ladybird/Toolchain/BuildIt.sh

302 lines
10 KiB
Bash
Raw Normal View History

#!/usr/bin/env bash
set -e
# This file will need to be run in bash, for now.
# === CONFIGURATION AND SETUP ===
2019-04-15 16:16:47 +03:00
DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
echo "$DIR"
ARCH=${ARCH:-"i686"}
TARGET="$ARCH-pc-serenity"
2019-04-15 16:16:47 +03:00
PREFIX="$DIR/Local"
BUILD=$(realpath "$DIR/../Build")
SYSROOT="$BUILD/Root"
2019-04-15 16:16:47 +03:00
MAKE="make"
MD5SUM="md5sum"
NPROC="nproc"
Travis: Reduce Toolchain cache item size by 73% Empirically, every single push or PR has to download *and then upload* about 3.6 GiB of "cache stuff", which takes up about 400 seconds: https://travis-ci.com/github/SerenityOS/serenity/builds/177500795 On every single push/PR! No matter what! Those 3.6 GB consist of: - 3.2 GB Toolchain cache (around 260 MB per compressed item) - 0.4 GB ccache, but is capped at 0.5 GB: https://travis-ci.com/github/BenWiederhake/serenity/builds/177528549 - (And 200 KB for some weird debian package? Dunno.) Investigating in the size, the Toolchain consists mostly of *DEBUG SYMBOLS IN THE COMPILER BINARIES* which comically misses the point. If we ever run into compiler crashes, any stacktrace would be lost anyway as soon as the Travis VM shuts down. Furthermore, Travis will only ever compile Serenity itself, and Serenity forbids C in it's Contribution Guidelines. That's another 20 MB we don't need to cache. Stripping the binaries and deleting the C compiler reduces the uncompressed size from 1200 MB down to 220 MB. The compressed size gets reduced from 260 MB to 70MB. That's a reduction of 73%. It'll take a while until the 'old' toolchains get deleted. I guess it'll take less than a week. From that point onward, the Travis cache will be 1.2 GB, consisting of: - 0.7 GB Toolchain cache - 0.5 GB ccache - (And that weird 200 KB deb file) If network speeds are linear, then this should reduce the "cache network overhead time" from about 400 seconds to about 120 seconds. tl;dr: Strip unnecessary debug infos, delete an unused files, and speed everything up by two minutes. (Both Toolchain cache hits and Toolchain rebuilds!)
2020-07-29 01:11:28 +03:00
# Each cache entry is 70 MB. 10 entries are 700 MiB.
# It seems that Travis starts having trouble around a total
# cache size of 9 GiB, so I think this is a good amount.
KEEP_CACHE_COUNT=10
if command -v ginstall &>/dev/null; then
INSTALL=ginstall
else
INSTALL=install
fi
if [ "$(uname -s)" = "OpenBSD" ]; then
MAKE=gmake
MD5SUM="md5 -q"
NPROC="sysctl -n hw.ncpuonline"
export CC=egcc
export CXX=eg++
export with_gmp=/usr/local
export LDFLAGS=-Wl,-z,notext
elif [ "$(uname -s)" = "FreeBSD" ]; then
2020-03-21 11:46:30 +03:00
MAKE=gmake
MD5SUM="md5 -q"
NPROC="sysctl -n hw.ncpu"
fi
git_patch=
while [ "$1" != "" ]; do
case $1 in
--dev ) git_patch=1
;;
esac
shift
done
echo PREFIX is "$PREFIX"
echo SYSROOT is "$SYSROOT"
2019-04-15 16:16:47 +03:00
mkdir -p "$DIR/Tarballs"
BINUTILS_VERSION="2.33.1"
BINUTILS_MD5SUM="1a6b16bcc926e312633fcc3fae14ba0a"
BINUTILS_NAME="binutils-$BINUTILS_VERSION"
BINUTILS_PKG="${BINUTILS_NAME}.tar.gz"
BINUTILS_BASE_URL="http://ftp.gnu.org/gnu/binutils"
2020-05-16 07:52:49 +03:00
GCC_VERSION="10.1.0"
GCC_MD5SUM="8a9fbd7e24d04c5d36e96bc894d3cd6b"
GCC_NAME="gcc-$GCC_VERSION"
GCC_PKG="${GCC_NAME}.tar.gz"
GCC_BASE_URL="http://ftp.gnu.org/gnu/gcc"
# === CHECK CACHE AND REUSE ===
pushd "$DIR"
if [ "${TRY_USE_LOCAL_TOOLCHAIN}" = "y" ] ; then
echo "Checking cached toolchain:"
DEPS_CONFIG="
uname=$(uname),TARGET=${TARGET},
BuildItHash=$($MD5SUM "$(basename "$0")"),
MAKE=${MAKE},MD5SUM=${MD5SUM},NPROC=${NPROC},
CC=${CC},CXX=${CXX},with_gmp=${with_gmp},LDFLAGS=${LDFLAGS},
BINUTILS_VERSION=${BINUTILS_VERSION},BINUTILS_MD5SUM=${BINUTILS_MD5SUM},
GCC_VERSION=${GCC_VERSION},GCC_MD5SUM=${GCC_MD5SUM}"
echo "Config is:${DEPS_CONFIG}"
if ! DEPS_HASH=$("$DIR/ComputeDependenciesHash.sh" "$MD5SUM" <<<"${DEPS_CONFIG}"); then
echo "Dependency hashing failed"
echo "Will rebuild toolchain from scratch, and NOT SAVE THE RESULT."
echo "Someone should look into this, but for now it'll work, albeit inefficient."
# Should be empty anyway, but just to make sure:
DEPS_HASH=""
elif [ -r "Cache/ToolchainLocal_${DEPS_HASH}.tar.gz" ] ; then
echo "Cache at Cache/ToolchainLocal_${DEPS_HASH}.tar.gz exists!"
echo "Extracting toolchain from cache:"
tar xzf "Cache/ToolchainLocal_${DEPS_HASH}.tar.gz"
echo "Done 'building' the toolchain."
exit 0
else
echo "Cache at Cache/ToolchainLocal_${DEPS_HASH}.tar.gz does not exist."
echo "Will rebuild toolchain from scratch, and save the result."
echo "But first, getting rid of old, outdated caches. Current caches:"
pushd "Cache/"
ls -l
# Travis preserves timestamps. Don't ask me why, but it does.
# We can exploit this to get an easy approximation of recent-ness.
# Our purging algorithm is simple: keep only the newest X entries.
# Note that `find` doesn't easily support ordering by date,
# and we control the filenames anyway.
# shellcheck disable=SC2012
ls -t | tail "-n+${KEEP_CACHE_COUNT}" | xargs -r rm -v
echo "After deletion:"
ls -l
popd
fi
fi
popd
# === DOWNLOAD AND PATCH ===
2019-04-15 16:16:47 +03:00
pushd "$DIR/Tarballs"
md5="$($MD5SUM $BINUTILS_PKG | cut -f1 -d' ')"
echo "bu md5='$md5'"
if [ ! -e $BINUTILS_PKG ] || [ "$md5" != ${BINUTILS_MD5SUM} ] ; then
rm -f $BINUTILS_PKG
curl -LO "$BINUTILS_BASE_URL/$BINUTILS_PKG"
else
echo "Skipped downloading binutils"
2019-04-15 16:16:47 +03:00
fi
md5="$($MD5SUM ${GCC_PKG} | cut -f1 -d' ')"
echo "gc md5='$md5'"
if [ ! -e $GCC_PKG ] || [ "$md5" != ${GCC_MD5SUM} ] ; then
rm -f $GCC_PKG
curl -LO "$GCC_BASE_URL/$GCC_NAME/$GCC_PKG"
else
echo "Skipped downloading gcc"
2019-04-15 16:16:47 +03:00
fi
if [ ! -d ${BINUTILS_NAME} ]; then
echo "Extracting binutils..."
tar -xzf ${BINUTILS_PKG}
2019-04-15 16:16:47 +03:00
pushd ${BINUTILS_NAME}
if [ "$git_patch" = "1" ]; then
git init > /dev/null
git add . > /dev/null
git commit -am "BASE" > /dev/null
git apply "$DIR"/Patches/binutils.patch > /dev/null
else
patch -p1 < "$DIR"/Patches/binutils.patch > /dev/null
fi
2019-04-15 16:16:47 +03:00
popd
else
echo "Skipped extracting binutils"
2019-04-15 16:16:47 +03:00
fi
if [ ! -d $GCC_NAME ]; then
echo "Extracting gcc..."
tar -xzf $GCC_PKG
pushd $GCC_NAME
if [ "$git_patch" = "1" ]; then
git init > /dev/null
git add . > /dev/null
git commit -am "BASE" > /dev/null
git apply "$DIR"/Patches/gcc.patch > /dev/null
else
2020-05-16 07:52:49 +03:00
patch -p1 < "$DIR/Patches/gcc.patch" > /dev/null
fi
2019-04-15 16:16:47 +03:00
popd
else
echo "Skipped extracting gcc"
2019-04-15 16:16:47 +03:00
fi
if [ "$(uname)" = "Darwin" ]; then
pushd "gcc-${GCC_VERSION}"
./contrib/download_prerequisites
popd
fi
2019-04-15 16:16:47 +03:00
popd
# === COMPILE AND INSTALL ===
mkdir -p "$PREFIX"
2019-04-15 16:16:47 +03:00
mkdir -p "$DIR/Build/binutils"
mkdir -p "$DIR/Build/gcc"
if [ -z "$MAKEJOBS" ]; then
MAKEJOBS=$($NPROC)
fi
2019-04-15 16:16:47 +03:00
pushd "$DIR/Build/"
unset PKG_CONFIG_LIBDIR # Just in case
pushd binutils
echo "XXX configure binutils"
"$DIR"/Tarballs/binutils-2.33.1/configure --prefix="$PREFIX" \
--target="$TARGET" \
--with-sysroot="$SYSROOT" \
--enable-shared \
--disable-nls \
${TRY_USE_LOCAL_TOOLCHAIN:+"--quiet"} || exit 1
if [ "$(uname)" = "Darwin" ]; then
# under macOS generated makefiles are not resolving the "intl"
# dependency properly to allow linking its own copy of
# libintl when building with --enable-shared.
"$MAKE" -j "$MAKEJOBS" || true
pushd intl
"$MAKE" all-yes
popd
fi
echo "XXX build binutils"
"$MAKE" -j "$MAKEJOBS" || exit 1
"$MAKE" install || exit 1
2019-04-15 16:16:47 +03:00
popd
pushd gcc
if [ "$(uname -s)" = "OpenBSD" ]; then
perl -pi -e 's/-no-pie/-nopie/g' "$DIR/Tarballs/gcc-$GCC_VERSION/gcc/configure"
fi
echo "XXX configure gcc and libgcc"
"$DIR/Tarballs/gcc-$GCC_VERSION/configure" --prefix="$PREFIX" \
--target="$TARGET" \
--with-sysroot="$SYSROOT" \
--disable-nls \
--with-newlib \
--enable-shared \
--enable-languages=c,c++ \
${TRY_USE_LOCAL_TOOLCHAIN:+"--quiet"} || exit 1
echo "XXX build gcc and libgcc"
"$MAKE" -j "$MAKEJOBS" all-gcc all-target-libgcc || exit 1
echo "XXX install gcc and libgcc"
"$MAKE" install-gcc install-target-libgcc || exit 1
echo "XXX serenity libc and libm"
mkdir -p "$BUILD"
pushd "$BUILD"
CXXFLAGS="-DBUILDING_SERENITY_TOOLCHAIN" cmake ..
cmake --build . --target LibC
"$INSTALL" -D Libraries/LibC/libc.a Libraries/LibM/libm.a Root/usr/lib/
SRC_ROOT=$(realpath "$DIR"/..)
FILES=$(find "$SRC_ROOT"/Libraries/LibC "$SRC_ROOT"/Libraries/LibM -name '*.h' -print)
for header in $FILES; do
target=$(echo "$header" | sed -e "s@$SRC_ROOT/Libraries/LibC@@" -e "s@$SRC_ROOT/Libraries/LibM@@")
$INSTALL -D "$header" "Root/usr/include/$target"
done
unset SRC_ROOT
popd
echo "XXX build libstdc++"
"$MAKE" all-target-libstdc++-v3 || exit 1
echo "XXX install libstdc++"
"$MAKE" install-target-libstdc++-v3 || exit 1
if [ "$(uname -s)" = "OpenBSD" ]; then
cd "$DIR/Local/libexec/gcc/i686-pc-serenity/$GCC_VERSION" && ln -sf liblto_plugin.so.0.0 liblto_plugin.so
fi
2019-04-15 16:16:47 +03:00
popd
popd
# == SAVE TO CACHE ==
pushd "$DIR"
if [ "${TRY_USE_LOCAL_TOOLCHAIN}" = "y" ] ; then
echo "Caching toolchain:"
if [ -z "${DEPS_HASH}" ] ; then
echo "NOT SAVED, because hashing failed."
echo "It's computed in the beginning; see there for the error message."
elif [ -e "Cache/ToolchainLocal_${DEPS_HASH}.tar.gz" ] ; then
# Note: This checks for *existence*. Initially we checked for
# *readability*. If Travis borks permissions, there's not much we can do.
echo "Cache exists but was not used?!"
echo "Not touching cache then."
else
mkdir -p Cache/
Travis: Reduce Toolchain cache item size by 73% Empirically, every single push or PR has to download *and then upload* about 3.6 GiB of "cache stuff", which takes up about 400 seconds: https://travis-ci.com/github/SerenityOS/serenity/builds/177500795 On every single push/PR! No matter what! Those 3.6 GB consist of: - 3.2 GB Toolchain cache (around 260 MB per compressed item) - 0.4 GB ccache, but is capped at 0.5 GB: https://travis-ci.com/github/BenWiederhake/serenity/builds/177528549 - (And 200 KB for some weird debian package? Dunno.) Investigating in the size, the Toolchain consists mostly of *DEBUG SYMBOLS IN THE COMPILER BINARIES* which comically misses the point. If we ever run into compiler crashes, any stacktrace would be lost anyway as soon as the Travis VM shuts down. Furthermore, Travis will only ever compile Serenity itself, and Serenity forbids C in it's Contribution Guidelines. That's another 20 MB we don't need to cache. Stripping the binaries and deleting the C compiler reduces the uncompressed size from 1200 MB down to 220 MB. The compressed size gets reduced from 260 MB to 70MB. That's a reduction of 73%. It'll take a while until the 'old' toolchains get deleted. I guess it'll take less than a week. From that point onward, the Travis cache will be 1.2 GB, consisting of: - 0.7 GB Toolchain cache - 0.5 GB ccache - (And that weird 200 KB deb file) If network speeds are linear, then this should reduce the "cache network overhead time" from about 400 seconds to about 120 seconds. tl;dr: Strip unnecessary debug infos, delete an unused files, and speed everything up by two minutes. (Both Toolchain cache hits and Toolchain rebuilds!)
2020-07-29 01:11:28 +03:00
# We *most definitely* don't need debug symbols in the linker/compiler.
# This cuts the uncompressed size from 1.2 GiB per Toolchain down to about 250 MiB.
pushd "Local/libexec/gcc/i686-pc-serenity/${GCC_VERSION}"
2020-07-29 16:22:02 +03:00
for binary in cc1 cc1plus lto1; do
Travis: Reduce Toolchain cache item size by 73% Empirically, every single push or PR has to download *and then upload* about 3.6 GiB of "cache stuff", which takes up about 400 seconds: https://travis-ci.com/github/SerenityOS/serenity/builds/177500795 On every single push/PR! No matter what! Those 3.6 GB consist of: - 3.2 GB Toolchain cache (around 260 MB per compressed item) - 0.4 GB ccache, but is capped at 0.5 GB: https://travis-ci.com/github/BenWiederhake/serenity/builds/177528549 - (And 200 KB for some weird debian package? Dunno.) Investigating in the size, the Toolchain consists mostly of *DEBUG SYMBOLS IN THE COMPILER BINARIES* which comically misses the point. If we ever run into compiler crashes, any stacktrace would be lost anyway as soon as the Travis VM shuts down. Furthermore, Travis will only ever compile Serenity itself, and Serenity forbids C in it's Contribution Guidelines. That's another 20 MB we don't need to cache. Stripping the binaries and deleting the C compiler reduces the uncompressed size from 1200 MB down to 220 MB. The compressed size gets reduced from 260 MB to 70MB. That's a reduction of 73%. It'll take a while until the 'old' toolchains get deleted. I guess it'll take less than a week. From that point onward, the Travis cache will be 1.2 GB, consisting of: - 0.7 GB Toolchain cache - 0.5 GB ccache - (And that weird 200 KB deb file) If network speeds are linear, then this should reduce the "cache network overhead time" from about 400 seconds to about 120 seconds. tl;dr: Strip unnecessary debug infos, delete an unused files, and speed everything up by two minutes. (Both Toolchain cache hits and Toolchain rebuilds!)
2020-07-29 01:11:28 +03:00
echo "Before: $(du -h "${binary}")"
strip "${binary}"
echo "After: $(du -h "${binary}")"
done
popd
binary=Local/bin/i686-pc-serenity-lto-dump
echo "Before: $(du -h "${binary}")"
strip "${binary}"
echo "After: $(du -h "${binary}")"
tar czf "Cache/ToolchainLocal_${DEPS_HASH}.tar.gz" Local/
fi
fi
popd