Metric	Baseline	Current	Delta	Threshold	Status
Shrinking Score	51.15%	51.64%	`+0.49% pts`	`max -2.00% pts`	✅ PASS
Optimization Score	50.81%	51.30%	`+0.49% pts`	`max -2.00% pts`	✅ PASS
Obfuscation Score	51.11%	51.61%	`+0.50% pts`	`max -2.00% pts`	✅ PASS
Uncompressed DEX Size	18.32 MB	17.53 MB	`-0.79 MB (-4.3%)`	`max +5.0% / +500KB`	✅ PASS

📚 Library Shrinking Percentage (Min Threshold: `90.0%`)

Library (`.aar` / `.jar`)	D8 DEX	R8 Kept DEX	DEX Shrinking %	Classes Shrunk	Methods Shrunk	Pkg-Wide Rules	Status
`network-release.aar`	64.35 KB	0.00 KB	100.00%	100.0% (49/49)	100.0% (250/250)	0	✅ PASS
`testing-release.aar`	29.22 KB	0.00 KB	100.00%	100.0% (17/17)	100.0% (119/119)	0	✅ PASS
`theme-release.aar`	209.49 KB	0.00 KB	100.00%	100.0% (51/51)	100.0% (405/405)	0	✅ PASS
`util-release.aar`	44.71 KB	0.00 KB	100.00%	100.0% (24/24)	100.0% (111/111)	0	✅ PASS
`xr-release.aar`	44.73 KB	0.00 KB	100.00%	100.0% (13/13)	100.0% (64/64)	0	✅ PASS
`data-release.aar`	55.13 KB	0.00 KB	100.00%	100.0% (54/54)	100.0% (193/193)	0	✅ PASS
`camera-release.aar`	289.75 KB	0.00 KB	100.00%	100.0% (93/93)	100.0% (550/550)	0	✅ PASS
`creation-release.aar`	343.21 KB	0.00 KB	100.00%	100.0% (86/86)	100.0% (610/610)	0	✅ PASS
`home-release.aar`	186.55 KB	0.00 KB	100.00%	100.0% (47/47)	100.0% (310/310)	0	✅ PASS
`results-release.aar`	602.10 KB	0.00 KB	100.00%	100.0% (200/200)	100.0% (1127/1127)	0	✅ PASS
`watchface-release.aar`	72.20 KB	2.73 KB	96.21%	95.7% (44/46)	96.1% (246/256)	11	⚠️ WARN
`common-release.aar`	45.32 KB	0.00 KB	100.00%	100.0% (31/31)	100.0% (229/229)	0	✅ PASS

On every PR, post a GitHub-style diff of the keep rules that changed compared with main, and the classes they keep. It covers rules that come with library upgrades and existing wildcard rules that start keeping new classes, neither of which shows up in the PR's own diff. It doesn't build anything. With AGP 9.3.2, every release build writes the keep rule .pb file (app/build/outputs/mapping/release/configanalyzer.pb) by default, so the gate uploads the file from the build it already runs. - keep_rule_diff.yml: reusable workflow, given the .pb file's path. On pushes to main it caches the file per commit; on PRs it diffs against main's cached file, writes the job summary and keeps one PR comment up to date. - r8_regression_check.yml: upload the .pb file after the release build and call keep_rule_diff.yml in a second job, which also runs when the gate fails. Optionally skip the diff when the R8 scores didn't change (KEEP_RULE_DIFF_ONLY_IF_SCORES_CHANGED). - scripts/keep_rule_diff.py and scripts/schema_pb2.py: compare two keep rule .pb files and write the Markdown report. Needs AGP 9.3 or newer to write the .pb file; until then the diff is skipped.

zizmor's excessive-permissions audit flags workflow-level write permissions once a workflow has more than one job, which the new keep-rule-diff job made it. Move them to the r8-regression-gate job; keep-rule-diff already sets its own (contents: read, pull-requests: write).

main doesn't run the keep rule diff yet, so the cache has no keep rules from main to compare this PR against. A temporary job builds the PR's base commit once, with AGP 9.3.2 so that it writes the keep rule .pb file, and caches it under the key the keep-rule-diff job looks for. To be removed once the diff has run.

github-actions · 2026-10-29T12:55:30Z

🔍 R8 keep rule diff

+ and - lines are keep rules the PR adds or removes. The indented lines under a rule are the classes it keeps. A rule with no sign didn't change, but now keeps different classes.

@@ app/proguard-rules.pro @@
- -keep class com.firebase.** { *; }
- -keep interface com.firebase.** { *; }
- -keep class okhttp3.** { *; }
-     okhttp3.Address
-     okhttp3.Authenticator
-     okhttp3.Authenticator$Companion
-     okhttp3.Authenticator$Companion$AuthenticatorNone
-     okhttp3.Cache
-     okhttp3.Cache$CacheResponseBody
-     okhttp3.Cache$CacheResponseBody$1
-     okhttp3.Cache$Companion
-     okhttp3.Cache$Entry
-     okhttp3.Cache$Entry$Companion
-     okhttp3.Cache$RealCacheRequest
-     okhttp3.Cache$RealCacheRequest$1
-     okhttp3.Cache$urls$1
-     okhttp3.CacheControl
-     okhttp3.CacheControl$Builder
-     okhttp3.CacheControl$Companion
-     okhttp3.Call
-     okhttp3.Call$Factory
-     okhttp3.Callback
-     okhttp3.CertificatePinner
-     okhttp3.CertificatePinner$Builder
-     okhttp3.CertificatePinner$Companion
-     okhttp3.CertificatePinner$Pin
-     okhttp3.Challenge
-     okhttp3.CipherSuite
-     okhttp3.CipherSuite$Companion
-     okhttp3.CipherSuite$Companion$ORDER_BY_NAME$1
-     okhttp3.CompressionInterceptor
-     okhttp3.CompressionInterceptor$DecompressionAlgorithm
-     okhttp3.Connection
-     okhttp3.ConnectionPool
-     okhttp3.ConnectionSpec
-     okhttp3.ConnectionSpec$Builder
-     okhttp3.ConnectionSpec$Companion
-     okhttp3.Cookie
-     okhttp3.Cookie$Builder
-     okhttp3.Cookie$Companion
-     okhttp3.CookieJar
-     okhttp3.CookieJar$Companion
-     okhttp3.CookieJar$Companion$NoCookies
-     okhttp3.Credentials
-     okhttp3.Dispatcher
-     okhttp3.Dispatcher$promoteAndExecute$Effects
-     okhttp3.Dns
-     okhttp3.Dns$Companion
-     okhttp3.Dns$Companion$DnsSystem
-     okhttp3.EventListener
-     okhttp3.EventListener$AggregateEventListener
-     okhttp3.EventListener$Companion
-     okhttp3.EventListener$Companion$NONE$1
-     … and 323 more
- -keep interface okhttp3.** { *; }
-     okhttp3.Authenticator
-     okhttp3.Call
-     okhttp3.Call$Factory
-     okhttp3.Callback
-     okhttp3.CompressionInterceptor$DecompressionAlgorithm
-     okhttp3.Connection
-     okhttp3.CookieJar
-     okhttp3.Dns
-     okhttp3.EventListener$Factory
-     okhttp3.Interceptor
-     okhttp3.Interceptor$Chain
-     okhttp3.TrailersSource
-     okhttp3.WebSocket
-     okhttp3.WebSocket$Factory
-     okhttp3.internal.SuppressSignatureCheck
-     okhttp3.internal.cache.CacheRequest
-     okhttp3.internal.concurrent.Lockable
-     okhttp3.internal.concurrent.TaskRunner$Backend
-     okhttp3.internal.connection.BufferedSocket
-     okhttp3.internal.connection.ExchangeFinder
-     okhttp3.internal.connection.RoutePlanner
-     okhttp3.internal.connection.RoutePlanner$Plan
-     okhttp3.internal.http.ExchangeCodec
-     okhttp3.internal.http.ExchangeCodec$Carrier
-     okhttp3.internal.http2.FlowControlListener
-     okhttp3.internal.http2.Http2Reader$Handler
-     okhttp3.internal.http2.PushObserver
-     okhttp3.internal.platform.ContextAwarePlatform
-     okhttp3.internal.platform.android.DeferredSocketAdapter$Factory
-     okhttp3.internal.platform.android.SocketAdapter
-     okhttp3.internal.publicsuffix.PublicSuffixList
-     okhttp3.internal.sse.ServerSentEventReader$Callback
-     okhttp3.internal.tls.TrustRootIndex
-     okhttp3.internal.ws.WebSocketReader$FrameCallback
-     okhttp3.logging.HttpLoggingInterceptor$Logger
-     okhttp3.sse.EventSource
-     okhttp3.sse.EventSource$Factory

cwoodwar6

I'm not sure if having the keep rule diff is worth the added proto complexity. It is nice to have when a library changes a keep rule, but for anything changed within the project, those changes are already visible in the git diff.

cwoodwar6 · 2026-10-29T17:56:25Z

Should avoid checking in this file. Do you have the .proto file? Adding a proto dependency seems a bit risky, I'd avoid it if possible.

cwoodwar6 · 2026-10-29T17:57:17Z

+
+    python3 scripts/keep_rule_diff.py base.pb pr.pb -o keep_rule_diff.md
+
+Reading .pb files needs `pip install 'protobuf>=6.33.4'` and schema_pb2.py next to this script.


This dependency should be installed in your workflow yaml.

AjeshRPai and others added 5 commits October 23, 2026 21:45

Add R8 regression & library shrinking gate with android analyze

577f604

Merge branch 'main' into r8-regression-gate-demo

15d23c5

Fix android analyze JSON results unwrapping, zizmor env inputs, and A…

a87895b

…GP 9.2.1 baseline

Pre-commit-hook

583e359

Split R8 regression gate into modular app, library, and staged-diff s…

c5181a4

…cripts

AjeshRPai requested review from riggaroo and tiwiz as code owners October 27, 2026 09:28

AjeshRPai removed request for riggaroo and tiwiz October 27, 2026 09:28

cwoodwar6 reviewed Oct 28, 2026

View reviewed changes

AjeshRPai added 3 commits October 29, 2026 17:00

Updates the keep rules and agp version

8d47ff1

android deleted a comment from github-actions Bot Oct 29, 2026

AjeshRPai mentioned this pull request Oct 29, 2026

[Dont merge] Add R8 regression & library shrinking gate with android analyze #218

Closed

cwoodwar6 approved these changes Oct 29, 2026

View reviewed changes



		def _unwrap_optimization(data: dict[str, Any]) -> dict[str, Any]:
		"""Supports both `{"results": [{"optimization": ...}]}` and `{"optimization": ...}`."""


		python3 scripts/keep_rule_diff.py base.pb pr.pb -o keep_rule_diff.md

		Reading .pb files needs `pip install 'protobuf>=6.33.4'` and schema_pb2.py next to this script.

Conversation

AjeshRPai commented Oct 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

1. Prerequisites & Setup Requirements

A. Initial R8 Baseline File (r8-baseline.json)

B. GitHub Actions Secrets & Permissions

C. Android Gradle Plugin (AGP) & R8 Build Configuration

D. Android CLI & Agent Skill

2. Modular Check Scripts (scripts/)

3. GitHub Actions Workflow (.github/workflows/r8_regression_check.yml)

Triggers & Inputs

Workflow Pipeline Steps

4. Presubmit Git Hooks (scripts/git-hooks/ & .pre-commit-config.yaml)

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🛡️ R8 Optimization & Shrinking Gate (android analyze)

📦 App Bundle R8 Scores & DEX Size

📚 Library Shrinking Percentage (Min Threshold: 90.0%)

Uh oh!

github-actions Bot commented Oct 29, 2026

🔍 R8 keep rule diff

Uh oh!

cwoodwar6 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

AjeshRPai commented Oct 27, 2026 •

edited

Loading

A. Initial R8 Baseline File (`r8-baseline.json`)

2. Modular Check Scripts (`scripts/`)

3. GitHub Actions Workflow (`.github/workflows/r8_regression_check.yml`)

4. Presubmit Git Hooks (`scripts/git-hooks/` & `.pre-commit-config.yaml`)

github-actions Bot commented Oct 29, 2026 •

edited

Loading

🛡️ R8 Optimization & Shrinking Gate (`android analyze`)

📚 Library Shrinking Percentage (Min Threshold: `90.0%`)