Cockatrice/.ci/compile.sh
Lukas Brübach 6f3297a4a0 [CI] Size compiler caches for one configuration and split the Fedora leg
The bounds in #7667 were set below what one build configuration writes, so
ccache filled to its limit and evicted the entries the next build was about to
ask for. The probe measured what that costs: a 6G cache on one leg served 1059
of 1059 calls and the 700M bound on the same leg served 14%. The quota also
went to legs that finish under the Windows leg, which has no compiler cache.

Debug information is most of an object's size. A measured Arch Debug
configuration is 1.2 GB with the built-in flags and 660 MB at -g1; Debian 13
Release is 1.1 GB and 614 MB. CI therefore builds every non-tag leg at level 1
through a new COCKATRICE_DEBUG_INFO_LEVEL option, which stays empty by default
so local, distro and release builds are unchanged. Level 1 keeps file and line,
which is what a failing test or a sanitizer backtrace reports. Tags and the
nightly legs keep full DWARF for the Sentry symbols.

With one configuration at half the size, the quota goes to fewer legs at a size
that holds one whole configuration, and the legs that duplicate another leg's
coverage give their cache up: Servatrice, Ubuntu 26.04 behind the sanitize leg
of the same distro, macOS 15 and macOS 26 Debug behind macOS 14. Their keys
leave LIVE_CCACHE_KEYS, so prune-caches reclaims the existing entries.

Fedora 44 built Debug for the suite and Release for the package in one job, so
the package compile queued behind the debug one and the leg took 30.8 min - the
only leg above Windows - while one 1.6 GB cache held two configurations. It is
split into two legs with one cache each, running in parallel.

The option reaches compile.sh inside the container through docker.sh, and
compile.sh turns it into the CMake flag.

Bounds: Arch 850M, Debian 13 800M, Fedora 44 debug 850M and release 800M,
Ubuntu 26.04 sanitize 1200M, macOS 14 800M. That is 5.3 GB of compiler cache
on top of roughly 4.5 GB of buildkit, vcpkg, Qt and CodeQL entries.

Windows still sets the pull-request wall clock: 29.5 min, 23.7 of it compiling
with no compiler cache, because ccache and sccache cannot store the 711 of 1054
translation units that use PCH under MSVC (#7381). That is a separate trial.
2026-11-04 13:52:05 +01:00

853 lines
37 KiB
Bash
Executable file

#!/bin/bash
# This script is to be used by the ci environment from the project root directory, do not use it from somewhere else.
# Compiles cockatrice inside of a ci environment
# --install runs make install
# --package [<package type>] runs make package, optionally force the type
# --suffix <suffix> renames package with this suffix, requires arg
# --server compiles servatrice
# --test runs tests
# --sanitize builds with AddressSanitizer and UndefinedBehaviorSanitizer, gcc/clang only
# --coverage instruments the test run for line coverage and prints a gcovr summary, gcc only
# --debug or --release sets the build type ie CMAKE_BUILD_TYPE
# --ccache [<size>] uses ccache and shows stats, optionally provide size
# --evict-ccache <age> runs ccache eviction based on given age after build
# --dir <dir> sets the name of the build dir, default is "build"
# --cmake-generator <generator> sets CMAKE_GENERATOR as used by cmake
# --target-macos-version <version> sets the min os version - only used for macOS builds
# --profile writes an MSBuild performance summary and binary log, Windows only
# uses env: BUILDTYPE MAKE_INSTALL MAKE_PACKAGE PACKAGE_TYPE PACKAGE_SUFFIX MAKE_SERVER MAKE_NO_CLIENT MAKE_TEST USE_CCACHE CCACHE_SIZE CCACHE_EVICTION_AGE BUILD_DIR CMAKE_GENERATOR TARGET_MACOS_VERSION MAKE_SANITIZE MAKE_COVERAGE VCPKG_TARGET_MACOS_VERSION BUILD_PARALLEL_LEVEL BUILD_PROFILE
# (correspond to args: --debug/--release --install --package <package type> --suffix <suffix> --server --test --sanitize --coverage --ccache <ccache_size> --dir <dir> --profile)
# exitcode: 1 for failure, 3 for invalid arguments
# Read arguments
while [[ $# != 0 ]]; do
case "$1" in
'--')
shift
;;
'--install')
MAKE_INSTALL=1
shift
;;
'--package')
MAKE_PACKAGE=1
shift
if [[ $# != 0 && ${1:0:1} != - ]]; then
PACKAGE_TYPE="$1"
shift
fi
;;
'--suffix')
shift
if [[ $# == 0 ]]; then
echo "::error file=$0::--suffix expects an argument"
exit 3
fi
PACKAGE_SUFFIX="$1"
shift
;;
'--server')
MAKE_SERVER=1
shift
;;
'--no-client')
MAKE_NO_CLIENT=1
shift
;;
'--test')
MAKE_TEST=1
shift
;;
'--sanitize')
MAKE_SANITIZE=1
shift
;;
'--coverage')
MAKE_COVERAGE=1
shift
;;
'--debug')
BUILDTYPE="Debug"
shift
;;
'--release')
BUILDTYPE="Release"
shift
;;
'--ccache')
USE_CCACHE=1
shift
if [[ $# != 0 && ${1:0:1} != - ]]; then
CCACHE_SIZE="$1"
shift
fi
;;
'--evict-ccache')
shift
if [[ $# == 0 ]]; then
echo "::error file=$0::--evict-ccache expects an argument"
exit 3
fi
CCACHE_EVICTION_AGE=$1
shift
;;
'--vcpkg')
USE_VCPKG=1
shift
;;
'--dir')
shift
if [[ $# == 0 ]]; then
echo "::error file=$0::--dir expects an argument"
exit 3
fi
BUILD_DIR="$1"
shift
;;
'--cmake-generator')
shift
if [[ $# == 0 ]]; then
echo "::error file=$0::--cmake-generator expects an argument"
exit 3
fi
export CMAKE_GENERATOR=$1
shift
;;
'--profile')
BUILD_PROFILE=1
shift
;;
'--target-macos-version')
shift
if [[ $# == 0 ]]; then
echo "::error file=$0::--target-macos-version expects an argument"
exit 3
fi
TARGET_MACOS_VERSION="$1"
shift
;;
*)
echo "::error file=$0::unrecognized option: $1"
exit 3
;;
esac
done
set -e
# Setup
./servatrice/check_schema_version.sh
if [[ ! $BUILDTYPE ]]; then
BUILDTYPE=Release
fi
if [[ ! $BUILD_DIR ]]; then
BUILD_DIR="build"
fi
mkdir -p "$BUILD_DIR"
if [[ $MAKE_COVERAGE ]]; then
# Resolved while the working directory is still the source root. $0 is relative to the
# directory the script was invoked from, so after the cd below it no longer resolves, and
# under `set -e` the failed subshell would abort the build rather than just yielding an empty
# variable. Only the coverage report needs this, so only coverage pays for it.
SRC_ROOT="$(cd "$(dirname "$0")/.." && pwd)"
fi
cd "$BUILD_DIR"
# Set minimum CMake Version
export CMAKE_POLICY_VERSION_MINIMUM=3.10
# Add cmake flags
flags=("-DCMAKE_BUILD_TYPE=$BUILDTYPE")
if [[ $MAKE_SERVER ]]; then
flags+=("-DWITH_SERVER=1")
fi
if [[ $MAKE_NO_CLIENT ]]; then
flags+=("-DWITH_CLIENT=0" "-DWITH_ORACLE=0")
fi
if [[ $MAKE_TEST ]]; then
flags+=("-DTEST=1")
fi
if [[ $MAKE_SANITIZE ]]; then
flags+=("-DENABLE_SANITIZERS=ON")
else
# Every leg states the flag explicitly rather than relying on the CMake
# default, so a build directory reused from a sanitized configuration can
# never leak libasan/libubsan into a shipped binary.
flags+=("-DENABLE_SANITIZERS=OFF")
fi
if [[ $MAKE_COVERAGE ]]; then
flags+=("-DENABLE_COVERAGE=ON")
fi
if [[ $USE_CCACHE ]]; then
flags+=("-DUSE_CCACHE=1")
if [[ $CCACHE_SIZE ]]; then
# note, this setting persists after running the script
ccache --max-size "$CCACHE_SIZE"
fi
fi
if [[ $PACKAGE_TYPE ]]; then
flags+=("-DCPACK_GENERATOR=$PACKAGE_TYPE")
fi
if [[ $USE_VCPKG ]]; then
flags+=("-DUSE_VCPKG=1")
flags+=("-DVCPKG_INSTALL_OPTIONS=--x-abi-tools-use-exact-versions")
fi
if [[ ${WITH_CRASH_REPORTING:-} == "ON" ]]; then
# Release CI opts into the sentry-native SDK (default is OFF so from-source, distro and
# offline package builds never fetch it). Debug builds stay SDK-free so crashes from
# developer builds never reach Sentry.
flags+=("-DWITH_CRASH_REPORTING=ON")
fi
if [[ ${COCKATRICE_DEBUG_INFO_LEVEL:-} ]]; then
# CI lowers the DWARF level so one build configuration fits its compiler-cache bound;
# see the option in CMakeLists.txt. Empty keeps the built-in flags.
flags+=("-DCOCKATRICE_DEBUG_INFO_LEVEL=$COCKATRICE_DEBUG_INFO_LEVEL")
fi
if [[ ${BUILDTYPE:-Release} == "Release" && -n ${COCKATRICE_SENTRY_DSN:-} ]]; then
# Release-only: Debug builds stay DSN-free so crashes from developer builds
# never reach Sentry. Empty (local builds, forks without the secret, and non-tag
# CI builds, which never receive the secret) keeps today's mock fallback via the
# compiled-in default. CI passes the value from the SENTRY_DSN secret on tag
# Release builds only, so pre-release test artifacts never report to Sentry.
flags+=("-DCOCKATRICE_SENTRY_DSN=$COCKATRICE_SENTRY_DSN")
fi
# Add cmake --build flags
buildflags=(--config "$BUILDTYPE")
function ccachestatsverbose() {
# note, verbose only works on newer ccache, discard the error
local got
if got="$(ccache --show-stats --verbose 2>/dev/null)"; then
echo "$got"
else
ccache --show-stats
fi
}
function cpuinfo() {
# The number of cores decides how much build parallelism is useful; the hosted
# runners are not all sized alike, so read it off the machine instead of assuming.
case "$RUNNER_OS" in
Windows)
echo "cores: ${NUMBER_OF_PROCESSORS:-unknown}"
;;
macOS)
echo "cores: $(sysctl -n hw.ncpu)"
;;
*)
echo "cores: $(nproc 2>/dev/null || echo unknown)"
;;
esac
}
# Compile
if [[ $RUNNER_OS == macOS ]]; then
# QTDIR is needed for macOS since we actually only use the cached thin Qt binaries instead of the install-qt-action,
# which sets a few environment variables
if QTDIR=$(find "$GITHUB_WORKSPACE/Qt" -depth -maxdepth 2 -name macos -type d -print -quit); then
echo "found QTDIR at $QTDIR"
else
echo "could not find QTDIR!"
exit 2
fi
# the qtdir is located at Qt/[qtversion]/macos
# we use find to get the first subfolder with the name "macos"
# this works independent of the qt version as there should be only one version installed on the runner at a time
export QTDIR
if [[ $TARGET_MACOS_VERSION ]]; then
# CMAKE_OSX_DEPLOYMENT_TARGET is a vanilla cmake flag needed to compile to target macOS version
flags+=("-DCMAKE_OSX_DEPLOYMENT_TARGET=$TARGET_MACOS_VERSION")
# vcpkg dependencies need a vcpkg triplet file to compile to the target macOS version
# an easy way is to copy the x64-osx.cmake file and modify it
triplets_dir="/tmp/cmake/triplets"
triplet_version="custom-triplet"
triplet_file="$triplets_dir/$triplet_version.cmake"
arch=$(uname -m)
if [[ $arch == x86_64 ]]; then
arch="x64"
fi
mkdir -p "$triplets_dir"
triplet_source="../vcpkg/triplets/$arch-osx.cmake"
if [[ ! -f "$triplet_source" ]]; then
triplet_source="../vcpkg/triplets/community/$arch-osx.cmake"
fi
cp "$triplet_source" "$triplet_file"
# The triplet is what vcpkg hashes into the ABI hash of every port, so two legs that
# agree here reuse each other's binary cache. Targets 14, 15 and 26 on the same Xcode
# produce identical binaries when they share the lowest deployment target, so they all
# build against 14 to keep one vcpkg cache scope instead of three.
# Keep in sync with "vcpkg_scope" and "vcpkg_deployment_target" in desktop-build.yml.
vcpkg_target="${VCPKG_TARGET_MACOS_VERSION:-$TARGET_MACOS_VERSION}"
echo "set(VCPKG_CMAKE_SYSTEM_VERSION $vcpkg_target)" >>"$triplet_file"
echo "set(VCPKG_OSX_DEPLOYMENT_TARGET $vcpkg_target)" >>"$triplet_file"
flags+=("-DVCPKG_OVERLAY_TRIPLETS=$triplets_dir")
flags+=("-DVCPKG_HOST_TRIPLET=$triplet_version")
flags+=("-DVCPKG_TARGET_TRIPLET=$triplet_version")
echo "::group::Generated triplet $triplet_file"
cat "$triplet_file"
echo "::endgroup::"
fi
echo "::group::Signing Certificate"
if [[ -n "$MACOS_CERTIFICATE_NAME" ]]; then
echo "$MACOS_CERTIFICATE" | base64 --decode >"certificate.p12"
security create-keychain -p "$MACOS_CI_KEYCHAIN_PWD" build.keychain
security default-keychain -s build.keychain
security set-keychain-settings -t 3600 -l build.keychain
security unlock-keychain -p "$MACOS_CI_KEYCHAIN_PWD" build.keychain
security import certificate.p12 -k build.keychain -P "$MACOS_CERTIFICATE_PWD" -T /usr/bin/codesign
security set-key-partition-list -S apple-tool:,apple:,codesign: -s -k "$MACOS_CI_KEYCHAIN_PWD" build.keychain
echo "macOS signing certificate successfully imported and keychain configured."
else
echo "No signing certificate configured. Skipping set up of keychain in macOS environment."
fi
echo "::endgroup::"
if [[ $MAKE_PACKAGE ]]; then
# Workaround https://github.com/actions/runner-images/issues/7522
# have hdiutil repeat the command 10 times in hope of success.
# The wrapper has to exit non-zero once it gives up: CPack reads that status to
# decide whether the disk image was created, so a zero exit turns a failed hdiutil
# into a much later, misleading "Problem copying the package" error that hides
# hdiutil's own message.
hdiutil_script="/tmp/hdiutil.sh"
# shellcheck disable=SC2016
echo '#!/bin/bash
attempts=10
i=0
while true; do
hdiutil "$@" && exit 0
rc=$?
if (( ++i >= attempts )); then
echo "::error::hdiutil $1 failed $i times (last status $rc), giving up" >&2
exit "$rc"
fi
echo "hdiutil $1 attempt $i/$attempts failed (status $rc), retrying" >&2
sleep 1
done' >"$hdiutil_script"
chmod +x "$hdiutil_script"
flags+=(-DCPACK_COMMAND_HDIUTIL="$hdiutil_script")
fi
elif [[ $RUNNER_OS == Windows ]]; then
if [[ $CMAKE_GENERATOR == "Visual Studio"* ]]; then
# cmake only hands MSBuild a job count when one is asked for, and without it
# MSBuild builds a single node: one project at a time, whatever MTT does inside
# it. Ask for a node per core.
if [[ ! $BUILD_PARALLEL_LEVEL ]]; then
BUILD_PARALLEL_LEVEL="${NUMBER_OF_PROCESSORS:-1}"
fi
buildflags+=(--parallel "$BUILD_PARALLEL_LEVEL")
# Enable MTT, see https://devblogs.microsoft.com/cppblog/improved-parallelism-in-msbuild/
# and https://devblogs.microsoft.com/cppblog/cpp-build-throughput-investigation-and-tune-up/#multitooltask-mtt
# EnforceProcessCountAcrossBuilds makes the CL_MPCount ceiling apply across all
# the projects the nodes are building, instead of each node running its own
# cl.exe processes and /m and MTT multiplying into an oversubscribed machine.
buildflags+=(-- -p:UseMultiToolTask=true -p:EnableClServerMode=true -p:EnforceProcessCountAcrossBuilds=true -p:CL_MPCount="$BUILD_PARALLEL_LEVEL")
if [[ $BUILD_PROFILE ]]; then
# A performance summary says which projects and tasks cost what, the binary log
# keeps the full timeline; both are far too noisy to always produce. The files
# land next to the solution, i.e. in the build dir the workflow uploads. These
# join the MTT options above, which already carried the "--" separator.
buildflags+=("-flp:PerformanceSummary;v=q;LogFile=msbuild-perf.log" -bl:msbuild.binlog)
fi
else
# Ninja runs the compiler itself, so it neither takes the MSBuild switches
# (it rejects -p: options outright) nor a job count: it sizes its own pool
# for the cores it finds.
echo "ninja $(ninja --version)"
fi
fi
if [[ $USE_CCACHE ]]; then
echo "::group::Show ccache stats"
ccachestatsverbose
echo "::endgroup::"
fi
echo "::group::Configure cmake"
cmake --version
echo "Running cmake with flags: ${flags[*]}"
cmake .. "${flags[@]}"
echo "::endgroup::"
if [[ $USE_CCACHE ]]; then
# PCH-aware caching is required or ccache refuses to cache any TU that consumes a
# precompiled header, silently recompiling everything on every run. The mtime and ctime
# entries cover files created during the build itself: the moc/uic/rcc output and the
# protobuf headers are generated after ccache starts, so without them every TU including
# one is read as a race and never cached. Checked-out sources are older than the compile
# that follows them and do not trip the check.
#
# This has to run after configure: CMakeLists.txt sets sloppiness=pch_defines,time_macros
# at configure time, and --set-config replaces the whole list rather than extending it, so
# setting it earlier would leave the build below with two entries. Re-asserting all four
# here is what actually reaches the compiler; the order is the point.
ccache --set-config sloppiness=pch_defines,time_macros,include_file_mtime,include_file_ctime
fi
echo "::group::Build project"
cpuinfo
echo "Running cmake --build with flags: ${buildflags[*]}"
cmake --build . "${buildflags[@]}"
echo "::endgroup::"
if [[ $BUILD_PROFILE && $RUNNER_OS == Windows && -f msbuild-perf.log ]]; then
# Repeat the summary in the job log so the numbers are readable without
# downloading the binary log, which is hundreds of megabytes.
echo "::group::MSBuild performance summary"
cat msbuild-perf.log
echo "::endgroup::"
fi
if [[ $USE_CCACHE ]]; then
if [[ $CCACHE_EVICTION_AGE ]]; then
echo "::group::evict ccache files older than $CCACHE_EVICTION_AGE"
ccache --evict-older-than "$CCACHE_EVICTION_AGE"
echo "::endgroup::"
fi
echo "::group::Show ccache stats again"
ccachestatsverbose
echo "::endgroup::"
elif [[ $CCACHE_EVICTION_AGE ]]; then
echo "::error file=$0::ccache eviction is enabled while ccache is disabled!"
fi
if [[ $RUNNER_OS == macOS ]]; then
echo "::group::Inspect Mach-O binaries"
for app in cockatrice oracle servatrice; do
binary="$GITHUB_WORKSPACE/build/$app/$app.app/Contents/MacOS/$app"
echo "Inspecting $app..."
vtool -show-build "$binary"
file "$binary"
lipo -info "$binary"
echo ""
done
echo "::endgroup::"
fi
if [[ $MAKE_TEST ]]; then
echo "::group::Run tests"
if [[ $MAKE_SANITIZE ]]; then
# detect_leaks=0: Qt and the sanitizers' own runtimes both leak at exit by design, and
# LeakSanitizer's default is aggressive enough to bury our real findings under hundreds
# of process-teardown reports. Malloc/free misuse and use-after-free are what ASan is for
# and they stay fully on.
# abort_on_error=1: a bad access leaves the process in an unusable state, so continuing
# only produces a second, more confusing report from the wreckage.
# UBSAN halt_on_error=1: without it UBSan prints "runtime error: ..." but the
# process exits 0 (recover mode), so gtest/ctest report a pass and the job
# stays green. Combined with -fno-sanitize-recover=undefined at compile time,
# the first finding aborts the test process with a non-zero exit.
export ASAN_OPTIONS="detect_leaks=0:abort_on_error=1:print_stacktrace=1:symbolize=1"
export UBSAN_OPTIONS="print_stacktrace=1:halt_on_error=1:symbolize=1"
fi
# Every TEST() is its own CTest entry, so the suite is process-spawn bound rather than CPU
# bound and -j recovers nearly all of the per-process overhead. The test binaries are safe to
# run concurrently: the only one that claims a fixed resource is single_instance_manager_test,
# which already scopes its local socket name to its own pid so parallel copies cannot collide.
ctest -C "$BUILDTYPE" --output-on-failure -j "${CTEST_JOBS:-4}"
echo "::endgroup::"
fi
if [[ $MAKE_COVERAGE ]]; then
# gcovr is not a build dependency and is not available everywhere, so a missing one is reported
# and skipped rather than failing an otherwise green build.
echo "::group::Coverage report"
GCOVR="$(command -v gcovr || true)"
if [[ -z $GCOVR ]]; then
echo "gcovr not found; install it to produce a coverage report"
else
"$GCOVR" --version
# This script configures with `cmake ..`, so the build dir it was started from is the object
# directory and the source root is one level up. Both are passed explicitly: gcovr otherwise
# walks the current directory and trips over in-tree build directories left by other tooling,
# whose stale .gcno files cannot be resolved. --config is named too, so the filters apply no
# matter what directory gcovr happens to start from.
GCOVR_ARGS=("." --root "$SRC_ROOT" --object-directory "." --config "$SRC_ROOT/gcovr.cfg")
"$GCOVR" "${GCOVR_ARGS[@]}" --txt --decisions
"$GCOVR" "${GCOVR_ARGS[@]}" --html-details coverage-details.html --html-title "Cockatrice line coverage"
"$GCOVR" "${GCOVR_ARGS[@]}" --lcov -o coverage.lcov
"$GCOVR" "${GCOVR_ARGS[@]}" --cobertura-pretty -o coverage.xml
"$GCOVR" "${GCOVR_ARGS[@]}" --json-summary-pretty -o coverage-summary.json
echo "Coverage artifacts written to coverage-details.html, coverage.lcov, coverage.xml and coverage-summary.json"
fi
echo "::endgroup::"
fi
if [[ $RUNNER_OS == Windows && ! $MAKE_NO_CLIENT ]]; then
echo "::group::Probe the installer's holder lookup"
# The working directory is still the build dir here: the package steps below are what
# leave and re-enter it, and that is what made a $BUILD_DIR-prefixed path name
# build/build. Both paths are relative to the dir the probe was built into.
probe="./installer-holder-probe.exe"
if [[ ! -f $probe ]]; then
echo "::error file=$0::installer-holder-probe.exe was not built"
exit 1
fi
# The probe exists to name a holder that a process-and-module scan cannot see: one
# that has a file open without having loaded it. Holding a temporary DLL with
# FileShare.None is exactly that case, so this asserts the reason the probe was
# added instead of only that the Restart Manager answers at all. It needs no NSIS
# and no packaging, so it runs on every Windows build that gets this far.
# The holder is a process of its own: the probe never names the process that started it,
# which is what keeps an in-place uninstaller from being named as a holder of the
# directory it is emptying. The same DLL held by the script that starts the probe
# therefore has to come back as nothing, and that is the second case below.
probe_check="./holder-probe-check.ps1"
cat >"$probe_check" <<'POWERSHELL'
$ErrorActionPreference = "Stop"
$dir = Join-Path ([IO.Path]::GetTempPath()) "cockatrice-holder-probe-$PID"
Remove-Item -Recurse -Force $dir -ErrorAction SilentlyContinue
New-Item -ItemType Directory -Force $dir | Out-Null
$probe = (Resolve-Path -LiteralPath $env:PROBE_EXE).Path
function Invoke-Probe([string] $guardDir) {
$start = New-Object System.Diagnostics.ProcessStartInfo
$start.FileName = $probe
$start.UseShellExecute = $false
$start.RedirectStandardOutput = $true
$start.EnvironmentVariables["COCKATRICE_GUARD_DIR"] = $guardDir
$process = [System.Diagnostics.Process]::Start($start)
$read = $process.StandardOutput.ReadToEndAsync()
if (-not $process.WaitForExit(20000)) {
try { $process.Kill() } catch { }
throw "the holder probe did not finish within 20 s for $guardDir"
}
$output = $read.Result
$code = $process.ExitCode
$process.Dispose()
return @{ Code = $code; Output = $output }
}
function Wait-UntilLocked([string] $path) {
$deadline = (Get-Date).AddSeconds(15)
while ((Get-Date) -lt $deadline) {
try {
$open = [IO.File]::Open($path, "Open", "ReadWrite", "None")
$open.Dispose()
} catch {
return
}
Start-Sleep -Milliseconds 100
}
throw "the holder never locked $path"
}
try {
$file = Join-Path $dir "locked.dll"
[IO.File]::WriteAllText($file, "x")
# The holder is a process of its own, not this script: the probe never names the process
# that started it (the case below asserts that), so a file held here would be left out on
# purpose and the check would pass for the wrong reason.
$holderScript = Join-Path $dir "hold.ps1"
Set-Content -LiteralPath $holderScript -Value @'
param([string] $Path)
$handle = [IO.File]::Open($Path, "Open", "ReadWrite", "None")
Start-Sleep -Seconds 120
'@
$powershell = Join-Path $env:SystemRoot "System32\WindowsPowerShell\v1.0\powershell.exe"
$holder = Start-Process -FilePath $powershell -PassThru -ArgumentList @(
"-NoProfile", "-ExecutionPolicy", "Bypass", "-File", "`"$holderScript`"", "-Path", "`"$file`"")
try {
Wait-UntilLocked $file
$locked = Invoke-Probe $dir
if ($locked.Code -ne 1) { throw "expected exit 1 for a locked file, got $($locked.Code) : $($locked.Output)" }
if ($locked.Output -notmatch "pid\s+$($holder.Id)\b") {
throw "the lookup did not name the process holding the file: $($locked.Output)"
}
} finally {
try { $holder.Kill() } catch { }
try { $holder.WaitForExit(5000) | Out-Null } catch { }
}
# The probe is started by the installer, or by an uninstaller that runs in place from the
# directory it is emptying, so the asking process holds files there itself. Naming it would
# ask the user to close the program that is asking - a question neither the close nor the
# end offer can answer, so the guard repeated it until the update was cancelled. A holder
# that is only the asking process has to come back as nothing.
$handle = [IO.File]::Open($file, "Open", "ReadWrite", "None")
try {
$self = Invoke-Probe $dir
} finally {
$handle.Dispose()
}
if ($self.Code -ne 0) { throw "expected exit 0 when the asking process is the only holder, got $($self.Code) : $($self.Output)" }
if ($self.Output -match "pid\s+$PID\b") { throw "the lookup named the process that started it: $($self.Output)" }
# A payload executable that refuses to be replaced is running, whether or not this probe
# may name it: the Restart Manager does not report every mapping, and the image path of a
# process under another account cannot be read from here, so both process lookups can
# come back empty while the program is still there. Held here by the asking process, so
# that it cannot be named - the answer has to be "could not name it", never "nothing is
# holding it", which is what sent an update into an uninstall that could only fail
# halfway. Exit 3 is the code the guard answers by asking the user to close it and retry.
$image = Join-Path $dir "oracle.exe"
[IO.File]::WriteAllText($image, "x")
$handle = [IO.File]::Open($image, "Open", "ReadWrite", "None")
try {
$unnameable = Invoke-Probe $dir
} finally {
$handle.Dispose()
}
if ($unnameable.Code -ne 3) {
throw "expected exit 3 for a payload executable that is in use but cannot be named, got $($unnameable.Code) : $($unnameable.Output)"
}
if ($unnameable.Output -notmatch "oracle\.exe is in use") {
throw "the answer did not name the executable it could not replace: $($unnameable.Output)"
}
$empty = Join-Path $dir "empty"
New-Item -ItemType Directory -Force $empty | Out-Null
$nothing = Invoke-Probe $empty
if ($nothing.Code -ne 0) { throw "expected exit 0 for a directory with nothing open, got $($nothing.Code) : $($nothing.Output)" }
# A process running from the directory is a holder even when it has no file open: an
# instance holds its payload as mapped images, not as handles, and that is the state
# the guard used to miss - it passed, the purge failed, and the install ended in a
# warning with no offer to close and try again. The image is a copy of a native
# Windows program, so it starts without its usual directory next to it, and it carries
# one of the payload's own names because that is what the lookup names.
$copy = Join-Path $dir "cockatrice.exe"
Copy-Item -LiteralPath (Join-Path $env:SystemRoot "System32\cmd.exe") -Destination $copy
$running = Start-Process -FilePath $copy -ArgumentList "/c", "ping -n 20 127.0.0.1 >nul" -PassThru
try {
Start-Sleep -Milliseconds 1500
$byImage = Invoke-Probe $dir
} finally {
try { $running.Kill() } catch { }
# Wait for the image to be released, or the cleanup below leaves the copy behind.
try { $running.WaitForExit(5000) | Out-Null } catch { }
}
if ($byImage.Code -ne 1) { throw "expected exit 1 for a process running from the directory, got $($byImage.Code) : $($byImage.Output)" }
if ($byImage.Output -notmatch "cockatrice\.exe \(pid\s+$($running.Id)") {
throw "the lookup did not name the process running from the directory: $($byImage.Output)"
}
# A program in a folder below the directory is not an instance of the installation, even
# when it carries a payload program's name: the install directory can be a folder the
# user keeps other applications in, each in its own subfolder, and naming one of those
# asked for a close the probe never sends and an end it refuses to perform, so the guard
# could only abort the update. A program that really does hold a payload file is still
# named, by the Restart Manager lookup exercised above.
$nested = Join-Path $dir "other"
New-Item -ItemType Directory -Force $nested | Out-Null
$nestedImage = Join-Path $nested "cockatrice.exe"
Copy-Item -LiteralPath (Join-Path $env:SystemRoot "System32\cmd.exe") -Destination $nestedImage
$nestedRunning = Start-Process -FilePath $nestedImage -ArgumentList "/c", "ping -n 20 127.0.0.1 >nul" -PassThru
try {
Start-Sleep -Milliseconds 1500
$inSubfolder = Invoke-Probe $dir
} finally {
try { $nestedRunning.Kill() } catch { }
try { $nestedRunning.WaitForExit(5000) | Out-Null } catch { }
}
if ($inSubfolder.Code -ne 0) {
throw "expected exit 0 for a program in a folder below the directory, got $($inSubfolder.Code) : $($inSubfolder.Output)"
}
# A program the user keeps in the directory itself is not a payload program either: the
# guard ends only what the payload installs, and naming anything else asks for an answer
# that cannot free a single file.
$foreign = Join-Path $dir "tool.exe"
Copy-Item -LiteralPath (Join-Path $env:SystemRoot "System32\cmd.exe") -Destination $foreign
$foreignRunning = Start-Process -FilePath $foreign -ArgumentList "/c", "ping -n 20 127.0.0.1 >nul" -PassThru
try {
Start-Sleep -Milliseconds 1500
$ignored = Invoke-Probe $dir
} finally {
try { $foreignRunning.Kill() } catch { }
try { $foreignRunning.WaitForExit(5000) | Out-Null } catch { }
}
if ($ignored.Code -ne 0) {
throw "expected exit 0 for a program the payload does not install, got $($ignored.Code) : $($ignored.Output)"
}
Write-Host "holder lookup named pid $($holder.Id) for a file another process holds, named nothing for a file only the asking process holds, reported a payload executable it could not name as in use instead of free, named nothing for an empty directory, named pid $($running.Id) for a payload program running from the directory, and named nothing for a payload program in a folder below it or for a program the payload does not install"
} finally {
Remove-Item -Recurse -Force $dir -ErrorAction SilentlyContinue
}
POWERSHELL
if ! PROBE_EXE="$probe" powershell -NoProfile -File "$probe_check"; then
rm -f "$probe_check"
exit 1
fi
rm -f "$probe_check"
echo "::endgroup::"
fi
if [[ $MAKE_INSTALL ]]; then
echo "::group::Install"
cmake --build . --target install --config "$BUILDTYPE"
echo "::endgroup::"
fi
if [[ $MAKE_PACKAGE ]]; then
echo "::group::Create package"
cmake --build . --target package --config "$BUILDTYPE"
echo "::endgroup::"
if [[ $PACKAGE_SUFFIX ]]; then
echo "::group::Update package name"
cd ..
BUILD_DIR="$BUILD_DIR" .ci/name_build.sh "$PACKAGE_SUFFIX"
echo "::endgroup::"
fi
if [[ $RUNNER_OS == Windows ]]; then
echo "::group::Check installer for build-tree artifacts"
cd "$BUILD_DIR"
package="$(find . -maxdepth 1 -type f -name 'Cockatrice-*.exe' -print -quit)"
if [[ ! $package ]]; then
echo "::error file=$0::Could not find installer to inspect"
exit 1
fi
seven_zip="$(command -v 7z || true)"
if [[ ! $seven_zip ]]; then
seven_zip="/c/Program Files/7-Zip/7z.exe"
fi
if [[ ! -f $seven_zip ]]; then
echo "::warning file=$0::7-Zip not found, skipping installer content check"
else
echo "Inspecting $package"
# Fail the build if the installer contains any path left behind by the MSBuild or
# Qt AUTOMOC tooling (build-tree artifacts must live in the build dir, not the install)
if "$seven_zip" l "$package" |
grep -E "_autogen|\.dir[\\/]|\.tlog|(^|[\\/])x64[\\/]|(^|[\\/])\.qt[\\/]|(^|[\\/])\.qsb[\\/]|(^|[\\/])\.lupdate[\\/]|CMakeFiles"; then
echo "::error file=$0::Installer contains build-tree artifacts"
exit 1
fi
echo "Installer content is clean"
# Fail the build if the installer does not carry the runtime the
# applications link against - a file that must be there, or a DLL that must
# not be. The runtime DLLs reach the package by globbing every *.dll out of
# the build output directories, so anything that build produced can ride
# along and nothing else in the build objects to it: a Qt module the client
# links statically (Qt6Multimedia.dll, Qt6QuickWidgets.dll) can be left out,
# and the test framework's gtest.dll can be shipped. Either only surfaces on
# a user's machine after an update has already replaced their install, as
# "The code execution cannot proceed because Qt6Multimedia.dll was not
# found" or "abseil_dll.dll is either not designed to run on Windows or it
# contains an error". required-runtime.txt is written by the top level
# CMakeLists.txt from the same module lists the targets are built against
# plus the vcpkg bin directory the app-local copies come from.
if [[ ! -f required-runtime.txt ]]; then
echo "::error file=$0::required-runtime.txt not found in the build dir, cannot verify the runtime"
exit 1
fi
# 7-Zip writes CRLF on Windows, so the \r has to come off before any
# whole-line comparison below. Normalize separators to /, and strip the
# install-root prefixes an NSIS listing may carry, so the entries can be
# compared as exact relative paths. -xF keeps a '.' in a DLL name from
# being a wildcard; -i because that is how the installed tree behaves -
# the uninstaller clears both $INSTDIR\plugins and $INSTDIR\Plugins.
# The archive side and the manifest side have to agree on line endings.
# CMake writes required-runtime.txt with the platform's line ending and
# 7-Zip writes CRLF on Windows, so a trailing \r survives on $entry and
# defeats the whole-line match - every entry then reads as missing even
# though the listing holds it verbatim.
# [$] rather than \$ so the literal dollar is a bracket expression and the
# linter does not read these as unexpanded shell variables (SC2016).
installer_paths="$("$seven_zip" l -slt "$package" | tr -d '\r' |
sed -n 's/^Path = //p' |
sed -e 's|\\|/|g' -e 's|^[$]INSTDIR/||' -e 's|^[$]OUTDIR/||' -e 's|^[$]PLUGINSDIR/||' -e 's|^\./||' -e 's|^/||')"
manifest="$(tr -d '\r' <required-runtime.txt)"
missing=""
checked=0
while IFS= read -r entry; do
if [[ -z $entry || $entry == \#* ]]; then
continue
fi
checked=$((checked + 1))
if ! grep -qxiF "$entry" <<<"$installer_paths"; then
missing+=" $entry"
fi
done <<<"$manifest"
if [[ -n $missing ]]; then
echo "::error file=$0::Installer is missing required runtime files:$missing"
# Print both sides of the comparison. A mismatch here has twice been
# caused by the shape of 7-Zip's output rather than by a missing file,
# and a bare "missing" line cannot tell those apart.
echo "required-runtime.txt asked for $checked path(s); 7-Zip listed $(grep -c '' <<<"$installer_paths") path(s)"
# Indent with parameter expansion rather than sed 's/^/ /', which the
# linter reports as SC2001.
echo "--- required ---"
echo "${manifest//$'\n'/$'\n '}"
echo "--- first 40 paths 7-Zip lists in the installer ---"
listed_head="$(head -40 <<<"$installer_paths")"
echo "${listed_head//$'\n'/$'\n '}"
exit 1
fi
echo "Installer carries the Qt runtime of every module the three applications link ($checked files)"
# Now the other direction: a top level DLL in the package that the manifest
# does not name. The missing check above cannot see a runtime that was
# never expected.
#
# Two things are reported here and neither may fail the build.
#
# The test framework is a fact rather than a guess - no shipped executable
# links gtest or gmock, and the build output directories the "*.dll" glob
# reads are shared with the test targets, so gtest.dll and gmock.dll are
# one filter away from shipping in every release. Even so it is reported
# instead of enforced: a check that has never passed on a real package has
# not earned the right to block a release, and the first Windows run of an
# enforcing version is exactly when its own wrong assumptions show up.
#
# Everything else is expected to be there. required-runtime.txt names the
# Qt modules the three applications link, and the protobuf, Abseil,
# OpenSSL, liblzma, ffmpeg and Qt debug DLLs in the package come in on the
# build output glob instead, because nothing in the build declares them by
# name. So this list is a report of what the package carries, not a verdict.
# Read it when a dependency is added; do not fail on it.
forbidden=""
unexpected=""
while IFS= read -r entry; do
[[ -z $entry ]] && continue
# Top level only. A DLL under Plugins/ or a nested image format plugin is
# a plugin the platform loads by name, not a runtime the executable
# imports, and requiring one here would list all of them every run.
[[ $entry == */* ]] && continue
if grep -qxiF "$entry" <<<"$manifest"; then
continue
fi
# NSIS extracts its plugins to $PLUGINSDIR, which the path normalization
# above strips, so they arrive here looking like payload files. They are
# part of the installer, not of the installed tree.
if [[ $entry == [Nn][Ss][Dd][Ii][Aa][Ll][Oo][Gg][Ss].dll ||
$entry == [Nn][Ss][Ee][Xx][Ee][Cc].dll ||
$entry == [Ss][Yy][Ss][Tt][Ee][Mm].dll ||
$entry == [Uu][Ss][Ee][Rr][Ii][Nn][Ff][Oo].dll ||
$entry == [Ii][Nn][Ee][Tt].dll ]]; then
continue
fi
if [[ $entry == [Gg][Tt][Ee][Ss][Tt]* || $entry == [Gg][Mm][Oo][Cc][Kk]* ]]; then
forbidden+=" $entry"
else
unexpected+=" $entry"
fi
done < <(grep -i '\.dll$' <<<"$installer_paths")
if [[ -n $forbidden ]]; then
echo "::warning file=$0::Installer contains test framework DLLs that no shipped executable links:$forbidden"
echo "A build output directory shared with the test targets has leaked into the package."
fi
if [[ -n $unexpected ]]; then
echo "Top level DLLs in the package beyond the linked Qt modules:$unexpected"
echo "These arrive on the build output glob, not from required-runtime.txt, so they are reported and not required."
fi
if [[ -z $forbidden && -z $unexpected ]]; then
echo "Installer carries no runtime files beyond the linked Qt modules"
fi
fi
echo "::endgroup::"
fi
fi