Skip to content

build: native build flags from #70 (asm, arch baseline, hardening) #71

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
dfa1 merged 3 commits into from
Jul 17, 2026
Merged

build: native build flags from #70 (asm, arch baseline, hardening) #71

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
3a4c32b
build: enable x86-64 BMI2 Huffman-decode assembly
dfa1 Jul 17, 2026
6b9e622
build: target ARMv8-A+CRC baseline on aarch64
dfa1 Jul 17, 2026
a55266c
build: full RELRO + immediate binding on Linux
dfa1 Jul 17, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 21 additions & 2 deletions scripts/build-zstd.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,14 +60,20 @@ echo "[build-zstd] Building zstd $CLASSIFIER$CROSS with zig cc (jobs=$JOBS)..."

# Library sources: core (common/compress/decompress) plus the dictionary
# builder (ZDICT_* API). legacy/ and deprecated/ are intentionally excluded.
# huf_decompress_amd64.S is hand-written BMI2 Huffman-decode asm; its body is
# entirely wrapped in `#if ZSTD_ENABLE_ASM_X86_64_BMI2` (gated on __x86_64__),
# so it compiles to a no-op object on every non-x86_64 target and only kicks
# in there via zstd's own runtime BMI2 detection (DYNAMIC_BMI2) — safe to
# always include rather than special-case it per classifier.
SRCS=$(find "$ZSTD_LIB/common" "$ZSTD_LIB/compress" "$ZSTD_LIB/decompress" \
"$ZSTD_LIB/dictBuilder" -name '*.c' | sort)
SRCS="$SRCS
$ZSTD_LIB/decompress/huf_decompress_amd64.S"

WORK="$(mktemp -d)"
trap 'rm -rf "$WORK"' EXIT

# Compile flags:
# -DZSTD_DISABLE_ASM=1 -> skip x86-only .S files, identical codegen every target
# -DZSTD_MULTITHREAD off -> single-threaded, no pthread dependency (hermetic)
# -DXXH_NAMESPACE -> matches zstd's own build, avoids xxhash symbol clashes
# ELF/Mach-O: -fvisibility=hidden + zstd's ZSTDLIB_VISIBLE keeps the surface
Expand All @@ -83,8 +89,21 @@ LINK_EXTRA=""
STRIP_FLAG="-s"
case "$CLASSIFIER" in
windows-*) VIS_FLAG=""; LINK_EXTRA="-Wl,--export-all-symbols"; STRIP_FLAG="" ;;
# Full RELRO + immediate binding: GOT is remapped read-only after startup
# relocation, closing off the classic GOT-overwrite exploit primitive.
# ELF-only (-z is a GNU ld/lld ELF flag; Mach-O/PE have no equivalent).
linux-*) LINK_EXTRA="-Wl,-z,relro,-z,now" ;;
esac
CFLAGS="-O3 -DNDEBUG -DZSTD_DISABLE_ASM=1 -DXXH_NAMESPACE=ZSTD_ $VIS_FLAG \

# Modern ARM baseline: ARMv8-A + CRC (zig's -mcpu syntax; clang's GCC-style
# -march=armv8-a+crc isn't accepted by zig's driver). "generic" keeps every
# aarch64 CPU supported, unlike pinning to e.g. apple_m1.
ARCH_FLAG=""
case "$CLASSIFIER" in
*-aarch64) ARCH_FLAG="-mcpu=generic+crc" ;;
esac

CFLAGS="-O3 $ARCH_FLAG -DNDEBUG -DXXH_NAMESPACE=ZSTD_ $VIS_FLAG \
-I$ZSTD_LIB -I$ZSTD_LIB/common -fPIC"

i=0
Expand Down
Loading