free-claude-code/scripts/ci.sh
Ali Khokhar c9b75088b0
Some checks are pending
Dependency Cache / Warm dependency cache (push) Waiting to run
Parallelize browser CI and fix chat scroll reset after Stop (#1687)
## Why

Browser checks force serial execution despite the project's automatic
pytest parallelism. Running them in parallel also exposed a chat bug: a
title update after Stop could briefly clear the conversation before
saved turns finished loading, resetting the reader's scroll position.

## How

Let both local CI scripts and GitHub Actions inherit `-n auto`. Keep the
visible answer until saved turns replace it, then refresh the operation
controls. Update the existing scroll test to hold the reload while the
title changes, wait for the reply to start before stopping, and scroll
the current panel. Update the existing CI command assertions and bump
the package to 5.22.8.

The scroll test fails against the original UI and passes 96 repetitions
with 24 workers after the fix.

The complete local CI run passed: formatting, lint, typing, 4,828
regular tests (146 skipped), and all 65 browser tests.
2026-09-06 00:40:54 -07:00

222 lines
5 KiB
Bash
Executable file

#!/bin/sh
set -eu
CHECK_ORDER="suppressions ruff-format ruff-check ty pytest playwright"
dry_run=0
only_checks=""
skip_checks=""
show_usage() {
cat <<'USAGE'
Usage: ci.sh [options]
Runs the local sequence for the same check IDs enforced by GitHub CI.
Requires uv on PATH when running ruff, ty, pytest, or Playwright checks.
Local ruff checks repair formatting and autofixable lint before later checks.
Checks (in order):
suppressions Ban type ignores and legacy future annotations
ruff-format uv run ruff format
ruff-check uv run ruff check --fix
ty uv run ty check
pytest uv run pytest -v --tb=short
playwright Install Chromium and run deterministic Admin browser tests
Options:
--only ID Run only the given check (repeatable)
--skip ID Skip the given check (repeatable)
--dry-run Print commands without running them.
--help Show this help text.
USAGE
}
fail() {
printf 'error: %s\n' "$*" >&2
exit 1
}
step() {
printf '\n==> %s\n' "$1"
}
quote_arg() {
case "$1" in
*[!A-Za-z0-9_./:@%+=,-]*|"")
escaped=$(printf '%s' "$1" | sed 's/\\/\\\\/g; s/"/\\"/g')
printf '"%s"' "$escaped"
;;
*)
printf '%s' "$1"
;;
esac
}
print_command() {
printf '+'
for arg in "$@"; do
printf ' '
quote_arg "$arg"
done
printf '\n'
}
run() {
print_command "$@"
if [ "$dry_run" -eq 0 ]; then
"$@"
fi
}
valid_check_id() {
case "$1" in
suppressions | ruff-format | ruff-check | ty | pytest | playwright) return 0 ;;
*) return 1 ;;
esac
}
validate_check_id() {
if ! valid_check_id "$1"; then
fail "unknown check id: $1 (expected one of: $CHECK_ORDER)"
fi
}
contains_check_id() {
list=$1
id=$2
case " $list " in
*" $id "*) return 0 ;;
*) return 1 ;;
esac
}
should_run_check() {
check_id=$1
if [ -n "$only_checks" ] && ! contains_check_id "$only_checks" "$check_id"; then
return 1
fi
if contains_check_id "$skip_checks" "$check_id"; then
return 1
fi
return 0
}
assert_uv_available() {
if ! command -v uv >/dev/null 2>&1; then
fail "uv is required but was not found on PATH. Install uv first (see README or scripts/install.sh)."
fi
}
selected_checks_need_uv() {
if [ "$dry_run" -ne 0 ]; then
return 1
fi
for check_id in $CHECK_ORDER; do
if should_run_check "$check_id" && [ "$check_id" != "suppressions" ]; then
return 0
fi
done
return 1
}
run_suppressions() {
step "Ban suppressions and legacy annotations"
pattern='# type: ignore|# ty: ignore|from __future__ import annotations'
print_command grep -rE "$pattern" --include='*.py' . \
--exclude-dir=.venv --exclude-dir=.git
if [ "$dry_run" -eq 0 ]; then
if grep -rE "$pattern" --include='*.py' . \
--exclude-dir=.venv --exclude-dir=.git; then
fail "type: ignore / ty: ignore comments and legacy future annotations are not allowed. Fix the underlying type/import issue instead."
fi
fi
}
run_ruff_format() {
step "ruff format"
run uv run ruff format
}
run_ruff_check() {
step "ruff check --fix"
run uv run ruff check --fix
}
run_ty() {
step "ty check"
run uv run ty check
}
run_pytest() {
step "pytest"
run uv run pytest -v --tb=short
}
run_playwright() {
step "playwright"
run uv run playwright install chromium
run uv run pytest e2e -v --tb=short \
--tracing=retain-on-failure --screenshot=only-on-failure \
--full-page-screenshot --output=test-results
}
run_check() {
case "$1" in
suppressions) run_suppressions ;;
ruff-format) run_ruff_format ;;
ruff-check) run_ruff_check ;;
ty) run_ty ;;
pytest) run_pytest ;;
playwright) run_playwright ;;
*) fail "unknown check id: $1" ;;
esac
}
parse_args() {
while [ "$#" -gt 0 ]; do
case "$1" in
--only)
shift
[ "$#" -gt 0 ] || fail "--only requires a check id."
validate_check_id "$1"
only_checks="${only_checks} $1"
;;
--skip)
shift
[ "$#" -gt 0 ] || fail "--skip requires a check id."
validate_check_id "$1"
skip_checks="${skip_checks} $1"
;;
--dry-run)
dry_run=1
;;
--help|-h)
show_usage
exit 0
;;
*)
show_usage >&2
fail "unknown option: $1"
;;
esac
shift
done
}
parse_args "$@"
if selected_checks_need_uv; then
assert_uv_available
fi
for check_id in $CHECK_ORDER; do
if should_run_check "$check_id"; then
run_check "$check_id"
fi
done
printf '\nAll selected CI checks passed.\n'