Files
element-x-android/.github/workflows/recordScreenshots.yml
T
joel.poyet 1cf01cdc62 Add member directory tab to home navigation (#3)
* feat(healthchat): add member directory tab to home navigation

Third bottom-nav tab backed by the HealthChat API: searchable member
cards with native start-chat, approval-pending banner, retry on
failure and navigation into the native profile screens.

* fix(healthchat): stop tab back handler swallowing back presses

While a screen is pushed on top of Home, the still-composed tab
BackHandler intercepted back presses during the transition and
invisibly switched to the Chats tab. Gate it on the node's target
visibility in both the parent and Home's own backstack.

* fix(healthchat): satisfy detekt, ktlint and konsist rules

Import alias order, composable parameter order in HomeView, and konsist presenter helper shape.

* ci: pass the real repository name to the screenshot recorder

recordScreenshots.yml set GITHUB_REPOSITORY from `secrets.GITHUB_REPOSITORY`.
No such secret exists on this fork, so the expression expands to an empty
string and overrides the built-in environment variable of the same name.
recordScreenshots.sh checks it up front and exits 1 with "No repo specified",
so the workflow could never record anything here.

Use the `github.repository` context instead, which is always populated.

Claude-Session: https://claude.ai/code/session_01MQbp1EPM4TSXq8P8Pyq44F

* Update screenshots

* Revert "Update screenshots"

This reverts commit b99e0d643f.

* Reapply "Update screenshots"

This reverts commit 85bceb6d67.

* test(screenshots): stop a LayoutLib teardown crash failing the shard tests

Paparazzi 2.0.0-alpha05 / LayoutLib 16.2.1 races on HandlerThread startup:
HandlerThread_Delegate.run calls Process.setThreadPriority, which throws
NoSuchMethodError for Thread.setPosixNicenessInternal and kills the
"android.bg" thread. If the test thread reaches getLooper() after that, it
gets null, BackgroundThread.ensureThreadReady() NPEs, and the static
initialiser of LatencyTracker fails. The JVM then marks that class erroneous
for the lifetime of the fork, so every later preview showing the soft
keyboard dies with NoClassDefFoundError.

LayoutLibErrorFilterStatement already swallowed the bare NoSuchMethodError,
but PaparazziLogger re-throws collected background errors from teardown()
wrapped in a MultipleFailuresException when there is more than one, and that
wrapper slipped through. A recent run reported 17 failures with zero image
comparisons failing and zero delta images written.

Unwrap the wrapper by reflection and swallow it only when every collected
cause matches the known-bug list; anything unrecognised, and anything that
cannot be decomposed, is rethrown.

Also close a hole this filter had from the start. Paparazzi runs the test as
`try { body } finally { teardown() }`, so a teardown error discards whatever
the body threw — including a genuine snapshot mismatch. renderFailureCapture
is registered inside the Paparazzi rule to record the body's exception so it
wins over any teardown error, and the filter refuses to swallow anything if
that rule is not registered. Two HorizontalFloatingToolbar previews have been
mismatching silently behind this, and may now surface as real failures.

Claude-Session: https://claude.ai/code/session_01MQbp1EPM4TSXq8P8Pyq44F

---------

Co-authored-by: ElementBot <android@element.io>
2026-10-19 08:56:59 +02:00

95 lines
4.8 KiB
YAML

name: Record screenshots
on:
workflow_dispatch:
pull_request:
types: [ labeled ]
permissions:
# Private fork: actions/checkout needs contents:read to clone. Upstream is public,
# where an empty token can still read the repo, so an empty permissions block works there and
# fails here with `fatal: repository not found`.
contents: read
# Enrich gradle.properties for CI/CD
env:
# Reduced from upstream's -Xmx8g / 4 GB Kotlin daemon for the self-hosted runners.
# Each incus container reports 48 GiB and 8 cores, so a single build has ample room —
# but the containers are oversubscribed against their host, and two concurrent heavy
# Gradle jobs get their daemons SIGKILLed ("the daemon has disappeared", no
# OutOfMemoryError, no heap dump). Running the same job alone succeeds. Lower heaps
# make two builds likelier to coexist; the real fix is sizing the containers to the
# host. See the "Report runner capacity" step.
# Kept character-identical to tests.yml so the record path and the verify path cannot drift
# apart again. These only affect the Gradle daemon — Paparazzi renders in forked test workers
# configured entirely by build.gradle.kts — but keeping them aligned removes a whole class of
# "did record and verify run the same way?" questions.
GRADLE_OPTS: -Dorg.gradle.jvmargs="-Xmx6g -XX:MaxMetaspaceSize=768m -Dfile.encoding=UTF-8 -XX:+HeapDumpOnOutOfMemoryError" -Dkotlin.daemon.jvm.options=-Xmx2g -XX:+UseG1GC
CI_GRADLE_ARG_PROPERTIES: --stacktrace --no-daemon -Dsonar.gradle.skipCompile=true --no-configuration-cache --max-workers=4
jobs:
record:
permissions:
# Need write permissions on PRs to remove the label "Record-Screenshots"
pull-requests: write
contents: write
name: Record screenshots on branch ${{ github.event.pull_request.head.ref || github.ref_name }}
runs-on: [self-hosted, Linux, X64]
if: github.event_name == 'workflow_dispatch' || github.event.label.name == 'Record-Screenshots'
steps:
- name: Free Disk Space (Ubuntu)
# Only meaningful on GitHub's throwaway VMs. On a persistent self-hosted runner this
# would delete dotnet/haskell/docker images from the host on every run.
if: runner.environment == 'github-hosted'
uses: jlumbroso/free-disk-space@54081f138730dfa15788a46383842cd2f914a1be
with:
# This might remove tools that are actually needed, if set to "true" but frees about 6 GB
tool-cache: true
# All of these default to true, but we should only need the 'android' one (and maybe swap-storage?)
android: false
dotnet: true
haskell: true
# This takes way too long to run (~2 minutes) and it saves only ~5.5GB
large-packages: false
docker-images: true
swap-storage: false
- name: Remove Record-Screenshots label
if: github.event.label.name == 'Record-Screenshots'
uses: actions-ecosystem/action-remove-labels@2ce5d41b4b6aa8503e285553f75ed56e0a40bae0 # v1.3.0
with:
labels: Record-Screenshots
- name: ⏬ Checkout with LFS (PR)
if: github.event.label.name == 'Record-Screenshots'
uses: nschloe/action-cached-lfs-checkout@385a8ecc719e50b8c71af6ab01a624b486b7c3bc # v1.2.5
with:
persist-credentials: false
ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.ref || github.ref }}
- name: ⏬ Checkout with LFS (Branch)
if: github.event_name == 'workflow_dispatch'
uses: nschloe/action-cached-lfs-checkout@385a8ecc719e50b8c71af6ab01a624b486b7c3bc # v1.2.5
with:
persist-credentials: false
- name: ☕️ Use JDK 21
uses: actions/setup-java@de7274f081f381c8f8158605e0321c36c376e2e6 # v6.0.1
with:
distribution: 'temurin' # See 'Supported distributions' for available options
java-version: '21'
# Add gradle cache, this should speed up the process
- name: Configure gradle
uses: gradle/actions/setup-gradle@3f5f9adaf7d9fecd50b5935e54106014257a94e6 # v6.4.0
with:
cache-read-only: ${{ github.ref != 'refs/heads/develop' }}
- name: Record screenshots
id: record
run: ./.github/workflows/scripts/recordScreenshots.sh
env:
GITHUB_TOKEN: ${{ secrets.DANGER_GITHUB_API_TOKEN || secrets.GITHUB_TOKEN }}
# Must be the `github.repository` context, not a secret. There is no
# GITHUB_REPOSITORY secret on this fork, so `secrets.GITHUB_REPOSITORY` expands to ""
# and overrides the built-in env var of the same name — recordScreenshots.sh then
# exits 1 with "No repo specified" before Gradle ever runs.
GITHUB_REPOSITORY: ${{ github.repository }}
GRADLE_ARGS: ${{ env.CI_GRADLE_ARG_PROPERTIES }}