Compare commits

...
1 Commits
Author SHA1 Message Date
noahandClaude Opus 4.8 4154a50ed0 ci: measure coverage in the release job and auto-update the badge
Add a best-effort Coverage step to release.yml: it adds the llvm-tools-preview
rustup component, fetches cargo-llvm-cov as a prebuilt binary (the vendored
.cargo/config blocks `cargo install` from crates.io), runs ci/coverage.sh to
merge unit-test and headless self-test line coverage, and feeds the TOTAL
line-coverage percent to the badge refresh so badges/coverage.svg tracks the
real number on every release. The step is gated on a release and marked
continue-on-error, so coverage tooling can never block a release.

coverage.sh now runs its instrumented builds --offline by default (the
dependency tree is fully vendored), overridable via CARGO_LLVM_COV_FLAGS.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Djv3oncM4LxTS5tb61Ako6
2026-09-23 09:32:41 -06:00
2 changed files with 46 additions and 3 deletions
+40 -1
View File
@@ -97,6 +97,40 @@ jobs:
echo "cargo-audit not installed on this runner; skipping"
fi
# Merge coverage from the unit tests AND the headless self-test (ci/coverage.sh)
# so the committed badge reflects real coverage, not the tests-only number a
# plain `cargo llvm-cov` reports. Best-effort: cargo-llvm-cov is fetched as a
# prebuilt binary because the vendored .cargo/config blocks `cargo install`,
# and the instrumented rebuild is slow, so a failure here never blocks a
# release. Feeds steps.cov.outputs.pct to the badge refresh below. xvfb is
# already installed by the self-test step above.
- name: Coverage (best-effort)
id: cov
if: steps.ver.outputs.release == 'true'
continue-on-error: true
run: |
rustup component add llvm-tools-preview >/dev/null 2>&1 || true
if ! command -v cargo-llvm-cov >/dev/null 2>&1; then
V=0.8.7
URL="https://github.com/taiki-e/cargo-llvm-cov/releases/download/v${V}/cargo-llvm-cov-x86_64-unknown-linux-gnu.tar.gz"
DEST="${CARGO_HOME:-$HOME/.cargo}/bin"
mkdir -p "$DEST"
curl -fsSL "$URL" | tar -xz -C "$DEST" && export PATH="$DEST:$PATH" \
|| echo "cargo-llvm-cov download failed"
fi
if command -v cargo-llvm-cov >/dev/null 2>&1; then
PCT=$(bash ci/coverage.sh --summary-only 2>/dev/null \
| awk '/^TOTAL/{gsub(/%/,"",$10); print $10}')
if [ -n "$PCT" ]; then
echo "pct=${PCT}" >> "$GITHUB_OUTPUT"
echo "line coverage (tests + self-test): ${PCT}%"
else
echo "coverage report produced no TOTAL line; badge left as committed"
fi
else
echo "cargo-llvm-cov unavailable; coverage badge left as committed"
fi
# Refresh the committed README badges from live metrics and push any change
# back to main. Best-effort: a locked-down runner that can't push must not
# fail the release. A badges-only push does not bump the version (see
@@ -109,7 +143,12 @@ jobs:
run: |
PASS=$(cargo test -p alkane-gui --offline 2>&1 \
| grep -hoE '[0-9]+ passed' | grep -oE '[0-9]+' | awk '{s+=$1} END{print s+0}')
python3 ci/gen-badges.py --tests "${PASS:-0}" --bench "row-render + scan-parse"
# Update the coverage badge only when the coverage step produced a number;
# otherwise leave the committed coverage.svg untouched.
COV="${{ steps.cov.outputs.pct }}"
COV_ARG=""
[ -n "$COV" ] && COV_ARG="--coverage $COV"
python3 ci/gen-badges.py --tests "${PASS:-0}" $COV_ARG --bench "row-render + scan-parse"
if [ -n "$(git status --porcelain badges)" ]; then
git config user.name "alkane-ci"
git config user.email "ci@bluefla.re"
+6 -2
View File
@@ -12,18 +12,22 @@ set -euo pipefail
export WX_RS_WXCONFIG="${WX_RS_WXCONFIG:-$HOME/.cache/wx-3.3.3-linux/bin/wx-config}"
# The dependency tree is fully vendored, so the instrumented builds run offline by
# default; override with CARGO_LLVM_COV_FLAGS if you need something else.
read -ra CARGO_FLAGS <<<"${CARGO_LLVM_COV_FLAGS:---offline}"
# A display for the self-test's real widgets.
need_xvfb() { command -v xvfb-run >/dev/null 2>&1; }
cargo llvm-cov clean --workspace
# 1. Unit tests.
cargo llvm-cov --no-report -p alkane-gui --bin alkane-toolkit
cargo llvm-cov --no-report "${CARGO_FLAGS[@]}" -p alkane-gui --bin alkane-toolkit
# 2. Headless UI self-test, under the same instrumentation, its coverage appended.
if need_xvfb; then
xvfb-run -a env GDK_BACKEND=x11 ALKANE_SELFTEST=1 \
cargo llvm-cov run --no-report --bin alkane-toolkit
cargo llvm-cov run --no-report "${CARGO_FLAGS[@]}" --bin alkane-toolkit
else
echo "coverage.sh: xvfb-run not found; skipping the self-test run (unit tests only)" >&2
fi