[Fix] (CI): merge Arch WSL runner and prefer it over hosted runners

2026-11-07 01:53:24 +09:00 · 2026-10-05 09:16:06 -04:00
parent 24889447cc 5331106e90
commit ee9919ba39
7 changed files with 573 additions and 273 deletions
@@ -0,0 +1,33 @@
+name: Configure Test runner
+description: Select the installed compiler and graphics stack without modifying the host
+runs:
+  using: composite
+  steps:
+    - shell: bash
+      run: |
+        if [ "$RUNNER_ENVIRONMENT" = "self-hosted" ]; then
+          for tool in clang clang++ cmake ninja ccache; do command -v "$tool"; done
+          test -r /usr/share/vulkan/icd.d/dzn_icd.json
+          test -e /dev/dxg
+          {
+            echo 'MOBILEGL_CI_CC=clang'
+            echo 'MOBILEGL_CI_CXX=clang++'
+            echo 'MOBILEGL_CI_LLD=ld.lld'
+            echo 'MOBILEGL_CI_SOFTWARE_RENDERING=OFF'
+            echo 'VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/dzn_icd.json'
+            echo 'VK_DRIVER_FILES=/usr/share/vulkan/icd.d/dzn_icd.json'
+            echo 'EGL_PLATFORM=surfaceless'
+            echo 'LIBGL_ALWAYS_SOFTWARE=0'
+            echo 'GALLIUM_DRIVER=d3d12'
+            echo 'MESA_LOADER_DRIVER_OVERRIDE=d3d12'
+          } >> "$GITHUB_ENV"
+        else
+          {
+            echo 'MOBILEGL_CI_CC=clang-20'
+            echo 'MOBILEGL_CI_CXX=clang++-20'
+            echo 'MOBILEGL_CI_LLD=ld.lld-20'
+            echo 'MOBILEGL_CI_SOFTWARE_RENDERING=ON'
+            echo 'VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/lvp_icd.json'
+            echo 'VK_DRIVER_FILES=/usr/share/vulkan/icd.d/lvp_icd.json'
+          } >> "$GITHUB_ENV"
+        fi
@@ -0,0 +1,138 @@
+# MiniPC-8845 Arch WSL GPU Runner
+
+Configuration recorded on 2026-10-05.
+
+## Host
+
+| Component | Configuration |
+| --- | --- |
+| Hostname | `MiniPC-8845` |
+| Windows build | `10.0.26200.9445` |
+| GPU | AMD Radeon 780M Graphics |
+| Windows GPU driver | `32.0.12033.1030` |
+| WSL version | `2.7.3.0` |
+| Distribution | `archlinux`, Arch Linux on WSL 2 |
+| Linux kernel | `6.6.114.1-microsoft-standard-WSL2` |
+| Init system | systemd |
+| WSL networking | Mirrored |
+| GPU access | `/dev/dxg` and `/usr/lib/wsl/lib` |
+
+## GitHub Actions Runner
+
+| Setting | Value |
+| --- | --- |
+| Repository | `MobileGL-Dev/MobileGL` |
+| Runner name | `minipc-8845-arch-wsl-gpu` |
+| Runner version | `2.337.0`, automatic updates enabled |
+| Labels | `self-hosted`, `Linux`, `X64`, `arch`, `wsl`, `gpu`, `gles`, `vulkan`, `d3d12` |
+| Concurrent jobs | 1 |
+| Linux account | `github-runner`, UID 1001, no sudo configuration |
+| Installation directory | `/opt/actions-runner/mobilegl` |
+| Work directory | `/opt/actions-runner/mobilegl/_work` |
+| Environment file | `/etc/mobilegl-runner/gpu.env` |
+
+Workflow runner selector:
+
+```yaml
+runs-on: [self-hosted, Linux, X64, arch, wsl, gpu, gles, vulkan]
+```
+
+## Graphics
+
+| Component | Configuration |
+| --- | --- |
+| Mesa | `26.2.1-arch1.1` |
+| OpenGL / GLES driver | Mesa D3D12 |
+| Renderer | `D3D12 (AMD Radeon 780M Graphics)` |
+| EGL | 1.5 |
+| OpenGL | 4.6 |
+| OpenGL ES | 3.1 |
+| Vulkan driver | Mesa Dozen, translating Vulkan to D3D12 |
+| Vulkan device API | 1.2.354 |
+| Vulkan ICD | `/usr/share/vulkan/icd.d/dzn_icd.json` |
+| Default EGL platform | Surfaceless |
+
+Dozen reports a non-conformant Vulkan implementation. The exposed capabilities
+are GLES 3.1 and Vulkan 1.2; GLES 3.2 and full Vulkan conformance are not available.
+
+The runner service selects the Radeon GPU with these environment variables:
+
+```sh
+DISPLAY=:0
+WAYLAND_DISPLAY=wayland-0
+XDG_RUNTIME_DIR=/mnt/wslg/runtime-dir
+LD_LIBRARY_PATH=/usr/lib/wsl/lib
+MESA_LOADER_DRIVER_OVERRIDE=d3d12
+GALLIUM_DRIVER=d3d12
+MESA_D3D12_DEFAULT_ADAPTER_NAME="AMD Radeon 780M Graphics"
+VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/dzn_icd.json
+EGL_PLATFORM=surfaceless
+PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+```
+
+The installed GPU probe is `/usr/local/libexec/mobilegl-runner/gpu-smoke`.
+Its GLES source is `/usr/local/libexec/mobilegl-runner/gles-smoke.c`, and the
+recorded verification log is `/var/log/mobilegl-runner/gpu-smoke.log`.
+
+## Build Tools
+
+Installed tools include GCC, Clang 22, CMake, Ninja, ccache, LLD, libc++, Git LFS,
+GitHub CLI, and EGL/GLES/Vulkan development headers.
+
+## Service and Startup
+
+| Setting | Value |
+| --- | --- |
+| systemd unit | `actions.runner.MobileGL-Dev-MobileGL.minipc-8845-arch-wsl-gpu.service` |
+| Service drop-in | `/etc/systemd/system/actions.runner.MobileGL-Dev-MobileGL.minipc-8845-arch-wsl-gpu.service.d/gpu.conf` |
+| Service startup | Enabled at WSL systemd boot |
+| Restart policy | `Restart=always`, `RestartSec=15` |
+| Startup checks | GPU device access and WSLg X11/Wayland sockets |
+| Windows scheduled task | `MobileGL Arch WSL Runner` |
+| Task trigger | Windows login by `Swung0x48` |
+| Task behavior | Starts WSL in a hidden window and keeps it running |
+| Task execution limit | Unlimited |
+| Task retry policy | Every minute, up to 999 retries |
+
+Automatic startup requires the Windows user to log in. The runner cannot execute
+jobs while the host is asleep.
+
+## Network
+
+The runner uses the Windows Clash proxy through WSL mirrored networking:
+
+```sh
+http_proxy=http://127.0.0.1:7897
+https_proxy=http://127.0.0.1:7897
+no_proxy=localhost,127.0.0.1,::1
+```
+
+Clash must be running for the runner's configured GitHub connection to work.
+
+## Main Test workflow
+
+`test.yml` uses a small GitHub-hosted `select-runner` job to query this repository's
+runner inventory. Every subsequent job uses its single decision, so artifacts built
+against Arch's compiler/libc/Mesa cannot accidentally be consumed by Ubuntu jobs.
+An online runner is preferred even when busy. Offline, unregistered, or mismatched
+labels select `ubuntu-latest` for the whole run. Test failures do not cause fallback.
+
+Configure repository Actions secret `RUNNER_STATUS_TOKEN` with a fine-grained PAT
+restricted to this repository and **Administration: Read-only**. The ordinary
+`GITHUB_TOKEN` cannot generally list repository runners. A missing/expired token or
+API failure stops selection with an error: unknown availability is not treated as
+an offline machine. The token is passed only to the GitHub-hosted selector.
+
+Availability is sampled at workflow start. If the machine disconnects after
+selection, GitHub queues its remaining jobs; rerunning the workflow while it is
+offline chooses Ubuntu. GitHub Actions does not migrate already scheduled jobs
+between runner types. Keeping one platform per run avoids incompatible artifacts.
+
+The local composite setup action uses Arch's installed Clang/CMake and Dozen ICD.
+Ubuntu package installation, swap provisioning, SDK download, and privileged core
+pattern changes are skipped on this no-sudo host. Compiler caches are separated by
+runner pool. Trace tests disable forced software-rendering environment overrides
+on this machine; the Ubuntu fallback retains its existing Mesa software settings.
+Hardware capability differences remain visible as test results, not fallback triggers.
+The dedicated `wsl-gpu-smoke.yml` workflow has been removed; the host's installed
+probe remains available to its operator.
@@ -0,0 +1,61 @@
+#!/usr/bin/env python3
+"""Choose one execution platform for the entire Test artifact dependency graph."""
+import json
+import os
+import sys
+import urllib.error
+import urllib.request
+
+NAME = "minipc-8845-arch-wsl-gpu"
+LABELS = ["self-hosted", "Linux", "X64", "arch", "wsl", "gpu", "gles", "vulkan"]
+
+
+def choose(runners):
+    for runner in runners:
+        labels = {label["name"].lower() for label in runner.get("labels", [])}
+        if runner.get("name") == NAME and set(map(str.lower, LABELS)) <= labels:
+            if runner.get("status") == "online":
+                # Busy is still available: queue behind its current job.
+                return LABELS, "arch-wsl", "preferred runner online (busy=%s)" % runner.get("busy", False)
+    return ["ubuntu-latest"], "github", "preferred runner offline, absent, or missing required labels"
+
+
+def runner_inventory(repository, token):
+    if not token:
+        raise ValueError("Set RUNNER_STATUS_TOKEN with repository Administration: read; runner availability is unknown")
+    runners = []
+    page = 1
+    while True:
+        url = f"https://api.github.com/repos/{repository}/actions/runners?per_page=100&page={page}"
+        request = urllib.request.Request(url, headers={
+            "Authorization": "Bearer " + token,
+            "Accept": "application/vnd.github+json",
+            "X-GitHub-Api-Version": "2022-11-28",
+            "User-Agent": "MobileGL-test-runner-selector",
+        })
+        with urllib.request.urlopen(request, timeout=20) as response:
+            data = json.load(response)
+        batch = data["runners"]
+        runners.extend(batch)
+        if len(batch) < 100:
+            return runners
+        page += 1
+
+
+def main():
+    try:
+        runners = runner_inventory(os.environ["GITHUB_REPOSITORY"], os.environ.get("RUNNER_STATUS_TOKEN", ""))
+        labels, pool, reason = choose(runners)
+    except (ValueError, KeyError, OSError, urllib.error.URLError) as error:
+        # An API/auth error is not evidence that the machine is offline. Do not
+        # silently send the whole matrix to GitHub when the status cannot be read.
+        print(f"::error::Cannot determine runner availability: {error}", file=sys.stderr)
+        return 1
+    with open(os.environ["GITHUB_OUTPUT"], "a", encoding="utf-8") as output:
+        output.write("runner=" + json.dumps(labels) + "\npool=" + pool + "\n")
+    print(f"Test runner: {pool}: {reason}")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -44,7 +44,7 @@ def run_arm(build, out, case, forward):
    if Path('/usr/share/glvnd/egl_vendor.d/50_mesa.json').exists():
        env['__EGL_VENDOR_LIBRARY_FILENAMES'] = '/usr/share/glvnd/egl_vendor.d/50_mesa.json'
    if Path('/usr/share/vulkan/icd.d/lvp_icd.json').exists():
-        env['VK_ICD_FILENAMES'] = '/usr/share/vulkan/icd.d/lvp_icd.json'
+        env.setdefault('VK_ICD_FILENAMES', '/usr/share/vulkan/icd.d/lvp_icd.json')
    peer_env = dict(env, MOBILEGL_IPC_ROLE='server', MOBILEGL_IPC_DIAL='no',
                    MOBILEGL_IPC_LOG_FORWARD=str(int(forward)), MOBILEGL_LOG_FILE_PATH=str(out / 'peer.log'))
    for name in ('MOBILEGL_TRANSPORT', 'MOBILEGL_IPC_SERVER_PATH', 'MOBILEGL_IPC_RING_MB', 'MOBILEGL_IPC_STAGE_MB'):
@@ -0,0 +1,69 @@
+#!/usr/bin/env python3
+import contextlib
+import importlib.util
+import io
+import json
+import os
+from pathlib import Path
+import tempfile
+import unittest
+from unittest.mock import patch
+import urllib.error
+
+spec = importlib.util.spec_from_file_location("selector", Path(__file__).with_name("select_test_runner.py"))
+selector = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(selector)
+
+
+def runner(status="online", busy=False):
+    return {"name": selector.NAME, "status": status, "busy": busy,
+            "labels": [{"name": label} for label in selector.LABELS]}
+
+
+class SelectRunnerTest(unittest.TestCase):
+    def test_online_busy_and_idle_both_use_preferred(self):
+        for busy in (False, True):
+            labels, pool, _ = selector.choose([runner(busy=busy)])
+            self.assertEqual(labels, selector.LABELS)
+            self.assertEqual(pool, "arch-wsl")
+
+    def test_offline_or_absent_falls_back(self):
+        for rows in ([], [runner("offline")]):
+            self.assertEqual(selector.choose(rows)[:2], (["ubuntu-latest"], "github"))
+
+    def test_wrong_runner_or_missing_labels_cannot_queue_forever(self):
+        wrong = runner()
+        wrong["name"] = "another-machine"
+        missing = runner()
+        missing["labels"] = missing["labels"][:-1]
+        for rows in ([wrong], [missing]):
+            self.assertEqual(selector.choose(rows)[1], "github")
+
+    def test_inventory_reads_later_pages(self):
+        pages = [io.StringIO(json.dumps({"runners": [{}] * 100})),
+                 io.StringIO(json.dumps({"runners": [runner()]}))]
+        with patch.object(selector.urllib.request, "urlopen", side_effect=pages) as request:
+            rows = selector.runner_inventory("owner/repo", "not-a-real-token")
+        self.assertEqual(len(rows), 101)
+        self.assertTrue(request.call_args.args[0].full_url.endswith("page=2"))
+
+    def test_auth_failure_does_not_masquerade_as_offline(self):
+        for error in (ValueError("missing token"), urllib.error.HTTPError("url", 403, "Forbidden", {}, None)):
+            with patch.dict(os.environ, {"GITHUB_REPOSITORY": "owner/repo"}), \
+                 patch.object(selector, "runner_inventory", side_effect=error), \
+                 contextlib.redirect_stderr(io.StringIO()):
+                self.assertEqual(selector.main(), 1)
+
+    def test_outputs_are_valid_runs_on_json(self):
+        with tempfile.TemporaryDirectory() as root:
+            path = Path(root) / "outputs"
+            with patch.dict(os.environ, {"GITHUB_REPOSITORY": "owner/repo", "GITHUB_OUTPUT": str(path)}), \
+                 patch.object(selector, "runner_inventory", return_value=[runner()]), \
+                 contextlib.redirect_stdout(io.StringIO()):
+                self.assertEqual(selector.main(), 0)
+            values = dict(line.split("=", 1) for line in path.read_text().splitlines())
+            self.assertEqual(json.loads(values["runner"]), selector.LABELS)
+
+
+if __name__ == "__main__":
+    unittest.main()
@@ -5,6 +5,8 @@ if(NOT UNIX)
    return()
 endif()

+option(MOBILEGL_TRACE_SOFTWARE_RENDERING "Force Mesa software rendering for trace tests" ON)
+
 set(MOBILEGL_TRACE_ROOT "${CMAKE_CURRENT_LIST_DIR}")
 set(MOBILEGL_TRACE_SHARED_CPP_DIR "${CMAKE_SOURCE_DIR}/android-plugin/app/src/trace/cpp")
 set(APITRACE_ROOT "${CMAKE_SOURCE_DIR}/3rdparty/apitrace")
@@ -445,15 +447,16 @@ function(add_trace_replay_test CASE_NAME BACKEND)
        set(trace_case_labels "retrace\;retrace-${TRACE_CASE_VARIANT}")
        string(TOLOWER "${trace_case_labels}" trace_case_labels)
    endif()
-    if(BACKEND STREQUAL "DirectGLES")
-        set_tests_properties(${trace_case_name} PROPERTIES
-                ENVIRONMENT "EGL_PLATFORM=surfaceless;LIBGL_ALWAYS_SOFTWARE=1;MESA_GL_VERSION_OVERRIDE=3.3;MESA_GLSL_VERSION_OVERRIDE=330"
-                LABELS "${trace_case_labels}")
-    else()
-        set_tests_properties(${trace_case_name} PROPERTIES
-                ENVIRONMENT "LIBGL_ALWAYS_SOFTWARE=1;MESA_GL_VERSION_OVERRIDE=3.3;MESA_GLSL_VERSION_OVERRIDE=330"
-                LABELS "${trace_case_labels}")
+    set(trace_case_environment "")
+    if(MOBILEGL_TRACE_SOFTWARE_RENDERING)
+        list(APPEND trace_case_environment LIBGL_ALWAYS_SOFTWARE=1
+             MESA_GL_VERSION_OVERRIDE=3.3 MESA_GLSL_VERSION_OVERRIDE=330)
    endif()
+    if(BACKEND STREQUAL "DirectGLES")
+        list(APPEND trace_case_environment EGL_PLATFORM=surfaceless)
+    endif()
+    set_tests_properties(${trace_case_name} PROPERTIES
+        ENVIRONMENT "${trace_case_environment}" LABELS "${trace_case_labels}")
    if(TRACE_CASE_IPC_CONTROL)
        set_tests_properties(${trace_case_name} PROPERTIES RESOURCE_LOCK mobilegl-tcp-trace)
    endif()