Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
19 commits
Select commit Hold shift + click to select a range
b3e3c2e
docs: specify cross-platform UI redesign
Shakibuzzaman3104 Aug 11, 2026
1ad1a5e
docs: plan cross-platform UI redesign
Shakibuzzaman3104 Aug 11, 2026
1e80fff
chore: ignore local worktrees
Shakibuzzaman3104 Aug 11, 2026
c8c0960
test(dashboard): define redesign contract
Shakibuzzaman3104 Aug 11, 2026
b5672f2
feat(dashboard): follow system theme by default
Shakibuzzaman3104 Aug 11, 2026
9499a0f
docs(dashboard): refine network list direction
Shakibuzzaman3104 Aug 11, 2026
e72eb28
style(dashboard): adopt graphite cobalt foundation
Shakibuzzaman3104 Aug 11, 2026
077a723
fix(dashboard): strengthen control outlines
Shakibuzzaman3104 Aug 11, 2026
5242cde
style(dashboard): simplify shell and navigation
Shakibuzzaman3104 Aug 11, 2026
6aa1b90
feat: complete native siblings UI redesign for web and android
Shakibuzzaman3104 Aug 11, 2026
1ba4c0b
fix: remove excessive padding above titles in android inspector
Shakibuzzaman3104 Aug 11, 2026
a998538
style(compose): adopt M3 springs and flat list sections
Shakibuzzaman3104 Aug 11, 2026
67c9ff8
release: 1.1.0 — native-siblings UI pass and four sweep fixes
Shakibuzzaman3104 Aug 11, 2026
08f5a46
docs(mocks): document the two-way engine switch and its asymmetric ga…
Shakibuzzaman3104 Aug 12, 2026
c1758d5
fix(publishing): sign only when a key is configured; trim comments
Shakibuzzaman3104 Aug 12, 2026
3e0d473
build: add jitpack.yml
Shakibuzzaman3104 Aug 12, 2026
11121e8
docs: document JitPack as the unreleased-code channel
Shakibuzzaman3104 Aug 12, 2026
82484a6
release: 1.1.1 — buildable without the signing key, JitPack channel
Shakibuzzaman3104 Aug 12, 2026
54e5f0f
docs: refresh screenshots and humanize the reader-facing docs
Shakibuzzaman3104 Aug 12, 2026
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -20,3 +20,6 @@ captures/

# macOS
.DS_Store

# Local feature worktrees
.worktrees/
129 changes: 129 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,135 @@ joined this list.)

## Unreleased

## 1.1.1 — 2026-08-12

A build-and-distribution patch: no SDK code changed, and the public API is byte-for-byte the
1.1.0 surface. It exists because 1.1.0 could not be built by anything that lacks the maintainer's
signing key — including JitPack, which this release adds as a channel.

### Added

- **JitPack as a secondary channel, for unreleased code.** Maven Central remains the supported
one; JitPack covers trying a branch or an unmerged fix. Artifacts resolve under the group
`com.github.devconsole-android.DevConsole` with the same artifact ids
(`com.github.devconsole-android.DevConsole:devconsole:1.1.1`), and any branch resolves as
`<branch>-SNAPSHOT` (a `/` in the branch name is written `~`). Two limits worth knowing: the
Gradle plugin is **not** served there — it lives in an `includeBuild` and JitPack lists only the
31 library modules, so JitPack consumers name coordinates directly and give up the plugin's
variant-policy enforcement — and JitPack artifacts are unsigned.

### Fixed

- **`./gradlew publishToMavenLocal` failed for anyone without the maintainer's signing key.**
`signAllPublications()` registers a signing task *and* the resulting `.asc` files as publication
artifacts. With no key the task fails outright ("no configured signatory") rather than skipping,
and even when the task is explicitly excluded the publication still demands `.asc` files nothing
produced ("artifact file does not exist: …-release.aar.asc"). Both shapes broke the same three
audiences: contributors running the exact command `CONTRIBUTING.md` asks for, CI building from a
clean checkout, and JitPack. Publications are now signed only when a key is actually configured —
the key itself, not the id/password a maintainer's machine keeps between releases. Release
signing is unchanged and verified both ways: five signatures with the key present, none and a
green build without. `docs/MAVEN_PUBLISHING.md` has claimed since 1.0.0 that signing "no-ops when
no key is configured"; that is now true rather than aspirational.

- **`docs/COMPOSER_AND_MOCKS.md` described re-enabling mocks without saying how**, from before
[1.1.0](#110--2026-08-12) added `POST /api/v1/mocks/enabled` — until then the browser genuinely
could not. Now documents both directions and why turning mocking on is capability-gated while
turning it off is not.

## 1.1.0 — 2026-08-12

A minor in the strict sense the [policy](#versioning-and-stability-policy) commits to: two additive
methods on `DevConsoleRuntime`, one new REST route, and no change to any existing surface. Most of
the release is the native-siblings UI pass across both surfaces, plus four defects a full
feature sweep of the shipped build turned up.

### Added

- **`POST /api/v1/mocks/enabled` — the way back from the mock kill switch.** The browser could
disable all mocking (`POST /api/v1/mocks/disable-all`) but had no route to turn it back on, so
the only way to resume mocking was to restart the host app — while the Android in-app inspector
had a two-way toggle the whole time. The dashboard's "Disable all mocks" button is now a two-way
switch that reads its label from the engine's live state. The two directions gate differently and
deliberately: turning mocking *off* stays ungated (falling back to real traffic is always
allowed), while turning it back *on* requires the host's `mocks` editing capability, since it
changes how the app behaves.

- **`DevConsoleRuntime.recordPublishedEvent()` / `recordDroppedEvents(count)`**, the counters behind
the fix below.

### Fixed

- **Every list row in the dashboard rendered without its path or summary.** `.row-main` is the
flexible column between a row's fixed ones, and those fixed columns — checkbox 24px, method badge
54px, duration 74px, status 62px, flag 24px, five 12px gaps, 20px padding — added up to 318px
inside a list pane whose default width was 320px. The path text was laid out into the leftover
2px and clipped to nothing, so Network, Timeline, WebSockets, Push, and Crashes all showed rows
identifiable only by their method badge and duration; on the two-column rows a sliver of the
leading `/` survived and read as a stray comma. The detail pane had the full URL all along, which
is what kept this from being obvious. The row gap is now 8px and the list column defaults to
440px with a 360px floor (it stays drag-resizable), leaving the main text 142px at the default.
The `.col-*` header cells were re-aligned to match, including `.col-flag`, which had been 34px
against the 24px flag button it labels.

- **The SDK-health card reported "0 published events" no matter how much the SDK captured.**
`publishedEventCount` and `droppedEventCount` are surfaced on the dashboard's SDK Health screen
and the Android More screen's health rows, and both were hardcoded 0 in every shipping build:
the counters live on `EventPipeline`, which nothing in the runtime ever instantiates — real
captures funnel through `CaptureTimelineBridge.emit()` instead, and `DevConsoleRuntime`'s own
`SdkHealth` only ever tracked `initializationCount` and `state`. A dropped-event count that can
never rise is the more consequential half: it is the SDK's only signal that capture lost data to
buffer overflow. `emit()` now counts each published event and the persistence queue's existing
`onDrop` hook counts each dropped one.

- **A malformed request answered `500 INTERNAL_ERROR`.** Routes that read a form body — `POST
/api/v1/push/simulate`, the `POST` halves of the HAR and Postman exports, `POST /api/v1/exports`,
and the rest — call `receiveParameters()`, which throws when handed JSON or no `Content-Type` at
all. That landed on the module's catch-all boundary and came back as a server error, sending
anyone writing a third-party client hunting a bug in the SDK for what was their own malformed
request. Content-transformation failures now answer `400 VALIDATION_FAILED` and the typed
media-type rejection answers `415 UNSUPPORTED_MEDIA_TYPE`, so an `INTERNAL_ERROR` from this
server once again means only a genuine server-side fault. The dashboard was never affected — it
sets the right content type on every one of these calls.

- **A pill button clipped its own label.** `InspectorPillButton` laid its label out with `maxLines =
1` and the default clip overflow, so the More screen's "Code" action — squeezed by the "Copy URL"
pill's 2f weight — rendered as "Cod". The label now ellipsizes rather than silently dropping a
glyph, the pill's horizontal padding is 12dp, and the URL card's primary action takes 1.5f.

- **Starting the server on a port that was taken between the availability probe and the real bind
could crash the host app.** `KtorLocalServerEngine.start` preflights each candidate port with
`isPortAvailable`, then calls `embeddedServer(CIO, ...).start(wait = false)` — which returns
*before* CIO's accept loop has actually bound. Losing that race throws `BindException` inside
Ktor's own accept coroutine, long after the `runCatching` around `start()` has returned, and with
nothing on the server's parent context it reached the thread's default uncaught handler: on
Android, the host application's crash handler. A debug console taking down the app it exists to
observe, for a condition it already handles — `isServerBound` sees the failure and the loop falls
forward to the next port exactly as intended. The embedded server now runs in an engine-owned
`SupervisorJob` scope carrying a `CoroutineExceptionHandler`, so a detected-and-recovered bind
failure stays the SDK's business. Surfaced by a flaky test rather than a report: Gradle runs
`testDebugUnitTest` and `testReleaseUnitTest` in parallel, both binding this module's fixed
`8400..8419` range, and the loser's escaped exception landed on whichever unrelated test called
`runTest` next as `UncaughtExceptionsBeforeTest` — roughly two failures in every five suite runs,
on two tests that had nothing to do with it.

- **OkHttp responses without a `Content-Length` showed a blank response body unless the host read
the whole thing.** Those responses — `Transfer-Encoding: chunked`, and every transparently
gzipped response, which is most of them — are captured by the tee, which copies bytes *as the
host reads them*. A call site that reads only the status code and closes (`execute().use {
it.code }`, the shape all three samples had), or a parser that stops at the end of the value it
wanted, therefore left nothing to capture: the transaction was recorded with an empty body and
`bodyOmittedReason = "partial"`. The same response *with* a `Content-Length` was captured whole
by the eager `peekBody`, and the Ktor plugin — which drains its own half of the split channel —
was never affected, so identical app code showed the body on one client and a blank pane on the
other. `TeeCapturingSource.close()` now drains whatever the host left behind before completing
the close, bounded by the existing 512 KiB cap and a 300 ms deadline (the same
bounded-drain-on-close OkHttp itself performs for connection reuse). A body that runs out of that
budget is still recorded `"partial"`, and one no bytes were recovered from is now recorded with
no body instead of an empty one — an empty preview beside a 200 reads as "the server sent
nothing", which is a different claim. The compose sample's OkHttp card now reads its response
body, like its Ktor card already did.

## 1.0.1 — 2026-08-08

A patch in the strict sense the [policy](#versioning-and-stability-policy) now commits to: one
Expand Down
66 changes: 34 additions & 32 deletions CONTRIBUTING.md
Original file line number Diff line number Diff line change
@@ -1,18 +1,18 @@
# Contributing to DevConsole

Thanks for taking a look. This project is a 25-module Android SDK plus two included Gradle
builds (`build-logic`, `gradle-plugin`), so the fastest way in is to lean on the same commands CI
runs — see [Verify](.github/workflows/verify.yml).
Thanks for taking a look. This is a 31-module Android SDK plus two included Gradle builds
(`build-logic` and `gradle-plugin`), which sounds like a lot until you realise you rarely touch more
than one at a time. The quickest way in is to run the same commands CI runs, listed below and in
[Verify](.github/workflows/verify.yml).

## Prerequisites

- JDK 17 (matches `compileOptions`/`setup-java` in CI; the `build-logic` and `gradle-plugin`
included builds target the same toolchain).
- Android SDK with `compileSdk`/`targetSdk` 35 installed (Android Gradle Plugin 8.13.0, `minSdk`
23). See the [compatibility table](README.md#compatibility) in the README for the full matrix.
- No emulator or device is required to build, run unit tests, or lint — only the three sample
apps' instrumented tests would need one, and those aren't part of the default `build`/`test`
path below.
- No emulator or device needed to build, test, or lint. Only the sample apps' instrumented tests
want one, and those sit outside the default `build` and `test` paths below.

## Building and testing

Expand Down Expand Up @@ -54,48 +54,50 @@ All commands run from the repository root.
./gradlew :samples:views-java-app:assembleDebug :samples:views-java-app:assembleRelease :samples:views-java-app:verifyDevConsoleProtectedArtifacts
```

If you only touched one module, scope any of the above to it, e.g.
Touched one module? Scope the command to it and save yourself the wait:
`./gradlew :sdk:network-okhttp:test :sdk:network-okhttp:ktlintCheck`.

## Making a change

- Public API changes to a module that carries `org.jetbrains.kotlinx.binary-compatibility-validator`
(anything with a committed `api/*.api` file, e.g. `sdk:api`) need `./gradlew apiDump` committed
alongside the change — `apiCheck` in CI fails otherwise.
- Kotlin file header, view-id naming, and general style conventions used throughout this codebase
are documented informally by example — match the surrounding file. New Kotlin files start with:
- Changing public API on a module with a committed `api/*.api` baseline means committing a
`./gradlew apiDump` alongside it. `apiCheck` fails in CI otherwise, which is the point.
- Style is documented by example rather than by rule, so match the file you're in. New Kotlin files
start with:
```kotlin
/**
* @author <you>
* @since DD/MM/YY
*/
```
- Capture code (`sdk:network*`, `sdk:socket*`, `sdk:push*`) must never throw into the host app —
wrap recording paths defensively, matching the existing `runCatching`/`onFailure { logcatInfo(...) }`
pattern used throughout those modules.
- Redaction happens at capture time (`RedactionEngine`, applied via `config.redactionPolicy`); code
that formats or displays already-captured text (dashboard JSON/XML rendering, Compose/Views
detail screens) must not need to know about redaction — if you're adding a redaction concern to a
formatter, something further upstream is probably the better place for it.
- If your change touches a protected-variant boundary (a module with a release no-op counterpart,
or the Gradle plugin's variant-policy enforcement), run the sample assemble/verify commands above
for all three samples — that's what actually exercises the debug/release split.
- Capture code (`sdk:network*`, `sdk:socket*`, `sdk:push*`) must never throw into the host app. A
debugging tool that crashes the app it exists to observe is worse than no tool. Wrap recording
paths defensively, following the `runCatching` / `onFailure { logcatInfo(...) }` pattern already
in those modules.
- Redaction happens at capture time, in `RedactionEngine` via `config.redactionPolicy`. Code that
formats or displays already-captured text (dashboard JSON/XML rendering, the Compose and Views
detail screens) should never need to know redaction exists. If you find yourself adding a
redaction concern to a formatter, the fix probably belongs further upstream.
- If your change touches a protected-variant boundary (a module with a release no-op twin, or the
Gradle plugin's variant-policy enforcement), run the sample assemble and verify commands above for
all three samples. That is what actually exercises the debug/release split.

## Opening a pull request

- Keep PRs scoped to one logical change; the commit message should explain *why*, not just *what*.
- Keep a PR to one logical change, and let the commit message explain *why* rather than *what*. The
diff already says what.
- Include the specific `./gradlew` commands you ran to verify the change in the PR description.
- If you changed public API, confirm `apiDump` is included in the diff.
- If you changed anything security- or redaction-adjacent, call it out explicitly in the PR
description — see [SECURITY.md](SECURITY.md) for how to report a vulnerability privately instead
of through a public PR/issue.
- No DCO or commit-sign-off is currently required or enforced by this repository (no `DCO` check
or commit-signing requirement is configured in CI) — standard `git commit` is fine. This may
change; check [.github/workflows](.github/workflows) if you're unsure what's currently enforced.
- Changed anything security- or redaction-adjacent? Say so plainly in the PR description. If you're
reporting a vulnerability rather than fixing one, [SECURITY.md](SECURITY.md) explains how to do
that privately instead of in a public PR or issue.
- No DCO or sign-off is required; a plain `git commit` is fine. That could change, so check
[.github/workflows](.github/workflows) if you want to be sure what's enforced today.

## Reporting bugs / requesting features

Use the issue templates: [Bug report](.github/ISSUE_TEMPLATE/bug_report.md) or
[Feature request](.github/ISSUE_TEMPLATE/feature_request.md). Do **not** file a public issue for a
suspected credential exposure, production inclusion, authentication bypass, or remote code
execution — see [SECURITY.md](SECURITY.md) for private reporting instead.
[Feature request](.github/ISSUE_TEMPLATE/feature_request.md).

One exception. Never open a public issue for a suspected credential exposure, an accidental
production inclusion, an authentication bypass, or remote code execution.
[SECURITY.md](SECURITY.md) has the private reporting route for those.
Loading
Loading