Update protobufjs to v8#32
Closed
renovate[bot] wants to merge 75 commits into
Closed
Conversation
Bumps the protobufs submodule to de6de8e which introduces two optional top-level messages on TAKPacketV2 at tags 25 and 26 for weather readings and sensor field-of-view cones. Wires them through the full Kotlin SDK: - TakPacketV2Data gains nested EnvironmentData and SensorFovData data classes and two optional top-level fields. The annotations attach to any payload_variant rather than living inside the Payload sealed class, mirroring the proto layout. - CotXmlParser captures <environment> (temperature, windDirection, windSpeed) and <sensor> (azimuth, range, fov, vfov, elevation, roll, model) attributes from any CoT event. A small inferSensorType() helper maps ATAK-CIV's free-form model string to the SensorType enum via keyword heuristics (flir/thermal -> Thermal, lrf -> Laser, nvg -> Nvg, etc.). - CotXmlBuilder emits <environment> and <sensor> elements inside <detail> whenever the fields are set, at the same depth as the other payload-agnostic elements (contact, track, takv). - TakPacketV2Serializer bridges to the Wire-generated Environment / SensorFov messages with unit conversions: temperature as deci-degrees Celsius (sint32, ~0.1 degC resolution), wind speed as cm/s to match TAKPacketV2.speed, all angles as whole degrees. Absent optional fields round-trip through proto3 defaults back to null. Intentional omissions (ATAK-CIV renders these locally from client defaults): fovAlpha, fovRed/Green/Blue, strokeColor, strokeWeight, displayMagneticReference, hideFov, fovLabels, rangeLines. The 237 B LoRa MTU makes shipping RGBA tints for map cosmetics the wrong trade. Test coverage: - New pli_with_sensor.xml fixture (PLI + <environment> + FLIR <sensor>) drives the existing parameterized round-trip, compatibility, and compression tests automatically via TestFixtures auto-discovery. - Explicit "PLI with sensor parses environment and sensor annotations" test in RoundTripTest verifies parse -> compress -> decompress -> rebuild preserves every field, including the Thermal type inferred from the "FLIR-Boson-640" model string. - DictionaryTrainingTest grows a 50-sample envsensor_* block in its corpus generator so the NEXT zstd dictionary retrain learns the new field tags and common sensor models. The committed .zstd dicts are unchanged; this is purely seeding for a future retrain. Measurements on the new fixture: 774 B XML -> 165 B raw proto -> 152 B compressed wire 85 B headroom under the 237 B LoRa MTU Incidental cleanup: testdata/compression-report.md plus casevac.bin and casevac_medline.bin were regenerated by the `generate compression report` test. They were stale drift from the initial squash commit (casevac_medline.xml grew to 1046 B but the report still listed 808 B); the regenerated values now match the actual fixtures. No changes to Swift / Python / C# / TypeScript in this commit — the cross-language rollout follows the same pattern as the v2.x CASEVAC extension, one focused session per language. 269/269 jvmTest green. No dictionary retrain required: existing fixtures encode byte-identically (new optional fields absent), and the new fixture compresses fine without dict hits.
Bumps the protobufs submodule to 9b123f3 which renames the proto message type from `Environment` to `TAKEnvironment`. The bare name collided with SwiftUI's `@Environment` property wrapper, making the generated Swift module unusable in any iOS consumer that imports both SwiftUI and the proto bindings — the Meshtastic-Apple app had 78 such files. Only `TakPacketV2Serializer.kt` changes in the SDK: the Wire import and the `toWire` / `toData` bridge helpers now reference `TAKEnvironment` instead of `Environment`. The SDK's internal data class stays named `TakPacketV2Data.EnvironmentData` to match the source `<environment>` CoT XML element name (we only renamed the wire type, not the internal model). Wire format is unaffected — proto3 encoding is tag-number-based, so every fixture's compressed bytes are byte-identical before and after the rename. All 269 jvmTest cases stay green with no golden file regeneration required.
All 40 test fixture XML files now cluster within ~0.01 degrees of the T-or-C anchor point (33.1284, -107.2528) instead of being scattered across the Philippine Sea and San Francisco. This makes the 40-marker test visually coherent when viewed on iTAK/ATAK — every marker, shape, route, casevac, and emergency beacon lands near the same map viewport. Fixture-internal coordinates (link points for shapes/routes, chat anchor points, aircraft positions) are updated to match. Golden .bin/.pb files and the compression-report.md are regenerated against the new positions. RoundTripTest assertions that hard-coded old fixture coordinates (pli_basic lat/lon, drawing_rectangle_itak longitude range check) are updated to reflect the new T-or-C area. Also added roundToInt() import to avoid IEEE 754 off-by-one on the pli_basic longitude assertion. 269/269 jvmTest green on clean build.
Cross-platform audit identified correctness bugs, performance issues, and non-idiomatic patterns across all 5 SDK implementations. This commit addresses the HIGH and MEDIUM severity findings. C# (HIGH — correctness): - Add CultureInfo.InvariantCulture to all 33 double.Parse/TryParse/ int.Parse/uint.Parse call sites in CotXmlParser.cs. Without this, systems with comma decimal separators (de-DE, fr-FR) would parse "37.7749" as 377749.0 or throw FormatException. - Fix CotXmlBuilder.cs speed/course formatting to use the F() helper for culture-invariant output. - Convert 7 inline Regex allocations to private static readonly fields with RegexOptions.Compiled (6 aircraft-from-remarks patterns + detail extraction regex). - Fix stale pli_basic coordinate assertion in TakTests.cs. Python (HIGH — correctness): - Add `from __future__ import annotations` to all 7 source files for Python 3.9 compatibility. The `bytes | None` union syntax in tak_compressor.py requires 3.10+ without the future import, but pyproject.toml declares requires-python >= 3.9. - Migrate dictionary loading from fragile os.path relative paths to importlib.resources (Python 3.9+). Dictionaries moved into a meshtastic_tak.resources subpackage so they bundle correctly in wheel builds. - Remove sys.path hack from tests/conftest.py; add pythonpath = ["src"] to pyproject.toml [tool.pytest.ini_options] instead. - Fix stale pli_basic coordinate assertion in test_round_trip.py. Swift (MEDIUM — performance + consistency): - Cache ISO8601DateFormatter instances as static let properties in CotXmlParser and CotXmlBuilder. Previously allocated new formatters on every parse/build call. - Make parse() throw CotXmlParserError.prohibitedXmlConstruct for DOCTYPE/ENTITY instead of silently returning an empty TAKPacketV2. Aligns Swift with the other 4 platforms which all throw exceptions. - Update all test call sites to handle throws. - Fix stale pli_basic coordinate assertion in RoundTripTests.swift. TypeScript (MEDIUM — packaging): - Remove phantom fzstd dependency (listed but never imported). - Change moduleResolution from "node" to "node16" for proper ESM support with module: ES2022. - Add "exports" field to package.json for modern ESM resolution. - Fix stale pli_basic coordinate assertions in test files. New files: - CLAUDE.md — comprehensive instructions for Claude Code sessions - .github/copilot-instructions.md — concise instructions for Copilot and other AI assistants All platform test suites pass: Kotlin: 269 tests (unchanged — cleanest implementation) C#: 242 tests TypeScript: 211 tests Python: 193 tests Swift: 193 tests
…TypeScript The entire TypeScript codebase previously passed protobuf packets as Record<string, unknown>, requiring unsafe casts like (packet.cotTypeId as number) ?? 0 on every field access — zero compile-time type checking on the packet structure. New file: typescript/src/types.ts (281 lines) Hand-written TypeScript interfaces for the full TAKPacketV2 schema including all 13 nested message types: TAKPacketV2, GeoChat, AircraftTrack, DrawnShape, Marker, RangeAndBearing, Route, RouteLink, CasevacReport, ZMistEntry, EmergencyAlert, TaskRequest, TAKEnvironment, SensorFov, CotGeoPoint. All fields optional with correct types, matching protobufjs camelCase runtime conventions. Updated files: - CotXmlParser.ts: parseCotXml() returns TAKPacketV2, all payload construction uses specific nested types (GeoChat, DrawnShape, etc.), stripZeros generic constraint changed to object - CotXmlBuilder.ts: buildCotXml() accepts TAKPacketV2, ~80 unsafe casts removed, emit* helpers typed to specific payload interfaces - TakCompressor.ts: compress/decompress/fallback methods use TAKPacketV2 instead of Record<string, unknown> - index.ts: re-exports types.ts All 211 tests pass. Structurally compatible with existing plain objects via TypeScript structural typing — no breaking change for downstream consumers.
The JVM publication of the SDK previously bundled the Wire-generated
`org.meshtastic.proto.*` classes alongside the SDK's own
`org.meshtastic.tak.*` classes. Any consumer that also generated the
same proto classes from the same submodule (e.g. Meshtastic-Android's
core:proto module) hit R8 "Type org.meshtastic.proto.... is defined
multiple times" errors during release builds.
This commit splits the JAR into two artifacts published from the
same Gradle project:
takpacket-sdk-jvm -- SDK classes only (org.meshtastic.tak.**).
Declares takpacket-sdk-protos-jvm as a
transitive runtime dep in its POM.
takpacket-sdk-protos-jvm -- the Wire-generated proto classes
(org.meshtastic.proto.**). Depends on
wire-runtime-jvm.
Standalone consumers pick up both jars transitively — no behavior
change. Consumers with their own copy of the proto classes can exclude
the protos jar:
implementation("com.github.meshtastic.TAKPacket-SDK:takpacket-sdk-jvm:v0.2.0") {
exclude(group = "com.github.meshtastic.TAKPacket-SDK", module = "takpacket-sdk-protos-jvm")
}
This is implemented via a single Gradle project rather than a
multi-project layout (the latter fights KMP's strong assumption that
the kotlin("multiplatform") plugin owns the whole project) — the
jvmJar task is mutated in place to drop org/meshtastic/proto/**, and
a sibling jvmProtosJar task packages just those classes from the
already-compiled JVM classpath.
iOS klibs (iosArm64, iosSimulatorArm64) are unaffected — the proto
classes still ship as part of their commonMain since iOS targets
don't have the duplicate-class problem.
Verified locally with publishToMavenLocal:
takpacket-sdk-jvm-0.2.0.jar : 0 proto, 35 tak classes
takpacket-sdk-protos-jvm-0.2.0.jar: 115 proto, 0 tak classes
Bumps the SDK version 0.1.x -> 0.2.0 (packaging-shape change affects
consumers).
testdata/compression-report.md is regenerated by the jvmTest run --
reflects the v2 dictionary (16KB non-aircraft + 4KB aircraft) that
the dictionaries on disk have used since the last retrain session;
the previous report was stale at v1 (8KB).
Closes #5
…ssue #5) v0.2.0 split the proto classes into a sibling `takpacket-sdk-protos-jvm` publication. That solved the R8 duplicate-class problem but it asked consumers to take on a second JitPack dependency entry — which is out of scope for downstream Meshtastic-Android, the canonical consumer that already has a `:core:proto` module providing those same classes. v0.2.1 takes the simpler route: just strip `org/meshtastic/proto/**` from the JVM JAR. The SDK's compiled bytecode still REFERENCES the proto types (TAKPacketV2, CotType, GeoChat, ...), but the consumer is expected to provide them on the runtime classpath. For Meshtastic ecosystem consumers (the only known consumers), that classpath already includes them. Behavioral / packaging shape: takpacket-sdk-jvm-0.2.1.jar : 0 proto classes, 35 tak classes No sibling protos artifact. iosArm64 / iosSimulatorArm64 klibs unchanged — they still ship protos inline because iOS has no analogous duplicate-class problem and no equivalent proto module on the Apple side. The build script comment block at the bottom of kotlin/build.gradle.kts documents the requirement for standalone JVM consumers (run Wire codegen against meshtastic/atak.proto themselves, or depend on a module that does — Meshtastic-Android's `:core:proto` is the reference example).
The 0.2.x JAR stripping strategy (issue #5 fix) traded one fragility for another: it eliminated R8 duplicate-class errors at the cost of leaving the SDK's bytecode pointing at proto method signatures the consumer's own codegen had to materialize. When the protobufs submodule advanced in one repo but not the other, signatures drifted — the most recent break was SensorFov.getRange_m() flipping from `int` to `Integer` when `range_m` became optional, leaving the SDK linked against the old signature and the app's `core:proto` generating the new one. Switch to the inverse strategy: the SDK is now the single owner of atak.proto Wire codegen. The JVM JAR ships `org.meshtastic.proto.**` (115 classes including TAKPacket, TAKPacketV2, GeoChat, Contact, CotType, SensorFov, TAKEnvironment) alongside the existing `org.meshtastic.tak.**` (34 classes). Consumers stop generating those types themselves — they prune the atak.proto messages out of their own Wire config and pull the SDK's classes transitively. Changes: - Remove the `jvmJar { exclude("org/meshtastic/proto/**") }` block from kotlin/build.gradle.kts. Replace the issue-#5 explanatory comment with one documenting the new strategy and the ABI rationale. - Switch the protobufs submodule's tracked branch from `takv2_geometry` to `master` in .gitmodules. The current pointer (1c62540) is already an ancestor of origin/master, so no submodule bump is needed in this commit. This unblocks Renovate from keeping the submodule on the same line as Meshtastic-Android's tracking pointer. - Add .github/renovate.json with `git-submodules: enabled` so protobufs submodule advances arrive as automated PRs. Mirrors the Meshtastic-Android renovate setup so both repos auto-sync on the same proto cadence. Downstream consumer change (Meshtastic-Android, not in this commit): the app's `core:proto` Wire config needs `prune("meshtastic.TAKPacket")`, `prune("meshtastic.TAKPacketV2")`, and prunes for the other atak.proto message types so it stops generating duplicates. Verified locally: ./gradlew clean jvmJar jvmTest # BUILD SUCCESSFUL jar tf takpacket-sdk-jvm-0.2.1.jar | grep -c "^org/meshtastic/proto/" # 115 jar tf takpacket-sdk-jvm-0.2.1.jar | grep -c "^org/meshtastic/tak/" # 34 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The issue #6 fix (33f04a0) restored full proto-class shipping in the JVM JAR. Running the downstream Meshtastic-Android prune-and-consume pattern surfaced a wrinkle: `meshtastic/module_config.proto` defines message ModuleConfig { message TAKConfig { Team team = 1; MemberRole role = 2; } } so `core:proto`'s Wire codegen needs to keep generating `Team` and `MemberRole` — Wire's prune is cascading, and pruning the enums silently drops the `takConfig.team` / `takConfig.role` fields. The app side verified this empirically: with `prune("meshtastic.Team")` in place, the compiler reports `Unresolved reference 'team'` at every call site that reads from `TAKConfig`. Result: ONLY these two classes need to live on the consumer side; every other atak.proto type can ship from the SDK. Strip just `Team` and `MemberRole` (and their inner `$Companion`/`$ADAPTER$1`/etc. classes) from the JVM JAR. The SDK's bytecode references to them still resolve at the consumer's classpath, the same way the rest of the SDK's hand-written code resolves any other proto type. Verified locally: - `./gradlew clean publishToMavenLocal` builds the JAR - `jar tf takpacket-sdk-jvm-0.2.1.jar | grep -E "Team|MemberRole"` returns empty (previously listed Team.class, MemberRole.class, and their $Companion/$ADAPTER$1 inner classes) - Downstream `./gradlew :app:assembleGoogleRelease` in Meshtastic-Android passes R8 (previously failed at `:app:minifyGoogleReleaseWithR8` with `Type org.meshtastic.proto.MemberRole$Companion$ADAPTER$1 is defined multiple times`) iOS klibs are unchanged — they ship Team/MemberRole the same as before because Apple consumers don't run a parallel proto codegen. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replace JitPack-only maven-publish with com.vanniktech.maven.publish 0.30.0
to publish KMP artifacts to Maven Central under org.meshtastic:takpacket-sdk.
Changes:
- Add vanniktech maven-publish plugin with mavenPublishing {} config block
(Maven Central target, conditional GPG signing, full POM metadata)
- Create kotlin/gradle.properties with publishing coordinates
(GROUP, POM_ARTIFACT_ID, VERSION_NAME=0.2.3)
- Add pluginManagement to settings.gradle.kts for plugin resolution
- Update release workflow to publish to Maven Central on tag push
- Add snapshot publishing job to CI on master push
- Update VERSION file to 0.2.3 (matches latest git tag)
- Note in jitpack.yml that Maven Central is now preferred
Required repository secrets for CI:
OSSRH_USERNAME, OSSRH_PASSWORD (Sonatype OSSRH credentials)
SIGNING_KEY_ID, SIGNING_KEY, SIGNING_PASSWORD (GPG in-memory signing)
Consumers can now depend on:
implementation("org.meshtastic:takpacket-sdk-jvm:0.2.3")
from Maven Central instead of the JitPack coordinate.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
feat(kotlin): Add Maven Central publishing via Vanniktech plugin
Replace JitPack-only maven-publish with com.vanniktech.maven.publish 0.30.0
to publish KMP artifacts to Maven Central under org.meshtastic:takpacket-sdk.
Changes:
- Add vanniktech maven-publish plugin with mavenPublishing {} config block
(Maven Central target, conditional GPG signing, full POM metadata)
- Create kotlin/gradle.properties with publishing coordinates
(GROUP, POM_ARTIFACT_ID, VERSION_NAME=0.2.3)
- Add pluginManagement to settings.gradle.kts for plugin resolution
- Update release workflow to publish to Maven Central on tag push
- Add snapshot publishing job to CI on master push
- Update VERSION file to 0.2.3 (matches latest git tag)
- Note in jitpack.yml that Maven Central is now preferred
Required repository secrets for CI:
OSSRH_USERNAME, OSSRH_PASSWORD (Sonatype OSSRH credentials)
SIGNING_KEY_ID, SIGNING_KEY, SIGNING_PASSWORD (GPG in-memory signing)
Consumers can now depend on:
implementation("org.meshtastic:takpacket-sdk-jvm:0.2.3")
from Maven Central instead of the JitPack coordinate.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
fix(ci): remove signing password env var from workflows
Gradle's in-memory signing can auto-detect the key ID from the ASCII-armored key itself. Passing an explicit KeyId was causing 'key ID must be in a valid form' errors when the secret value didn't match the expected 8-char hex format. Only SIGNING_KEY is needed (no KeyId, no password). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
fix(ci): remove signingInMemoryKeyId from publish workflows
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
Add concurrency settings to both CI and Release workflows so that new runs on the same branch or tag automatically cancel stale in-progress runs, saving Actions minutes on repeated pushes and PR updates. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…ellation Add concurrency groups to cancel in-progress CI runs
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
renovate
Bot
force-pushed
the
renovate/protobufjs-8.x
branch
from
010923e to
9f0889f
Compare
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
TypeScript 5.5+ enforces that module must be 'Node16' when moduleResolution is 'Node16'. This was causing the release build to fail with TS5110. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Use exact versioned JAR filename instead of glob (vanniktech produces -sources.jar and -javadoc.jar that matched the wildcard) - Use ./gradlew instead of bare gradle to ensure wrapper version is used - Set cancel-in-progress: false (don't cancel mid-publish) - Remove unused id-token/attestations permissions Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Stage signed artifacts to local Maven, attest with actions/attest-build-provenance, then publish to Maven Central. Matches the pattern used by mqttastic-client-kmp. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Update protobufs digest to e4219a0
Update node to v24
Update vitest to v4
Update Gradle to v9
Update coverlet.collector to v10
Update typescript to v6
Update Microsoft.NET.Test.Sdk to 18.6.0
renovate
Bot
force-pushed
the
renovate/protobufjs-8.x
branch
from
9f0889f to
0791ab5
Compare
Contributor
Author
Renovate Ignore NotificationBecause you closed this PR without merging, Renovate will ignore this update. You will not get PRs for any future If you accidentally closed this PR, or if you changed your mind: rename this PR to get a fresh replacement PR. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR contains the following updates:
^7.4.0→^8.0.0Release Notes
protobufjs/protobuf.js (protobufjs)
v8.4.2Compare Source
Bug Fixes
v8.4.1Compare Source
Bug Fixes
v8.4.0Compare Source
Features
v8.3.0Compare Source
Features
v8.2.1Compare Source
Bug Fixes
Performance Improvements
v8.2.0Compare Source
Features
Bug Fixes
Performance Improvements
v8.0.3Compare Source
Bug Fixes
v8.0.2Compare Source
Bug Fixes
[@exports](https://redirect.github.com/exports)for enums (#1824) (eb256f0)v8.0.1Compare Source
Bug Fixes
v8.0.0Compare Source
⚠ BREAKING CHANGES
Features
Configuration
📅 Schedule: (UTC)
🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.
♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
🔕 Ignore: Close this PR and you won't be reminded about this update again.
This PR was generated by Mend Renovate. View the repository job log.