swung0x48/MobileGL
1
0
Fork 0
mirror of https://github.com/MobileGL-Dev/MobileGL synced 2026-09-17 00:28:31 +09:00
Code Issues Packages Projects Releases Wiki Activity

[Merge] (MGPipe, P5): integrate package t1

Browse Source
This commit is contained in:
swung0x48
2026-09-11 15:57:13 -04:00
parent 3089ce8c1a 949ed1f037
commit bfaf5f9d0e
15 changed files with 2347 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/test.yml
+629
View File
@@ -736,6 +736,412 @@ jobs:
path: /tmp/core.*
if-no-files-found: ignore
# P5's split build. The verify pair cloned, because the problem is the same one: a third
# configuration of the same sources whose whole value depends on the option having taken.
#
# IT CANNOT RIDE ON build-linux'S ARTIFACT. build-linux passes no -DMOBILEGL_PIPE_PUSH, so it is
# a PULL build; MOBILEGL_BUILD_DISAGGREGATED implies MOBILEGL_PIPE_PUSH (the split path decodes
# into the MGPipe applier and PIPE_PUSH is what compiles the applier), so the split arm needs a
# build of its own exactly as the verify arm does.
#
# THE `nm` STEP IS THE POINT OF THIS JOB. In a build without the option, MG_Remote is not
# compiled at all and MOBILEGL_TRANSPORT's parser does not exist - so the variable is accepted by
# the environment and silently ignored (CONTRACT-P5 5). Every downstream lane would then run
# monolith and go green under a name that says split. CMake will not complain about a typo'd -D,
# so the build-level assertion is the only guard, and its absence is precisely what would make
# the whole split arm meaningless.
build-linux-split:
runs-on: ubuntu-latest
timeout-minutes: 120
permissions:
actions: write
contents: read
env:
BUILD_DIR: build-split
CCACHE_BASEDIR: ${{ github.workspace }}
CCACHE_COMPRESS: "true"
CCACHE_DIR: ${{ github.workspace }}/.ccache
CCACHE_MAXSIZE: 4G
CCACHE_NOHASHDIR: "true"
steps:
- name: Set Swap Space
uses: pierotofy/set-swap-space@v1.0
with:
swap-size-gb: 32
- name: Checkout repo
uses: actions/checkout@v6
with:
# recursive, and load-bearing here beyond the usual: MOBILEGL_BUILD_DISAGGREGATED
# SHADOWS ITSELF BACK TO OFF when 3rdparty/flatbuffers/include is missing
# (CMakeLists.txt:471-486, a normal variable rather than a cache force, deliberately).
# A shallow checkout would therefore configure cleanly, build a monolith library, and
# be caught only by the nm step below.
submodules: recursive
- name: Get CMake
uses: lukka/get-cmake@v4.3.3
- name: Restore ccache
uses: actions/cache/restore@v5
with:
path: .ccache
key: ${{ runner.os }}-test-${{ github.job }}-ccache-v1
restore-keys: |
${{ runner.os }}-test-${{ github.job }}-ccache-
- name: Prepare Vulkan SDK
uses: humbletim/setup-vulkan-sdk@v1.2.1
with:
vulkan-query-version: 1.4.304.1
vulkan-components: Vulkan-Headers, Vulkan-Loader
vulkan-use-cache: true
- name: Update glslang external sources
working-directory: 3rdparty/glslang
run: python update_glslang_sources.py
- name: Install build dependencies
run: |
sudo apt-get update
sudo apt-get install -y ccache clang-20 clang++-20 lld-20 libc++-20-dev libc++abi-20-dev libvulkan-dev libegl1-mesa-dev libgles2-mesa-dev libgl1-mesa-dri mesa-vulkan-drivers ninja-build
- name: Show installed toolchain
run: |
ccache --version
clang-20 --version
clang++-20 --version
dpkg -l 'libc++*' 'libegl*' 'libgles*' 'mesa*' 'vulkan*' || true
- name: Configure CMake
# Release/INFO like the shipped build, for the same reason build-linux-verify gives.
# INFO specifically matters here: ConfigLoader logs the resolved transport at INFO and
# that line is what run_trace_case.cmake and the integration-split lane read back as
# proof the transport resolved in THIS process.
#
# _INPROC implies _DISAGGREGATED implies _PIPE_PUSH; all three are passed anyway, because
# an implication that is asserted in two places is an implication nobody has to remember.
run: |
cmake -S . -B "${BUILD_DIR}" -G Ninja \
-DCMAKE_C_COMPILER=clang-20 \
-DCMAKE_CXX_COMPILER=clang++-20 \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_BUILD_TYPE=Release \
-DMOBILEGL_LOG_ACTIVE_LEVEL=MOBILEGL_LOG_LEVEL_INFO \
-DMOBILEGL_BUILD_TEST=ON \
-DMOBILEGL_BUILD_BENCHMARK=OFF \
-DMOBILEGL_BUILD_INTEGRATION_TEST=ON \
-DMOBILEGL_ITEST_VK_ICD=/usr/share/vulkan/icd.d/lvp_icd.json \
-DMOBILEGL_BUILD_TRACE_REPLAY=OFF \
-DMOBILEGL_PIPE_PUSH=ON \
-DMOBILEGL_BUILD_DISAGGREGATED=ON \
-DMOBILEGL_BUILD_DISAGGREGATED_INPROC=ON \
-DCMAKE_POLICY_VERSION_MINIMUM=3.5
- name: Build
run: cmake --build "${BUILD_DIR}" --parallel "$(nproc)"
# `nm` and NOT `nm -D`, for the reason build-linux-verify spells out: everything under
# MG_Remote is a plain namespace symbol in a CXX_VISIBILITY_PRESET=hidden Release build and
# none of it reaches the dynamic table. The symbol count guards the remaining hole - a
# stripped library would make the grep fail for a third, silent reason.
#
# The mirror of this assertion already exists and is the G1 control: monolith-symbol-report
# asserts that a -DMOBILEGL_BUILD_DISAGGREGATED=OFF library defines NO MG_Remote symbol. The
# two together are ARCHITECTURE.md:524's surviving byte-level equality, in both directions.
- name: The split library really carries MG_Remote
run: |
test -f "${BUILD_DIR}/libMobileGL.so"
defined=$(nm --defined-only "${BUILD_DIR}/libMobileGL.so" | wc -l)
if [ "${defined}" -lt 1000 ]; then
echo "::error::nm --defined-only sees only ${defined} symbols in ${BUILD_DIR}/libMobileGL.so - it looks stripped, so the checks below could not have failed honestly"
exit 1
fi
remote=$(nm --defined-only "${BUILD_DIR}/libMobileGL.so" | grep -c -i "MG_Remote" || true)
if [ "${remote}" -lt 1 ]; then
echo "::error::libMobileGL.so defines no MG_Remote symbol: -DMOBILEGL_BUILD_DISAGGREGATED=ON did not take (a typo'd -D is not a CMake error, and the option shadows itself OFF when 3rdparty/flatbuffers/include is missing). Every lane that consumes this artifact would run MONOLITH while claiming to run split, because the MOBILEGL_TRANSPORT parser does not exist in such a build and the variable is accepted and ignored."
exit 1
fi
# The transport parser itself, which is the symbol the runtime evidence depends on.
if ! nm --defined-only "${BUILD_DIR}/libMobileGL.so" | grep -q "MGPipeApply"; then
echo "::error::libMobileGL.so defines no MGPipeApply* entry point, so MOBILEGL_PIPE_PUSH did not take either and there is no applier for the split path to decode into"
exit 1
fi
echo "libMobileGL.so defines ${remote} MG_Remote symbol(s) and the MGPipe applier (${defined} defined symbols)"
- name: Show ccache stats
if: always()
run: ccache --show-stats
- name: Release superseded ccache entry
if: github.ref_name == github.event.repository.default_branch
env:
GH_TOKEN: ${{ github.token }}
CACHE_KEY: ${{ runner.os }}-test-${{ github.job }}-ccache-v1
run: gh cache delete "${CACHE_KEY}" || true
- name: Save ccache
if: github.ref_name == github.event.repository.default_branch
continue-on-error: true
uses: actions/cache/save@v5
with:
path: .ccache
key: ${{ runner.os }}-test-${{ github.job }}-ccache-v1
- name: Package Linux split runtime
run: |
mkdir -p ci-artifacts
mapfile -t SHARED_LIBS < <(find "${BUILD_DIR}" -type f \( -name '*.so' -o -name '*.so.*' \) -print | sort)
tar \
--exclude='*/CMakeFiles' \
--exclude='*.o' \
--exclude='*.a' \
--exclude='*.ninja*' \
--exclude='build.ninja' \
--exclude='cmake_install.cmake' \
-czf ci-artifacts/mobilegl-linux-runtime-split.tgz \
"${BUILD_DIR}/CTestTestfile.cmake" \
"${BUILD_DIR}/MobileGL/MG_Test" \
"${BUILD_DIR}/MobileGL/MG_IntegrationTest" \
"${SHARED_LIBS[@]}"
- name: Upload Linux split runtime
uses: actions/upload-artifact@v7
with:
name: mobilegl-linux-runtime-split
path: ci-artifacts/mobilegl-linux-runtime-split.tgz
if-no-files-found: error
# The split lane itself: the three-arm shape ARCHITECTURE.md:521 asks for verbatim, plus the
# first real run of the five MG_Test/Wire suites, plus P5's two exit-gate negative controls.
integration-split:
runs-on: ubuntu-latest
timeout-minutes: 120
needs: build-linux-split
steps:
- name: Checkout repo
uses: actions/checkout@v6
- name: Get CMake
uses: lukka/get-cmake@v4.3.3
- name: Install runtime dependencies
run: |
sudo apt-get update
sudo apt-get install -y libvulkan1 libegl1 libegl-mesa0 libgles2 libgl1-mesa-dri mesa-vulkan-drivers
- name: Download Linux split runtime
uses: actions/download-artifact@v8
with:
name: mobilegl-linux-runtime-split
path: .
- name: Unpack Linux split runtime
run: |
tar -xzf mobilegl-linux-runtime-split.tgz
test -f build-split/libMobileGL.so
- name: Normalize CTest command paths
run: |
python - <<'PY'
from pathlib import Path
import re
for path in Path('build-split').rglob('CTestTestfile.cmake'):
text = path.read_text()
text = re.sub(r'"[^"]*/cmake-[^"]*/bin/cmake"', '"cmake"', text)
path.write_text(text)
PY
# THE FIRST REAL RUN OF MG_Test/Wire. Those five suites - Framing, Ring, InProcessTransport,
# ProtocolSmoke and FdPassing - are registered only under MOBILEGL_BUILD_DISAGGREGATED
# (MG_Test/CMakeLists.txt), and before this job existed NO cmake invocation anywhere in this
# workflow passed that option. They had never been compiled by CI, let alone run.
- name: Unit tests on the split runtime
working-directory: build-split
run: ctest --output-on-failure -L unit --no-tests=error -j "$(nproc)"
# --no-tests=error is half the gate, exactly as in integration-verify: the integration-split
# entries exist only when the library was configured with -DMOBILEGL_BUILD_DISAGGREGATED=ON,
# so a build that lost the option matches nothing and reds here instead of reporting a green
# run of nothing. The other half is the transport-resolution check below.
#
# An entry whose owning package (c1 client, s1 session, v1 server) has not landed SKIPS with
# the reason and never goes green - MG_IntegrationTest/CMakeLists.txt probes MG_Remote for
# c0's signature stubs and disarms the lane while any remain.
- name: Split scenarios under MOBILEGL_TRANSPORT=inproc
working-directory: build-split
env:
MOBILEGL_ITEST_REQUIRE_GPU: "1"
run: |
ulimit -c unlimited
sudo sysctl -w kernel.core_pattern='/tmp/core.%e.%p'
ctest --output-on-failure -L integration-split --no-tests=error -j 4
# ARCHITECTURE.md:521 asks for `ctest -L integration-gpu` to be name-for-name identical
# between the monolith and the split arm of the SAME build - the G2 shape extended to a
# third arm. The knob goes in the JOB environment rather than in a ctest property, for the
# reason the MOBILEGL_ESPRYT_DISABLE_INVALIDATE_FLUSH precedent in `integration` gives: a
# property would override it and the arm would not be an arm.
#
# The entries that name MOBILEGL_TRANSPORT in their OWN property (the Split. lanes) keep
# their value in both passes, which is correct: they are the split family in both arms and
# the comparison is about the other 1100.
- name: The same integration entries under monolith and under inproc
working-directory: build-split
env:
MOBILEGL_ITEST_REQUIRE_GPU: "1"
MOBILEGL_MAGMA_FIX_ITERATIONRP_SUBGROUP_SCRATCH: "1"
MOBILEGL_MAGMA_DERIVE_NUM_SUBGROUPS: "1"
MOBILEGL_MAGMA_ITERATIONRP_FIX_BARRIER: "1"
run: |
ulimit -c unlimited
sudo sysctl -w kernel.core_pattern='/tmp/core.%e.%p'
count=$(ctest -N -L integration-gpu | grep -cE '^ *Test *#[0-9]+:')
if [ "${count}" -lt 1 ]; then
echo "::error::the split runtime registers ${count} integration-gpu entries"
exit 1
fi
echo "integration-gpu entries in the split build: ${count}"
MOBILEGL_TRANSPORT=monolith ctest --output-on-failure -L integration-gpu --no-tests=error -j 4 --output-junit "${RUNNER_TEMP}/arm-monolith.xml"
MOBILEGL_TRANSPORT=inproc ctest --output-on-failure -L integration-gpu --no-tests=error -j 4 --output-junit "${RUNNER_TEMP}/arm-inproc.xml"
# NAME **AND STATUS**, and it is the comparison this step claimed to make and did not
# (review finding N-3): the first version wrote a names file and never read it, and
# `--output-on-failure` treats a SKIPPED test as not-a-failure - so the very failure
# ARCHITECTURE.md:521 is about, "an inproc arm that skipped forty entries the monolith arm
# ran", was invisible here and caught only by the local gate. `ctest -N` cannot see it
# either: this is one build directory, so the two arms have identical name lists by
# construction and the difference is entirely in what each entry DID.
python3 - "${RUNNER_TEMP}/arm-monolith.xml" "${RUNNER_TEMP}/arm-inproc.xml" <<'PY'
import sys, xml.etree.ElementTree as ET
def rows(path):
out = {}
for case in ET.parse(path).getroot().iter('testcase'):
status = 'passed'
if case.find('failure') is not None or case.find('error') is not None:
status = 'failed'
elif case.find('skipped') is not None or case.get('status') in ('notrun', 'disabled'):
status = 'skipped'
out[case.get('name')] = status
return out
a, b = rows(sys.argv[1]), rows(sys.argv[2])
diff = sorted(set(a) ^ set(b)) + sorted(n for n in set(a) & set(b) if a[n] != b[n])
if diff:
for name in diff[:40]:
print(f"::error::{name}: monolith={a.get(name, '<absent>')} inproc={b.get(name, '<absent>')}")
print(f"::error::the monolith and inproc arms of ctest -L integration-gpu differ on "
f"{len(diff)} entries. ARCHITECTURE.md:521 requires them identical name for name "
f"AND status; an entry that SKIPPED on one arm and ran on the other is the "
f"failure this compares for, and it is not a failure to --output-on-failure.")
raise SystemExit(1)
print(f"the two arms agree on all {len(a)} entries, name and status")
PY
# THE RUNTIME HALF OF "THIS IS REALLY A SPLIT BUILD". The build-level nm check in
# build-linux-split proves the library CARRIES MG_Remote; this proves the transport
# RESOLVED in a process of this lane. ConfigLoader::InitTransport logs one INFO line when it
# selects InProcess, and the DirectGLES.Split.PersistentMapArm. entry is the one Split entry
# with a MOBILEGL_LOG_FILE_PATH of its own (nothing else writes it, so the grep means what it
# says). The line is written during bring-up, before any scenario decides to skip, so this
# check is live from the day the lanes land rather than from the day they stop skipping.
- name: The split lane really resolved the transport
working-directory: build-split
run: |
log=MobileGL/MG_IntegrationTest/persistent-map-arm-split-DirectGLES.log
if [ ! -f "${log}" ]; then
echo "::error::${log} does not exist: the DirectGLES.Split.PersistentMapArm. entry never ran, so nothing in this job establishes that MOBILEGL_TRANSPORT ever resolved to inproc in a live process"
exit 1
fi
# THE DISTINCTIVE PART OF THE INFO LINE, not the bare KEY=VALUE (review finding M-5):
# ConfigLoader logs `Config: Accepted env variable: MOBILEGL_TRANSPORT=inproc` for the
# env dump too, unconditionally and in a PULL build, and at DEBUG that line is live.
if ! grep -q "MOBILEGL_TRANSPORT=inproc - the MGPipe record stream" "${log}"; then
echo "::error::${log} carries no transport-resolution line. ConfigLoader::InitTransport logs it at INFO when it selects InProcess, and that code exists only in a MOBILEGL_BUILD_DISAGGREGATED build - so this lane ran a monolith library while claiming to be the split lane. (A bare MOBILEGL_TRANSPORT=inproc substring is NOT accepted: the env dump prints one in every build.)"
exit 1
fi
echo "the split lane resolved MOBILEGL_TRANSPORT=inproc"
# EXIT GATE E1's NEGATIVE CONTROL and EXIT GATE E3(a)'s, in one step because they have the
# same two-state shape and the same reason for it.
#
# Both controls turn a knob that MUST make a split scenario red: MOBILEGL_IPC_VERB_BARRIER=0
# removes the lockstep fence R-1 rests on, and MOBILEGL_IPC_PERSISTENT_BLOCK_KB=0 turns the
# persistent-map push off. A control of the verify lane's shape - "this step passes when
# ctest FAILS" - cannot be written yet, because while packages c1/s1/v1 are landing the
# Split entries SKIP and ctest reports green whatever the knob says, so an unconditional
# control would be red for the whole of P5 for a reason that is not a defect.
#
# So the expected state is DERIVED rather than assumed, from the same fact the lanes derive
# it from: MG_IntegrationTest/CMakeLists.txt puts MGITEST_REMOTE_CLIENT_PRESENT=1 into the
# Split entries' ENVIRONMENT exactly when MG_Remote carries no c0 signature stub, and that
# string is in the generated ctest include files this artifact ships. When it is there the
# controls MUST fire; when it is not, the step says so loudly and does not pretend.
- name: Negative controls - the verb barrier and the persistent-map push must be load-bearing
working-directory: build-split
env:
MOBILEGL_ITEST_REQUIRE_GPU: "1"
run: |
# THE ARMED STATE IS DERIVED FROM BEHAVIOUR, not from a marker string in the generated
# ctest files. The first version read MGITEST_REMOTE_CLIENT_PRESENT out of
# *_tests.cmake, which was a restatement of the CMake source probe review finding M-1
# falsified; the arming condition is now a runtime fact inside each test process, so the
# only honest way to ask it from a shell is to look at what the entries DID.
ctest -L integration-split -j 4 --no-tests=error --output-junit "${RUNNER_TEMP}/isplit.xml" || true
armed=$(python3 - "${RUNNER_TEMP}/isplit.xml" <<'PY'
import sys, xml.etree.ElementTree as ET
ran = 0
for case in ET.parse(sys.argv[1]).getroot().iter('testcase'):
if case.find('skipped') is None and case.get('status') not in ('notrun', 'disabled'):
ran += 1
print(ran)
PY
)
echo "split entries that actually ran: ${armed}"
if [ "${armed}" -lt 1 ]; then
echo "::warning::every DirectGLES.Split. entry SKIPPED, so neither negative control can fire. The arming condition is a runtime fact - MG_Config::Transport, ClientSession::Active() and ImplementedVerbCount(), read by Harness/SplitRuntimePeek - and it becomes true on the commit that lands the last of c1/s1/v1. This step becomes a gate then, with no edit; it is not a green that asserted anything today."
exit 0
fi
run_control() {
name="$1"; filter="$2"; shift 2
matched=$(ctest -N -L integration-split -R "${filter}" | grep -cE '^ *Test *#[0-9]+:')
if [ "${matched}" -lt 1 ]; then
echo "::error::${name} selected ${matched} tests; its filter no longer matches anything"
exit 1
fi
if env "$@" ctest --output-on-failure -L integration-split -R "${filter}" --no-tests=error; then
echo "::error::${name} left ${matched} split entries GREEN, so the knob it turns is not load-bearing and the gate it controls proves nothing."
exit 1
fi
echo "${name} turned ${matched} selected entries red, as it must"
}
# E1: R-1's lockstep verb barrier. Without it the client keeps pulling fields from a live
# GLContext while the server runs ahead, so the server reads future values.
run_control "negative control E1 (MOBILEGL_IPC_VERB_BARRIER=0)" \
'DirectGLES\.Split\.(Triangle|ClearThenReadPixels)' MOBILEGL_IPC_VERB_BARRIER=0
# E3(a): the persistent-map push. 0 is admitted by ConfigLoader on purpose and is
# documented there as this control.
run_control "negative control E3(a) (MOBILEGL_IPC_PERSISTENT_BLOCK_KB=0)" \
'DirectGLES\.Split\.PersistentCoherentMapScenario' MOBILEGL_IPC_PERSISTENT_BLOCK_KB=0
- name: Upload split lane logs
if: always()
uses: actions/upload-artifact@v7
with:
name: integration-split-logs
path: build-split/MobileGL/MG_IntegrationTest/*.log*
if-no-files-found: warn
- name: Upload core dumps
if: failure()
uses: actions/upload-artifact@v7
with:
name: integration-split-core-dumps
path: /tmp/core.*
if-no-files-found: ignore
# MobileGL/MG_Remote/Protocol/generated/protocol_generated.h is COMMITTED, and
# flatc is deliberately absent from the default build graph (a codegen step in
# the graph is how the earlier branch ended up cross-compiling an arm64 flatc
@@ -1013,6 +1419,7 @@ jobs:
matrix: ${{ steps.trace-cases.outputs.matrix }}
names: ${{ steps.trace-cases.outputs.names }}
verify-matrix: ${{ steps.trace-cases.outputs.verify-matrix }}
split-matrix: ${{ steps.trace-cases.outputs.split-matrix }}
steps:
- name: Checkout repo
uses: actions/checkout@v6
@@ -1025,6 +1432,20 @@ jobs:
# The subset the verify build retraces ("verify": true in trace_cases.json). It is a
# SUBSET of the matrix above, so retrace-verify needs no fixtures of its own.
echo "verify-matrix=$(python3 tools/trace_replay/trace_cases.py --ci --format github-verify-matrix)" >> "$GITHUB_OUTPUT"
# P5's split subset ("split": true, DirectGLES only). Also a SUBSET of the matrix above,
# so retrace-split needs no fixtures of its own either. It is one case today - OpenRA,
# which is what the phase gate names - and trace_cases.py refuses a `split` case that is
# not in CI or does not run DirectGLES, so the subset cannot silently become empty.
SPLIT_MATRIX=$(python3 tools/trace_replay/trace_cases.py --ci --format github-split-matrix)
# AND IT MUST NOT BE EMPTY. An empty `include` is not an error to GitHub - it skips the
# whole retrace-split job with no red anywhere - so the one way this subset can vanish
# silently is guarded here. trace_cases.py now also rejects an unknown manifest key, which
# was the hole: `"splitt": true` loaded clean and emptied the subset (review N-4).
if [ "$(printf '%s' "${SPLIT_MATRIX}" | python3 -c 'import json,sys; print(len(json.load(sys.stdin)["include"]))')" -lt 1 ]; then
echo "::error::the split retrace subset is EMPTY. No case in trace_cases.json carries \"split\": true, so retrace-split would be skipped with no red. Exit gate E2 names OpenRA."
exit 1
fi
echo "split-matrix=${SPLIT_MATRIX}" >> "$GITHUB_OUTPUT"
trace-fixtures:
name: trace fixture (${{ matrix.case }})
@@ -1397,6 +1818,214 @@ jobs:
build-retrace/tools/trace_replay/${{ matrix.case }}/${{ matrix.backend }}/output/**
if-no-files-found: warn
# P5's retrace arm: retrace-verify with MOBILEGL_PIPE_VERIFY=1 swapped for
# MOBILEGL_TRANSPORT=inproc and the comparator's symbol swapped for an MG_Remote one. Exit gate
# E2 is one row - OpenRA at SSIM >= 0.99 - and trace_cases.json's `split: true` is where that
# subset lives.
#
# IT RUNS THE UNCHANGED CTEST NAMES with the transport exported in the JOB environment, rather
# than the SPLIT-suffixed variant entries. build-retrace configures without
# -DMOBILEGL_BUILD_DISAGGREGATED, so the variant entries are deliberately not registered there
# (a name that says SPLIT in a build that cannot be one is worse than no name), and the
# unchanged names are what retrace-verify already proves this shape works with. The variant
# entries exist for a build that DOES configure the option - a local build-split with trace
# replay on - where `ctest -L retrace-split` is self-describing and needs no environment ritual.
#
# WHAT MAKES IT FALSIFIABLE is not the SSIM. A monolith run of OpenRA also scores 1.000: measured
# on this branch, a pull library under MOBILEGL_TRANSPORT=inproc produced ssim=1.0 and was caught
# only by run_trace_case.cmake's transport-resolution assertion. So there are two guards, and the
# SSIM is neither of them: the nm check below (this library carries MG_Remote) and the
# transport-resolution line in the library's own log (this PROCESS resolved inproc).
retrace-split:
name: retrace split (${{ matrix.backend }}, ${{ matrix.case }})
runs-on: ubuntu-latest
timeout-minutes: 240
needs:
- build-linux
- build-linux-split
- build-retrace
- trace-cases
- trace-fixtures
# build-linux is needed for the negative control ONLY: its pull library is what the control
# swaps in to prove the transport-resolution assertion is load-bearing.
if: ${{ always() && needs.build-linux.result == 'success' && needs.build-linux-split.result == 'success' && needs.build-retrace.result == 'success' && needs.trace-cases.result == 'success' }}
strategy:
fail-fast: false
max-parallel: 4
matrix: ${{ fromJSON(needs.trace-cases.outputs.split-matrix) }}
steps:
- name: Set Swap Space
uses: pierotofy/set-swap-space@v1.0
with:
swap-size-gb: 16
- name: Checkout repo
uses: actions/checkout@v6
- name: Download trace fixture
uses: actions/download-artifact@v8
with:
name: trace-fixture-${{ matrix.case }}
path: trace-fixture-download
- name: Install trace fixture
run: |
mkdir -p tools/trace_replay/fixtures
find trace-fixture-download -type f -exec cp {} tools/trace_replay/fixtures/ \;
- name: Get CMake
uses: lukka/get-cmake@v4.3.3
- name: Install runtime dependencies
run: |
sudo apt-get update
sudo apt-get install -y libvulkan1 libegl1-mesa-dev libgles2-mesa-dev libgl1-mesa-dri mesa-vulkan-drivers
test -e /usr/lib/x86_64-linux-gnu/libEGL.so
test -e /usr/lib/x86_64-linux-gnu/libGLESv2.so
- name: Download Linux split runtime
uses: actions/download-artifact@v8
with:
name: mobilegl-linux-runtime-split
path: .
- name: Download trace replay
uses: actions/download-artifact@v8
with:
name: mobilegl-trace-replay
path: .
# The PULL runtime, for the negative control at the end of this job and for nothing else.
- name: Download Linux pull runtime (negative control)
uses: actions/download-artifact@v8
with:
name: mobilegl-linux-runtime
path: pull-runtime
- name: Unpack the pull runtime (negative control)
run: |
tar -xzf pull-runtime/mobilegl-linux-runtime.tgz -C pull-runtime
test -f pull-runtime/build-linux/libMobileGL.so
- name: Unpack the SPLIT runtime as the library under test
# build-retrace's CTestTestfile.cmake has the absolute path
# <workspace>/build-linux/libMobileGL.so frozen into every case, so the swap happens here
# rather than through a variable, exactly as in retrace-verify. `nm`, not `nm -D`.
run: |
tar -xzf mobilegl-linux-runtime-split.tgz
tar -xzf mobilegl-trace-replay.tgz
test -f build-split/libMobileGL.so
test -f build-retrace/tools/trace_replay/mobilegl_trace_replay
mkdir -p build-linux
cp build-split/libMobileGL.so build-linux/libMobileGL.so
if ! nm --defined-only build-linux/libMobileGL.so | grep -q -i MG_Remote; then
echo "::error::the library unpacked at build-linux/libMobileGL.so defines no MG_Remote symbol, so this retrace would replay against a monolith build, ignore MOBILEGL_TRANSPORT entirely and match its golden having split nothing"
exit 1
fi
echo "the library at build-linux/libMobileGL.so is the split build"
- name: Retrace and validate under MOBILEGL_TRANSPORT=inproc
working-directory: build-retrace/tools/trace_replay
# run_trace_case.cmake turns MOBILEGL_TRANSPORT into assertions of its own - the library
# must have RESOLVED the transport in this process, and its log must carry no Fatal{ - so a
# case that somehow ran the wrong library reds here instead of passing on its golden. That
# log is also the only valid refusal census: the console sink is compiled out of this
# configuration, so a `ctest -V` transcript reports a false zero for Fatal{ lines.
run: |
ulimit -c unlimited
sudo sysctl -w kernel.core_pattern='/tmp/core.%e.%p'
export MOBILEGL_TRANSPORT=inproc
ctest -V --no-tests=error --timeout 10800 \
-R '^MobileGLTraceReplay\.${{ matrix.case }}\.${{ matrix.backend }}$'
# THE RETRACE LANE'S OWN ALWAYS-ON NEGATIVE CONTROL, and its absence was review finding M-6:
# this job was a clone of retrace-verify with the one step removed that makes the lane mean
# anything. retrace-verify carries "a corrupted snapshot field must red this retrace" so that
# "79 traces, zero divergences" is not a statement about a comparator nobody watched; the
# same sentence applies here word for word.
#
# The control swaps the PULL library into the frozen path and requires the SAME replay to
# fail. It is the sharpest one available, because OpenRA scores ssim 1.000000 either way -
# measured - so this step fails if and only if run_trace_case.cmake's transport-resolution
# assertion has stopped working, which is the single thing standing between this job and a
# green that ran monolith end to end.
#
# NOT the control BRIEF 7 E2 names ("patch the Clear emitter to drop one emission; SSIM must
# fall below threshold"). That one needs an emitter, i.e. package c1, and it is carried as an
# explicit debt in t1-v1.md rather than silently substituted - which is what the first
# version of this package did.
#
# The rerun replays into the same case directory, so the good run's images are put aside and
# restored whichever way the control goes; "Upload actual image" below runs `if: always()`
# and would otherwise ship the deliberately-wrong run's output under the good run's name.
- name: Negative control - the PULL library must red this split retrace
working-directory: build-retrace/tools/trace_replay
run: |
set +e
GOOD_OUTPUT="${RUNNER_TEMP}/split-verified-output"
rm -rf "${GOOD_OUTPUT}"
if [ -d "${{ matrix.case }}" ]; then cp -a "${{ matrix.case }}" "${GOOD_OUTPUT}"; fi
# The pull library, unpacked from build-linux's artifact, over the frozen path every
# case has baked in. It defines no MG_Remote symbol, so ConfigLoader has no transport
# parser and MOBILEGL_TRANSPORT=inproc is accepted and ignored - the exact shape of "the
# split lane ran monolith".
cp "${GITHUB_WORKSPACE}/pull-runtime/build-linux/libMobileGL.so" \
"${GITHUB_WORKSPACE}/build-linux/libMobileGL.so"
if nm --defined-only "${GITHUB_WORKSPACE}/build-linux/libMobileGL.so" | grep -q -i MG_Remote; then
echo "::error::the control's own library defines MG_Remote symbols, so it is not a pull build and this control would prove nothing"
exit 1
fi
export MOBILEGL_TRANSPORT=inproc
ctest -V --no-tests=error --timeout 10800 \
-R "^MobileGLTraceReplay\.${{ matrix.case }}\.${{ matrix.backend }}$"
control_rc=$?
set -e
if [ -d "${GOOD_OUTPUT}" ]; then
rm -rf "${{ matrix.case }}"; mv "${GOOD_OUTPUT}" "${{ matrix.case }}"
echo "restored the verified run's output over the control's"
fi
if [ "${control_rc}" -eq 0 ]; then
echo "::error::a PULL library passed the split retrace. OpenRA scores ssim 1.000000 under a monolith library too (measured), so the picture is not and cannot be this lane's gate - run_trace_case.cmake's transport-resolution assertion is, and it has stopped working. Every green in this job is then a monolith run under a name that says split."
exit 1
fi
echo "the pull library turned the split retrace red, as it must (ctest exit ${control_rc})"
# The refusal census, recorded rather than gated. run_trace_case.cmake already REDS the case
# on any Fatal{, so reaching here means the count is zero - but the number and the distinct
# slot names are what MEASUREMENTS wants from every split run, and reading them out of the
# library's own log is the only way to get them (ctest -V's transcript is a false zero).
- name: Refusal census from the library log
if: always()
working-directory: build-retrace/tools/trace_replay
run: |
log="${{ matrix.case }}/${{ matrix.backend }}/output/mobilegl.log"
if [ ! -f "${log}" ]; then
echo "no ${log} - the replay wrote no library log"
exit 0
fi
echo "Fatal{ lines: $(grep -c 'Fatal{' "${log}" || true)"
grep -o 'Fatal{[A-Za-z]*, "[^"]*"' "${log}" | sort | uniq -c | sort -rn | head -20 || true
grep -m1 "MOBILEGL_TRANSPORT=" "${log}" || echo "no transport line in ${log}"
- name: Upload core dumps
if: failure()
uses: actions/upload-artifact@v7
with:
name: retrace-split-core-dumps-${{ matrix.backend }}-${{ matrix.case }}
path: /tmp/core.*
if-no-files-found: ignore
- name: Upload actual image
if: always()
uses: actions/upload-artifact@v7
with:
name: retrace-split-result-${{ matrix.backend }}-${{ matrix.case }}
path: |
build-retrace/tools/trace_replay/${{ matrix.case }}/actual-images/**
build-retrace/tools/trace_replay/${{ matrix.case }}/${{ matrix.backend }}/output/**
if-no-files-found: warn
# G1's own job: the pull build must be the tree before P1, symbol for symbol and byte for byte.
# workflow_dispatch only - it builds the library twice from scratch, and its answer is about a
# BASELINE rather than about this push, so a per-push run would be measuring the wrong pair.