swung0x48/MobileGL
1
0
Fork 0
mirror of https://github.com/MobileGL-Dev/MobileGL synced 2026-09-07 19:58:32 +09:00
Code Issues Packages Projects Releases Wiki Activity

[CI] (Pipe): the third CI mode - a verify build, its integration and retrace lanes, and the two negative controls as always-on steps

Browse Source 
- build-linux-verify is a second Release/INFO build with -DMOBILEGL_PIPE_VERIFY=ON,
  because the comparator is a compile-time option and does not exist in the shipped
  library. It refuses to ship an artifact whose libMobileGL.so does not export
  MGPipeVerifyInputs and MGPipeFillForVerb: a typo'd -D is not an error in CMake,
  and every lane below would then be green having compared nothing.
- integration-verify runs the suite with the comparator armed and then proves it
  armed twice over: --no-tests=error reds a build whose verify entries were never
  registered, and a step greps every pipe-verify-*.log for the arming line.
- The two negative controls are steps of that job, not a manual exercise: a gate
  that can only be shown to work by someone remembering to break it has already
  stopped working. Each passes when ctest FAILS, and each first counts its own
  selection - an empty selection also exits non-zero under --no-tests=error, and a
  control that passed because it ran nothing would be worse than no control.
- Control B targets PoisonOmissionScenario.WithoutOmissionCompletes, the only
  integration entry in the tree that calls glGenerateMipmap at all; the case no
  longer skips itself when the omission knob is set, precisely so that the control
  has a green entry to turn red.
- retrace-verify replays the eight "verify": true cases against the verify library,
  copied over build-linux/libMobileGL.so because build-retrace freezes that absolute
  path into every case, with an nm check that the swap happened and an inverted
  OpenRA step that must go red under MOBILEGL_PIPE_VERIFY_CORRUPT.
  remove-artifact-clutter now waits for it: it deletes the trace fixtures these jobs
  download.
- monolith-symbol-report is G1 as a job: two pull builds with identical flags and
  LTO off, the baseline named by a workflow_dispatch input, symbol_report.py with
  both hard gates, and a refusal of any MG_Remote symbol in the monolith. It is
  dispatch-only because its answer is about a baseline, not about this push.
- pipe-gates gains the two --self-test steps. Regenerating and diffing cannot see a
  structural check that silently stopped checking; a broken gate and a clean tree
  produce the same green. Its dirty-surface comment now says P2, which is where
  ROADMAP.md:18 puts the first mapping round.
This commit is contained in:
swung0x48
2026-09-06 04:41:28 -04:00
parent 416cd23c28
commit 0fe7bf82d2
2 changed files with 588 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/test.yml
+572 -2
View File
@@ -7,6 +7,14 @@ on:
- Feat/Backend-Direct-GLES
- Feat/Backend-Direct-Vulkan
workflow_dispatch:
inputs:
baseline_sha:
description: >-
The commit monolith-symbol-report compares this tree against. P1's G1 says the pull
build is byte-identical to feat/disaggregated@087685d1, and that is what the default
names. The trigger set is unchanged: this job runs on workflow_dispatch only.
required: false
default: "087685d1"
jobs:
build-linux:
@@ -295,6 +303,293 @@ jobs:
path: /tmp/core.*
if-no-files-found: ignore
# THE THIRD CI MODE (ARCHITECTURE.md 13.2-(2)): the same library, built with the PipeInputs
# comparator compiled in, running the integration suite and a trace subset with two state models
# in one address space. It is a second build rather than a flag on the first because
# MOBILEGL_PIPE_VERIFY is a compile-time option - the snapshot, the entry compare and the
# compare-at-read hook do not exist in the shipped library, and are never meant to.
build-linux-verify:
runs-on: ubuntu-latest
timeout-minutes: 120
permissions:
actions: write
contents: read
env:
BUILD_DIR: build-verify
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:
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: Configure CMake
# Release/INFO like the shipped build on purpose. The poison arms in this configuration
# through MOBILEGL_PIPE_VERIFY (PipeInputs.h derives MOBILEGL_PIPE_POISON from it), so this
# job needs neither a Debug log level nor MOBILEGL_BUILD_DISAGGREGATED - and a Debug build
# would compare a different library from the one the other lanes measure.
run: |
cmake -S . -B "${BUILD_DIR}" -G Ninja \
-DCMAKE_C_COMPILER=clang-20 \
-DCMAKE_CXX_COMPILER=clang++-20 \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_BUILD_TYPE=Release \
-DMOBILEGL_LOG_ACTIVE_LEVEL=MOBILEGL_LOG_LEVEL_INFO \
-DMOBILEGL_BUILD_TEST=ON \
-DMOBILEGL_BUILD_BENCHMARK=OFF \
-DMOBILEGL_BUILD_INTEGRATION_TEST=ON \
-DMOBILEGL_ITEST_VK_ICD=/usr/share/vulkan/icd.d/lvp_icd.json \
-DMOBILEGL_BUILD_TRACE_REPLAY=OFF \
-DMOBILEGL_PIPE_VERIFY=ON \
-DCMAKE_POLICY_VERSION_MINIMUM=3.5
- name: Build
run: cmake --build "${BUILD_DIR}" --parallel "$(nproc)"
# The lane is worthless if the option silently did not take, and that is a one-character
# mistake away at all times (a typo'd -D is not an error in CMake). Two checks, both cheap:
# the comparator's entry point must be in the library, and the fill entry point with it.
- name: The verify library really carries the comparator
run: |
test -f "${BUILD_DIR}/libMobileGL.so"
nm -D --defined-only "${BUILD_DIR}/libMobileGL.so" | grep -q MGPipeVerifyInputs
nm -D --defined-only "${BUILD_DIR}/libMobileGL.so" | grep -q MGPipeFillForVerb
echo "libMobileGL.so exports MGPipeVerifyInputs and MGPipeFillForVerb"
- 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 verify 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-verify.tgz \
"${BUILD_DIR}/CTestTestfile.cmake" \
"${BUILD_DIR}/MobileGL/MG_Test" \
"${BUILD_DIR}/MobileGL/MG_IntegrationTest" \
"${SHARED_LIBS[@]}"
- name: Upload Linux verify runtime
uses: actions/upload-artifact@v7
with:
name: mobilegl-linux-runtime-verify
path: ci-artifacts/mobilegl-linux-runtime-verify.tgz
if-no-files-found: error
# The verify lane itself, plus the two negative controls that keep it falsifiable. The controls
# are ALWAYS-ON steps, not a manual exercise: a gate that can only be shown to work by someone
# remembering to break it on purpose is a gate that has already stopped working.
integration-verify:
runs-on: ubuntu-latest
timeout-minutes: 180
needs: build-linux-verify
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 verify runtime
uses: actions/download-artifact@v8
with:
name: mobilegl-linux-runtime-verify
path: .
- name: Unpack Linux verify runtime
run: |
tar -xzf mobilegl-linux-runtime-verify.tgz
test -f build-verify/libMobileGL.so
- name: Normalize CTest command paths
run: |
python - <<'PY'
from pathlib import Path
import re
for path in Path('build-verify').rglob('CTestTestfile.cmake'):
text = path.read_text()
text = re.sub(r'"[^"]*/cmake-[^"]*/bin/cmake"', '"cmake"', text)
path.write_text(text)
PY
- name: Integration scenarios under MOBILEGL_PIPE_VERIFY
working-directory: build-verify
# --no-tests=error is half the gate: the verify entries only exist when the library was
# configured with -DMOBILEGL_PIPE_VERIFY=ON, so a build that lost the option matches no
# tests and reds here instead of reporting a green run of nothing. The other half is
# PipeVerifyArmingScenario.Armed, which fails when the library never printed its arming
# line - the failure mode a bare `MOBILEGL_PIPE_VERIFY=1` cannot detect by itself.
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'
if [ "${{ secrets.ACTIONS_STEP_DEBUG }}" = "true" ]; then
ctest -V -L integration-verify --no-tests=error
else
ctest --output-on-failure -L integration-verify --no-tests=error
fi
- name: Every verify process really armed
working-directory: build-verify
run: |
shopt -s nullglob
logs=(MobileGL/MG_IntegrationTest/pipe-verify-*.log)
if [ ${#logs[@]} -eq 0 ]; then
echo "::error::the verify lane wrote no pipe-verify-*.log at all"
exit 1
fi
for log in "${logs[@]}"; do
if ! grep -q "MGPipe: verify armed" "${log}"; then
echo "::error::${log} carries no arming line: that lane ran without the comparator"
exit 1
fi
done
echo "arming line present in ${#logs[@]} lane log(s)"
# NEGATIVE CONTROL A (gate G4). The knob perturbs one field in the snapshot arm before the
# entry compare, so a working comparator must abort the run. This step passes when ctest
# FAILS - `if ctest ...; then error` - which is the only shape that can catch a comparator
# that silently compares nothing.
#
# The knob reaches the test process through the JOB environment: no ctest ENVIRONMENT
# property on the ambient Verify. entries names it (MG_IntegrationTest/CMakeLists.txt says
# so out loud), and a property entry would otherwise override this and the control would
# prove nothing. Same precedent as MOBILEGL_ESPRYT_DISABLE_INVALIDATE_FLUSH in `integration`.
- name: Negative control A - a corrupted snapshot field must turn the lane red
working-directory: build-verify
env:
MOBILEGL_ITEST_REQUIRE_GPU: "1"
MOBILEGL_PIPE_VERIFY_CORRUPT: GetRenderStateParameters
run: |
FILTER='DirectGLES\.Verify\..*ClearThenReadPixels'
# An empty selection would ALSO make ctest exit non-zero (--no-tests=error), and this
# step reads non-zero as "the control worked" - so the selection is counted first. A
# control that passes because it ran nothing is worse than no control.
matched=$(ctest -N -L integration-verify -R "${FILTER}" | grep -cE '^ *Test *#[0-9]+:')
if [ "${matched}" -lt 1 ]; then
echo "::error::negative control A selected ${matched} tests; its filter no longer matches anything"
exit 1
fi
if ctest --output-on-failure -L integration-verify -R "${FILTER}" --no-tests=error; then
echo "::error::MOBILEGL_PIPE_VERIFY_CORRUPT=GetRenderStateParameters left ${matched} verify entries GREEN. The comparator is not comparing, so every green entry above is green for no reason."
exit 1
fi
echo "the corrupted field turned ${matched} selected entries red, as it must"
# NEGATIVE CONTROL B (gate G5). The omission skips the STAMP of one field for one verb while
# still copying its value - indistinguishable from a fill row nobody wrote - so the poison
# must abort the glGenerateMipmap. Again: this step passes when ctest fails.
#
# The entry it targets is PoisonOmissionScenario.WithoutOmissionCompletes, which is green in
# the ambient lane above and is the ONLY integration entry in the tree that calls
# glGenerateMipmap at all. It deliberately does not skip itself when the knob is set, exactly
# so that this control has something to turn red.
- name: Negative control B - an omitted fill point must turn the lane red on that verb
working-directory: build-verify
env:
MOBILEGL_ITEST_REQUIRE_GPU: "1"
MOBILEGL_PIPE_POISON_OMIT: GenerateMipmap:GetActiveTextureUnit
run: |
FILTER='DirectGLES\.Verify\.PoisonOmissionScenario\.WithoutOmissionCompletes'
matched=$(ctest -N -L integration-verify -R "${FILTER}" | grep -cE '^ *Test *#[0-9]+:')
if [ "${matched}" -lt 1 ]; then
echo "::error::negative control B selected ${matched} tests; its filter no longer matches anything"
exit 1
fi
if ctest --output-on-failure -L integration-verify -R "${FILTER}" --no-tests=error; then
echo "::error::MOBILEGL_PIPE_POISON_OMIT=GenerateMipmap:GetActiveTextureUnit left the verify lane GREEN. The per-verb poison is not armed, so a forgotten fill row would ship silently."
exit 1
fi
echo "the omitted fill point turned the lane red, as it must"
- name: Upload verify lane logs
if: always()
uses: actions/upload-artifact@v7
with:
name: integration-verify-logs
path: build-verify/MobileGL/MG_IntegrationTest/pipe-*.log*
if-no-files-found: warn
- name: Upload core dumps
if: failure()
uses: actions/upload-artifact@v7
with:
name: integration-verify-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
@@ -562,6 +857,7 @@ jobs:
outputs:
matrix: ${{ steps.trace-cases.outputs.matrix }}
names: ${{ steps.trace-cases.outputs.names }}
verify-matrix: ${{ steps.trace-cases.outputs.verify-matrix }}
steps:
- name: Checkout repo
uses: actions/checkout@v6
@@ -571,6 +867,9 @@ jobs:
run: |
echo "matrix=$(python3 tools/trace_replay/trace_cases.py --ci --format github-test-matrix)" >> "$GITHUB_OUTPUT"
echo "names=$(python3 tools/trace_replay/trace_cases.py --ci --format names)" >> "$GITHUB_OUTPUT"
# The subset the verify build retraces ("verify": true in trace_cases.json). It is a
# SUBSET of the matrix above, so retrace-verify needs no fixtures of its own.
echo "verify-matrix=$(python3 tools/trace_replay/trace_cases.py --ci --format github-verify-matrix)" >> "$GITHUB_OUTPUT"
trace-fixtures:
name: trace fixture (${{ matrix.case }})
@@ -784,10 +1083,269 @@ jobs:
archive: false
if-no-files-found: error
# The trace half of the third CI mode. Same replay, same goldens, but the library underneath is
# the verify build and MOBILEGL_PIPE_VERIFY=1 is in the environment, so every backend read of
# frontend state is checked against a snapshot taken at the verb boundary. Eight cases rather
# than the full lane's 40 (tools/trace_replay/trace_cases.json, "verify": true): the comparator
# is budgeted at 5-10x, and the full sweep is a phase-exit / workflow_dispatch run.
retrace-verify:
name: retrace verify (${{ matrix.backend }}, ${{ matrix.case }})
runs-on: ubuntu-latest
timeout-minutes: 240
needs:
- build-linux-verify
- build-retrace
- trace-cases
- trace-fixtures
if: ${{ always() && needs.build-linux-verify.result == 'success' && needs.build-retrace.result == 'success' && needs.trace-cases.result == 'success' }}
strategy:
fail-fast: false
max-parallel: 4
matrix: ${{ fromJSON(needs.trace-cases.outputs.verify-matrix) }}
steps:
- name: Set Swap Space
uses: pierotofy/set-swap-space@v1.0
with:
swap-size-gb: 16
- name: Checkout repo
uses: actions/checkout@v6
- name: Download trace fixture
uses: actions/download-artifact@v8
with:
name: trace-fixture-${{ 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 verify runtime
uses: actions/download-artifact@v8
with:
name: mobilegl-linux-runtime-verify
path: .
- name: Download trace replay
uses: actions/download-artifact@v8
with:
name: mobilegl-trace-replay
path: .
- name: Unpack the VERIFY 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: the verify .so is put where that path points. The nm
# check is what makes the swap falsifiable - a run against the ordinary library would
# export no comparator, ignore MOBILEGL_PIPE_VERIFY entirely, and match its golden.
run: |
tar -xzf mobilegl-linux-runtime-verify.tgz
tar -xzf mobilegl-trace-replay.tgz
test -f build-verify/libMobileGL.so
test -f build-retrace/tools/trace_replay/mobilegl_trace_replay
mkdir -p build-linux
cp build-verify/libMobileGL.so build-linux/libMobileGL.so
nm -D --defined-only build-linux/libMobileGL.so | grep -q MGPipeVerifyInputs
echo "the library at build-linux/libMobileGL.so is the verify build"
- name: Retrace and validate under MOBILEGL_PIPE_VERIFY
working-directory: build-retrace/tools/trace_replay
# run_trace_case.cmake turns MOBILEGL_PIPE_VERIFY into three assertions of its own (the
# arming line, no Fatal{PipeVerifyDiffer, no Fatal{UnmigratedPipeInput), so a case that
# somehow ran the wrong library reds here instead of passing on its golden.
# --timeout 10800: the 1800s cases run 5-10x slower with both comparator arms live, which
# is well past ctest's 1500s default.
run: |
ulimit -c unlimited
sudo sysctl -w kernel.core_pattern='/tmp/core.%e.%p'
export MOBILEGL_PIPE_VERIFY=1
if [ '${{ matrix.backend }}' = 'DirectVulkan' ]; then
export MOBILEGL_MAGMA_R11G11B10F_FALLBACK=1
fi
if [ '${{ matrix.backend }}' = 'DirectVulkan' ] \
&& [ '${{ matrix.case }}' = 'improved-transparency-minecraft-26.3' ]; then
export MOBILEGL_MAGMA_DISABLE_BLENDED_DEPTH_WRITE=1
fi
ctest -V --no-tests=error --timeout 10800 \
-R '^MobileGLTraceReplay\.${{ matrix.case }}\.${{ matrix.backend }}$'
# The retrace lane's own always-on negative control, on one case so it costs one short trace:
# with a snapshot field corrupted, the SAME replay must fail. Without it, "40 traces, zero
# divergences" would be a statement about a comparator nobody watched.
- name: Negative control - a corrupted snapshot field must red this retrace
if: ${{ matrix.case == 'OpenRA' && matrix.backend == 'DirectGLES' }}
working-directory: build-retrace/tools/trace_replay
run: |
export MOBILEGL_PIPE_VERIFY=1
export MOBILEGL_PIPE_VERIFY_CORRUPT=GetRenderStateParameters
if ctest -V --no-tests=error --timeout 10800 \
-R '^MobileGLTraceReplay\.OpenRA\.DirectGLES$'; then
echo "::error::MOBILEGL_PIPE_VERIFY_CORRUPT=GetRenderStateParameters left the OpenRA retrace GREEN, so the comparator is not comparing and the whole verify retrace lane proves nothing."
exit 1
fi
echo "the corrupted field turned the retrace red, as it must"
- name: Upload core dumps
if: failure()
uses: actions/upload-artifact@v7
with:
name: retrace-verify-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-verify-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.
monolith-symbol-report:
name: monolith symbol report
runs-on: ubuntu-latest
timeout-minutes: 180
if: ${{ github.event_name == 'workflow_dispatch' }}
env:
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:
submodules: recursive
fetch-depth: 0
- 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: 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 binutils
# Both sides with IDENTICAL flags, LTO off, the same compiler and the same standard library:
# symbol_report.py's guard rails (scripts/symbol_report.py) say a mismatched pair "adds"
# thousands of symbols and the comparison then means nothing. The library alone - no tests,
# no benchmark, no integration test, no trace replay - because those targets do not ship.
- name: Build the baseline library (${{ inputs.baseline_sha }})
run: |
git worktree add ../baseline "${{ inputs.baseline_sha }}"
cd ../baseline
git submodule update --init --recursive
(cd 3rdparty/glslang && python update_glslang_sources.py)
cmake -S . -B build-sym-base -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=OFF -DMOBILEGL_BUILD_BENCHMARK=OFF \
-DMOBILEGL_BUILD_INTEGRATION_TEST=OFF -DMOBILEGL_BUILD_TRACE_REPLAY=OFF \
-DMOBILEGL_BUILD_DISAGGREGATED=OFF \
-DMOBILEGL_PIPE_PUSH=OFF -DMOBILEGL_PIPE_VERIFY=OFF \
-DMOBILEGL_ENABLE_LTO=OFF \
-DCMAKE_POLICY_VERSION_MINIMUM=3.5
cmake --build build-sym-base --parallel "$(nproc)"
cp build-sym-base/libMobileGL.so "${GITHUB_WORKSPACE}/libMobileGL-baseline.so"
- name: Build the head library
run: |
(cd 3rdparty/glslang && python update_glslang_sources.py)
cmake -S . -B build-sym-head -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=OFF -DMOBILEGL_BUILD_BENCHMARK=OFF \
-DMOBILEGL_BUILD_INTEGRATION_TEST=OFF -DMOBILEGL_BUILD_TRACE_REPLAY=OFF \
-DMOBILEGL_BUILD_DISAGGREGATED=OFF \
-DMOBILEGL_PIPE_PUSH=OFF -DMOBILEGL_PIPE_VERIFY=OFF \
-DMOBILEGL_ENABLE_LTO=OFF \
-DCMAKE_POLICY_VERSION_MINIMUM=3.5
cmake --build build-sym-head --parallel "$(nproc)"
# The monolith must not have grown a remote half. ARCHITECTURE.md:506: MG_Remote lives behind
# MOBILEGL_BUILD_DISAGGREGATED and nothing of it may reach a shipped pull build.
- name: No MG_Remote in the pull build
run: |
if nm --defined-only build-sym-head/libMobileGL.so | grep -q MG_Remote; then
echo "::error::the pull build defines MG_Remote symbols; the disaggregated half leaked into the monolith"
nm --defined-only build-sym-head/libMobileGL.so | grep MG_Remote | head -20
exit 1
fi
echo "no MG_Remote symbols in the pull build"
- name: Symbol report (G1)
run: |
python3 scripts/symbol_report.py \
--before libMobileGL-baseline.so \
--after build-sym-head/libMobileGL.so \
--threshold 0 \
--fail-on-symbol-set-change \
--fail-on-added-bytes 0 \
--markdown symbol-report.md \
--json symbol-report.json
- name: Upload the symbol report
if: always()
uses: actions/upload-artifact@v7
with:
name: monolith-symbol-report
path: |
symbol-report.md
symbol-report.json
if-no-files-found: error
remove-artifact-clutter:
name: remove artifact clutter
runs-on: ubuntu-latest
needs: retrace-summary
# (d) retrace-verify too: this job deletes the trace-fixture-* artifacts, and the verify
# retraces download the same ones.
needs:
- retrace-summary
- retrace-verify
if: always()
permissions:
actions: write
@@ -855,6 +1413,17 @@ jobs:
python3 scripts/gen_pipe.py
git diff --exit-code -- MobileGL/MG_Pipe/generated
# The generators' own negative controls: canned inputs that MUST trip each structural check
# (a field list that does not cover its struct's members, a verb set that is not the function
# table's). Regenerating and diffing above cannot see a check that silently stopped
# checking - a broken gate and a clean tree produce the same green.
- name: The MGPipe generators' checks can still fail
run: python3 scripts/gen_pipe.py --self-test
# The same question for the symbol tool the P1 gate is written in terms of.
- name: The symbol report's buckets and gates can still fail
run: python3 scripts/symbol_report.py --self-test
# Per-draw fprintf/printf instrumentation has repeatedly been committed by accident,
# once inside a mutex critical section. Nothing under these two trees prints to a
# stdio stream today - MGLOG_D compiles out in INFO builds and is the only channel
@@ -872,7 +1441,8 @@ jobs:
echo "no fprintf(stderr/stdout / printf( / puts( / std::cout|cerr under MobileGL/MG_Backend or MobileGL/MG_State"
# Informational: the frontend mutation surface an MGPipe aggregate generation has to
# cover. It becomes a gate in P1, when the mapping file exists to diff against.
# cover. It becomes a gate in P2, when the mapping file exists to diff against
# (ROADMAP.md:18 puts the first mapping round in P2, not P1).
- name: MGPipe dirty-surface report
run: python3 scripts/gen_pipe_dirty_surface.py --summary
MobileGL/MG_IntegrationTest/Scenarios/PoisonOmissionScenario.cpp
+16 -6
View File
@@ -335,6 +335,12 @@ void main() { o_color = vec4(0.25, 0.5, 0.75, 1.0); }
// The sibling control, in the ambient Verify. lanes: the same sequence with the knob UNSET
// must run to completion and log no Fatal at all.
//
// It deliberately does NOT skip when MOBILEGL_PIPE_POISON_OMIT is set. This is the entry
// CI's always-on negative control B exports the knob at: a green entry that the omission
// turns red is the whole proof that the poison is armed, and an entry that politely skipped
// itself would report that green either way. Nothing else in the integration suite calls
// glGenerateMipmap, so this case is also the only possible target for that control.
TEST_F(PoisonOmissionScenario, WithoutOmissionCompletes) {
if (!Ready()) return;
@@ -342,22 +348,26 @@ void main() { o_color = vec4(0.25, 0.5, 0.75, 1.0); }
GTEST_SKIP() << "the poison is only compiled into the push/verify builds; in an ordinary "
"build there is nothing for this control to be a control OF";
}
if (StringKnobIsSet("MOBILEGL_PIPE_POISON_OMIT")) {
GTEST_SKIP() << "MOBILEGL_PIPE_POISON_OMIT is armed for this process, so the abort is the "
"EXPECTED outcome here; OmittedFieldAbortsOnThatVerb owns that half and "
"runs in the PoisonOmitted. lane";
}
const bool omissionArmed = StringKnobIsSet("MOBILEGL_PIPE_POISON_OMIT");
int status = 0;
std::string reason;
ASSERT_TRUE(RunSequenceInAChildProcess(status, reason)) << reason;
const std::string childLog = ReadWholeFile(ChildLogPath());
const std::string note =
omissionArmed
? std::string(
" NOTE: MOBILEGL_PIPE_POISON_OMIT is set in this process, so this failure is "
"what CI's negative control B is asking for - the poison IS armed, and this "
"entry going red is the proof.")
: std::string();
ASSERT_TRUE(WIFEXITED(status))
<< "with no omission armed, a draw followed by glGenerateMipmap must complete; the child "
<< DescribeStatus(status)
<< ". If it aborted, the poison is firing on a field the verb's fill table SHOULD list - "
"add the row to MG_Pipe/FillPoints.def, never mark the field sticky. Child log:\n"
"add the row to MG_Pipe/FillPoints.def, never mark the field sticky."
<< note << " Child log:\n"
<< childLog;
EXPECT_EQ(WEXITSTATUS(status), 0) << "the child " << DescribeStatus(status)
<< ". Child log:\n"