httrack/tests/13_crawl_proxy_https.test

#!/bin/bash
#
# Issue #85: an https crawl must go through the configured proxy (CONNECT
# tunnel), not bypass it and hit the origin directly. Fully local: a self-signed
# TLS origin plus a logging CONNECT proxy, so no network access is needed.

set -euo pipefail

: "${top_srcdir:=..}"

# shellcheck source=tests/testlib.sh
. "$top_srcdir/tests/testlib.sh"

if test "${HTTPS_SUPPORT:-}" == "no"; then
    echo "no https support compiled, skipping"
    exit 77
fi
python=$(find_python) || python=
if test -z "$python" || ! command -v openssl >/dev/null 2>&1; then
    echo "python3/openssl missing, skipping"
    exit 77
fi

server=$(nativepath "$top_srcdir/tests/proxy-https-server.py")
tmpdir=$(mktemp -d)
pids=

cleanup() {
    for pid in $pids; do
        stop_server "$pid"
    done
    rm -rf "$tmpdir"
}
trap cleanup EXIT

# self-signed cert for the local TLS origin (httrack does not verify certs)
openssl req -x509 -newkey rsa:2048 -keyout "$tmpdir/key.pem" \
    -out "$tmpdir/cert.pem" -days 2 -nodes -subj "/CN=127.0.0.1" \
    >/dev/null 2>&1
cat "$tmpdir/key.pem" "$tmpdir/cert.pem" >"$tmpdir/both.pem"

# start_server <logdir> <mode>: launches a proxy+origin pair, sets $origin_port
# and $proxy_port from its announced ephemeral ports.
start_server() {
    local dir="$1" mode="$2" ports pem
    mkdir -p "$dir"
    ports="$dir/ports.txt"
    pem=$(nativepath "$tmpdir/both.pem")
    dir_native=$(nativepath "$dir")
    : >"$ports"
    "$python" "$server" "$pem" "$dir_native" "$mode" \
        >"$ports" 2>"$dir/server.err" &
    pids="$pids $!"
    for _ in $(seq 1 100); do
        grep -q "^ready" "$ports" 2>/dev/null && break
        sleep 0.1
    done
    grep -q "^ready" "$ports" 2>/dev/null || {
        echo "server ($mode) did not start" >&2
        cat "$dir/server.err" >&2
        exit 1
    }
    origin_port=$(awk '/^ORIGIN/{print $2}' "$ports")
    proxy_port=$(awk '/^PROXY/{print $2}' "$ports")
}

# Run httrack, but kill it after a deadline so a hang (e.g. a missing bound on
# the proxy response) surfaces as the kill code $HANG_RC instead of stalling the
# whole job. A portable stand-in for `timeout`, which macOS lacks.
HANG_RC=137 # 128 + SIGKILL
run_crawl() {
    local out="$1" proxy="$2" port="$3"
    rm -rf "$out"
    httrack "https://127.0.0.1:${port}/" --proxy "$proxy" \
        -O "$out" -r1 -s0 --timeout=10 >"$out.log" 2>&1 &
    local pid=$!
    (sleep 60 && kill -9 "$pid" 2>/dev/null) &
    local guard=$!
    local rc=0
    wait "$pid" 2>/dev/null || rc=$?
    kill "$guard" 2>/dev/null || true
    wait "$guard" 2>/dev/null || true
    return "$rc"
}

# --- working proxy ----------------------------------------------------------
ok="$tmpdir/ok"
start_server "$ok" ok

# 1. page retrieved AND the proxy saw a CONNECT to the origin
run_crawl "$ok/out" "127.0.0.1:${proxy_port}" "$origin_port"
grep -rq "ORIGIN-PAGE-85" "$ok/out" || {
    echo "FAIL: origin page not downloaded through proxy" >&2
    cat "$ok/out.log" >&2
    exit 1
}
grep -q "^CONNECT 127.0.0.1:${origin_port} " "$ok/proxy.log" || {
    echo "FAIL: proxy never received a CONNECT (https bypassed the proxy)" >&2
    cat "$ok/proxy.log" >&2
    exit 1
}
echo "OK: https tunneled through proxy via CONNECT"

# 2. authenticated proxy: creds ride the CONNECT, and NEVER reach the origin
: >"$ok/proxy.log"
: >"$ok/origin-headers.log"
run_crawl "$ok/out2" "user:secret@127.0.0.1:${proxy_port}" "$origin_port"
grep -rq "ORIGIN-PAGE-85" "$ok/out2" || {
    echo "FAIL: origin page not downloaded through authenticated proxy" >&2
    exit 1
}
got=$(awk '/^AUTH Basic /{print $3}' "$ok/proxy.log" | head -1)
# base64("user:secret"); compared as a literal to stay portable (no base64 -d,
# which differs between GNU and BSD)
test "$got" == "dXNlcjpzZWNyZXQ=" || {
    echo "FAIL: Proxy-Authorization not carried on CONNECT (got '$got')" >&2
    cat "$ok/proxy.log" >&2
    exit 1
}
if grep -qi "proxy-authorization" "$ok/origin-headers.log"; then
    echo "FAIL: proxy credentials leaked to the origin through the tunnel" >&2
    cat "$ok/origin-headers.log" >&2
    exit 1
fi
echo "OK: proxy credentials carried on CONNECT, not leaked to origin"

# --- hostile proxy ----------------------------------------------------------
# A proxy that answers 200 then streams headers forever must not hang the crawl:
# the client bounds the response. run_crawl kills a hung httrack after 60s, so a
# missing bound surfaces as $HANG_RC here.
flood="$tmpdir/flood"
start_server "$flood" flood
rc=0
run_crawl "$flood/out" "127.0.0.1:${proxy_port}" "$origin_port" || rc=$?
test "$rc" -ne "$HANG_RC" || {
    echo "FAIL: crawl hung on a flooding proxy (bounded read missing)" >&2
    exit 1
}
grep -rq "ORIGIN-PAGE-85" "$flood/out" 2>/dev/null && {
    echo "FAIL: flooding proxy unexpectedly served the page" >&2
    exit 1
}
echo "OK: bounded proxy response, no hang on a flooding proxy"