Files
httrack/.github/workflows/ci.yml
Xavier Roche c1a8c5ffa8 ci: install git-clang-format and shfmt from apt, drop the github.com downloads
Both linters fetched a tool over the network. The format job pulled the
git-clang-format driver from raw.githubusercontent.com, which 429 rate-limits
the shared runner egress IPs; a 429 failed the job and left the cache empty, so
every later run cold-missed and 429'd again. The lint job similarly fetched the
shfmt release binary from github.com.

Both are unnecessary. The clang-format-19 package already installed ships the
matching git-clang-format driver (/usr/bin/git-clang-format-19); symlink it to
the unsuffixed name. And ubuntu-24.04 (noble) ships shfmt 3.8.0 in universe,
exactly the pinned version, so install it from apt too. This drops both fetches,
both actions/cache steps, and the LLVM_TAG / SHFMT_VERSION env: no network call,
nothing to rate-limit. Each tool's version now tracks its apt package, same as
clang-format itself.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: Xavier Roche <roche@httrack.com>
2026-06-16 23:09:04 +02:00

396 lines
14 KiB
YAML

# Build and test on x86-64 and arm64, and lint the shell scripts.
name: CI
on:
push:
branches: [master]
pull_request:
workflow_dispatch:
# Least privilege: the workflow only needs to read the repo.
permissions:
contents: read
# Cancel superseded runs on the same branch or PR.
concurrency:
group: ci-${{ github.ref }}
cancel-in-progress: true
jobs:
build:
name: build (${{ matrix.arch }}, ${{ matrix.cc }})
runs-on: ${{ matrix.runner }}
strategy:
fail-fast: false
matrix:
include:
- { arch: x86-64, runner: ubuntu-24.04, cc: gcc }
- { arch: x86-64, runner: ubuntu-24.04, cc: clang }
- { arch: arm64, runner: ubuntu-24.04-arm, cc: gcc }
- { arch: arm64, runner: ubuntu-24.04-arm, cc: clang }
env:
CC: ${{ matrix.cc }}
steps:
- uses: actions/checkout@v6
with:
submodules: recursive
- name: Install build dependencies
run: |
set -euo pipefail
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
build-essential clang autoconf automake libtool autoconf-archive \
zlib1g-dev libssl-dev
- name: Configure
run: |
set -euo pipefail
# Regenerate from configure.ac/Makefile.am to validate them; the
# committed generated files already let a plain checkout build.
autoreconf -fi
./configure
- name: Build
run: make -j"$(nproc)"
- name: Test
run: make check
- name: Print the test log on failure
if: failure()
run: cat tests/test-suite.log 2>/dev/null || true
# Portability: build and test on macOS (Darwin/clang) on a native runner --
# no VM. The tree has no __APPLE__ branches, so Darwin exercises the
# generic-Unix path on a second libc and kernel. brew's openssl@3 is keg-only,
# so point configure at it; everything else is in the SDK or default paths.
macos:
name: build (macOS arm64, clang)
runs-on: macos-14
steps:
- uses: actions/checkout@v6
with:
submodules: recursive
- name: Install build dependencies
run: |
set -euo pipefail
brew install autoconf automake libtool autoconf-archive
- name: Configure
run: |
set -euo pipefail
ssl="$(brew --prefix openssl@3)"
autoreconf -fi
./configure CPPFLAGS="-I${ssl}/include" LDFLAGS="-L${ssl}/lib"
- name: Build
run: make -j"$(sysctl -n hw.ncpu)"
- name: Test
run: make check
- name: Print the test log on failure
if: failure()
run: cat tests/test-suite.log 2>/dev/null || true
# Portability/hardening: 32-bit (i386) build on the x86-64 runner via multilib
# -- no extra hardware. Exercises the 32-bit size_t/pointer ABI, where size
# and bounds math can truncate or wrap in ways 64-bit never reveals (the axis
# the overflow-safe bounds work targets). --build (not --host) keeps configure
# out of cross mode, so the i386 binary still runs the test suite here.
linux-i386:
name: build (linux i386, gcc -m32)
runs-on: ubuntu-24.04
steps:
- uses: actions/checkout@v6
with:
submodules: recursive
- name: Install build dependencies (multilib + 32-bit libs)
run: |
set -euo pipefail
sudo dpkg --add-architecture i386
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
build-essential gcc-multilib autoconf automake libtool \
autoconf-archive zlib1g-dev:i386 libssl-dev:i386
- name: Configure
run: |
set -euo pipefail
autoreconf -fi
./configure --build=i686-pc-linux-gnu CC="gcc -m32"
- name: Build
run: make -j"$(nproc)"
- name: Test
run: make check
- name: Print the test log on failure
if: failure()
run: cat tests/test-suite.log 2>/dev/null || true
# Memory safety: build and run the suite under AddressSanitizer +
# UndefinedBehaviorSanitizer. The offline engine self-tests drive the parsers
# that chew on untrusted crawled input (charset, mime, HTML, entities, IDNA,
# filters, cache) straight through the sanitizers, so a buffer overrun,
# use-after-free, or signed overflow there fails the build instead of slipping
# past a plain -O2 build. gcc's runtimes; one job is enough (the bug class is
# arch-independent and the matrix already covers compile portability).
sanitize:
name: sanitize (ASan+UBSan, gcc)
runs-on: ubuntu-24.04
steps:
- uses: actions/checkout@v6
with:
submodules: recursive
- name: Install build dependencies
run: |
set -euo pipefail
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
build-essential autoconf automake libtool autoconf-archive \
zlib1g-dev libssl-dev
- name: Configure (sanitized)
run: |
set -euo pipefail
autoreconf -fi
./configure CC=gcc \
CFLAGS="-fsanitize=address,undefined -fno-sanitize-recover=all -g -O1 -fno-omit-frame-pointer" \
LDFLAGS="-fsanitize=address,undefined"
- name: Build
run: make -j"$(nproc)"
- name: Test (sanitized)
# Leaks at exit are out of scope (the CLI frees little on the way out);
# we want memory-safety errors, so turn leak detection off and make every
# other finding abort the run.
#
# Poison fresh allocations with 0xCA and freed blocks with 0xCB (decimal
# 202/203) so memory never reads back as accidental zeros: a missing-NUL
# fread buffer then runs strlen off into the redzone instead of stopping
# at a lucky zero. Distinct bytes tell the two apart in a dump (0xCA =
# uninitialized, 0xCB = use-after-free). ASan caps its malloc fill at 4096
# bytes by default, so max_malloc_fill_size lifts it to cover large cache
# buffers; free_fill flags use-after-free reads.
env:
ASAN_OPTIONS: detect_leaks=0:abort_on_error=1:halt_on_error=1:strict_string_checks=1:malloc_fill_byte=202:max_malloc_fill_size=2147483647:free_fill_byte=203:max_free_fill_size=2147483647
UBSAN_OPTIONS: print_stacktrace=1:halt_on_error=1
run: make check
- name: Print the test log on failure
if: failure()
run: cat tests/test-suite.log 2>/dev/null || true
# Optional-dependency build: compile and test with HTTPS/OpenSSL disabled --
# the configuration users on minimal systems build, and one libssl is not even
# installed here so configure cannot silently re-enable it. The matrix above
# always has libssl, so the #if HTS_USEOPENSSL branches would otherwise never
# be compiled and could rot unnoticed.
no-ssl:
name: build (no openssl, --disable-https)
runs-on: ubuntu-24.04
steps:
- uses: actions/checkout@v6
with:
submodules: recursive
- name: Install build dependencies (no libssl)
run: |
set -euo pipefail
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
build-essential autoconf automake libtool autoconf-archive zlib1g-dev
- name: Configure (https disabled)
run: |
set -euo pipefail
autoreconf -fi
./configure --disable-https
- name: Build
run: make -j"$(nproc)"
- name: Test
run: make check
- name: Print the test log on failure
if: failure()
run: cat tests/test-suite.log 2>/dev/null || true
# Validate the Debian packaging via the same script maintainers release with.
# One amd64/gcc run is enough: packaging (control/rules/manifest/lintian/quilt
# source build) is arch- and compiler-independent, and the build matrix above
# already covers compile portability. lintian runs with --fail-on=error.
deb:
name: deb package (lintian)
runs-on: ubuntu-24.04
steps:
- uses: actions/checkout@v6
with:
submodules: recursive
- name: Install packaging toolchain
run: |
set -euo pipefail
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
build-essential autoconf automake libtool autoconf-archive \
zlib1g-dev libssl-dev \
debhelper devscripts lintian fakeroot
# --unsigned: CI has no GPG key (also skips the release sig/checksums).
# debuild builds every package, then lintian gates on errors.
#
# DEB_BUILD_OPTIONS trims work CI does not need (release builds via
# mkdeb.sh are untouched): noautodbgsym drops the -dbgsym packages whose
# LTO payloads are slow to compress and that CI never ships; parallel uses
# every core. We let debuild run its test pass -- the only one now that
# mkdeb no longer runs its own -- so CI exercises the packaged tests.
- name: Build Debian packages
run: |
export DEB_BUILD_OPTIONS="noautodbgsym parallel=$(nproc)"
bash tools/mkdeb.sh --unsigned --no-release-artifacts
# Release-tarball integrity: `make distcheck` rolls the dist tarball, then
# configures, builds and tests it out-of-tree from a read-only source tree and
# checks nothing is left behind. Catches a file referenced in *_SOURCES or
# EXTRA_DIST but missing from the tarball -- the same "ships broken to users"
# class as a stale committed Makefile.in.
distcheck:
name: distcheck (release tarball)
runs-on: ubuntu-24.04
steps:
- uses: actions/checkout@v6
with:
submodules: recursive
- name: Install build dependencies
run: |
set -euo pipefail
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
build-essential autoconf automake libtool autoconf-archive \
zlib1g-dev libssl-dev
- name: distcheck
run: |
set -euo pipefail
autoreconf -fi
./configure
make -j"$(nproc)" distcheck
dco:
name: DCO sign-off
# Only checkable on a PR, where we have the base..head commit range.
if: github.event_name == 'pull_request'
runs-on: ubuntu-24.04
steps:
- uses: actions/checkout@v6
with:
fetch-depth: 0
- name: Every commit must be signed off
env:
BASE: ${{ github.event.pull_request.base.sha }}
HEAD: ${{ github.event.pull_request.head.sha }}
run: |
set -euo pipefail
fail=0
# --no-merges: merge commits are GitHub-generated and carry no sign-off.
for sha in $(git rev-list --no-merges "$BASE..$HEAD"); do
if [ -z "$(git log -1 --format='%(trailers:key=Signed-off-by)' "$sha")" ]; then
echo "Missing Signed-off-by: $(git log -1 --format='%h %s' "$sha")"
fail=1
fi
done
if [ "$fail" -ne 0 ]; then
echo
echo "Sign commits with 'git commit -s'; fix a branch with 'git rebase --signoff $BASE'."
echo "See CONTRIBUTING.md (Developer Certificate of Origin)."
exit 1
fi
lint:
name: lint (shellcheck, shfmt)
runs-on: ubuntu-24.04
steps:
- uses: actions/checkout@v6
- name: Install linters
run: |
set -euo pipefail
sudo apt-get update
# noble ships shfmt 3.8.0 (universe), matching the pinned local dev
# version; use it rather than fetching a release binary from github.com.
sudo apt-get install -y --no-install-recommends shellcheck shfmt
shfmt --version
# Lint the scripts we maintain; the legacy scripts are a separate cleanup.
- name: shellcheck
run: shellcheck man/makeman.sh tools/mkdeb.sh .githooks/pre-commit tests/*.test tests/check-network.sh
- name: shfmt
run: shfmt -d -i 4 man/makeman.sh tools/mkdeb.sh .githooks/pre-commit
# Check clang-format on CHANGED LINES ONLY. The engine predates clang-format
# (it was shaped by an old Visual Studio formatter) and does not round-trip,
# so we never reformat the whole tree -- only the lines a PR touches.
format:
name: format (clang-format-19, changed lines)
if: github.event_name == 'pull_request'
runs-on: ubuntu-24.04
steps:
- uses: actions/checkout@v6
with:
fetch-depth: 0
- name: Install clang-format 19 (pinned, from apt.llvm.org)
run: |
set -euo pipefail
# ubuntu-24.04's native clang-format is 18; pin 19 to match local dev.
wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key \
| sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc >/dev/null
echo "deb http://apt.llvm.org/noble/ llvm-toolchain-noble-19 main" \
| sudo tee /etc/apt/sources.list.d/llvm-19.list >/dev/null
sudo apt-get update
sudo apt-get install -y --no-install-recommends clang-format-19
# The clang-format-19 package ships the git-clang-format driver;
# expose it unsuffixed so "git clang-format" finds it.
sudo ln -sf /usr/bin/git-clang-format-19 /usr/local/bin/git-clang-format
clang-format-19 --version
- name: Check formatting of changed lines
run: |
set -euo pipefail
git fetch --no-tags origin \
"+refs/heads/${{ github.base_ref }}:refs/remotes/origin/${{ github.base_ref }}"
base="origin/${{ github.base_ref }}"
set +e
diff="$(git clang-format --binary clang-format-19 --style=file \
--diff --extensions c,h "$base")"
rc=$?
set -e
# Classify by output, not exit code: a non-empty diff means "not
# clean" (git-clang-format may exit 0 or 1 on a diff). A nonzero exit
# with clean output is a real checker error.
case "$diff" in
"" | "no modified files to format" | *"did not modify any files"*)
if [ "$rc" -ne 0 ]; then
echo "::error::git clang-format failed (exit $rc): checker error."
exit 1
fi
echo "Formatting OK: changed C lines are clang-format-clean." ;;
*)
echo "$diff"
echo "::error::Changed C lines are not clang-format-clean."
echo "Fix locally with: git clang-format --binary clang-format-19 $base"
exit 1 ;;
esac