v0.1.7 build failed on ':react-native-screens:configureCMake' —
ZipException "Archive is not a ZIP archive" from a truncated cmake;3.22.1
download. Install ndk+cmake up front and retry up to 3x, clearing the partial
dir between attempts so a bad download self-heals.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
App:
- When the single-estimate fallback sees MaxTime 0 (or an all-$0.00 ladder),
show "Parking is currently free" instead of a bogus $0.00 purchase ladder.
Free-detection extracted to tested client helpers (isFreeEstimate/ladderAllFree).
Client + tests (node:test, zero deps):
- src/estimates.ts: pure isFreeEstimate/ladderAllFree helpers (exported).
- test/estimates + test/http (mocked fetch): headers, query serialization,
error->ParkSmarterApiError, rolling Auth_Token refresh, 401 clears token.
- test/integration: live checks against the Sandpoint (DSB) zones; targets the
`test` instance by default and SKIPS when it's unreachable (non-prod is
IP-restricted -> 403). Verified 3/3 green against prodv2 read-only.
- CI runs `npm test` as a build gate.
Note: dev/stage/test instances return 403 "Web App - Unavailable" from the
public internet (IP-restricted). API schema is otherwise unchanged vs Jul 6.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The refactor dropped the command-line -Djava.net.preferIPv4Stack from gradlew, so
artifact-transform/worker JVMs reverted to IPv6 and couldn't reach dl.google.com
("Network is unreachable"). JAVA_TOOL_OPTIONS is inherited by every spawned JVM.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- scripts/deploy-runner.sh: compose-managed BigBrainParking Forgejo runner
(adapted from the Camp Scan one), with valid_volumes for bigbrainparking-gradle
and bigbrainparking-npm. No android-sdk volume — the RN build image ships it.
- Workflow mounts those caches (GRADLE_USER_HOME + npm cache) and enables the
Gradle build cache (org.gradle.caching + --build-cache), so repeat builds skip
the dep re-download and recompile that dominate a cold ~35 min build.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Server report (20GB/4-core box) showed 3 Kotlin daemons at -Xmx5376m each + clang
swarm ~9GB thrashing swap. Cap kotlin.daemon.jvmargs and org.gradle.jvmargs to
1536m in gradle.properties, workers.max=2, no parallel. Container: --cpus=2
--memory=8g with swap disabled (--memory-swap==--memory) so it OOM-fails instead
of thrashing disk. Build now uses ~5GB.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Container hard limits: --cpus=2 --memory=6g (Docker-enforced regardless of Gradle).
Gradle: -Xmx3g heap, --max-workers=2, no parallel project builds. Adjust the
--cpus/--memory to the server's size.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The signing plugin added a release{} inside signingConfigs{}, so the debug-line
replacement matched that block instead of the release BUILD TYPE — CI produced
debug-signed APKs (would break Obtainium updates vs the release-signed v0.1.0).
Anchor on `buildTypes { ... release {`. Verified locally: release buildType now
uses the release key when BBP_UPLOAD_* is set. Also unify keystore path via
github.workspace in both steps.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
react-native-edge-to-edge's buildscript needs com.android.tools.build:gradle:8.1.1
from Google Maven; the runner's container has no IPv6 route so the AAAA lookup was
"Network is unreachable". Pass -Djava.net.preferIPv4Stack=true to Gradle.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- .forgejo/workflows/build-apk.yml (Forgejo's location): on v* tag, build signed
APK using the Actions secrets and publish via the Forgejo API (curl + auto token),
no external action. Best-effort NDK install for the RN version.
- Remove the old .gitea workflow to avoid double runs.
- Bump app version to 0.1.1.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
