| #!/bin/bash |
| # Copyright 2018 The Bazel Authors. All rights reserved. |
| # |
| # Licensed under the Apache License, Version 2.0 (the "License"); |
| # you may not use this file except in compliance with the License. |
| # You may obtain a copy of the License at |
| # |
| # http://www.apache.org/licenses/LICENSE-2.0 |
| # |
| # Unless required by applicable law or agreed to in writing, software |
| # distributed under the License is distributed on an "AS IS" BASIS, |
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| # See the License for the specific language governing permissions and |
| # limitations under the License. |
| # |
| # Tests of the bazel client. |
| |
| # Disable the package loader sanity check since many of these tests use fifos |
| # for controlling access to BUILD files. (This only has an effect at Google.) |
| export DONT_SANITY_CHECK_WITH_PACKAGE_LOADER=1 |
| |
| CURRENT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" |
| source "${CURRENT_DIR}/../integration_test_setup.sh" \ |
| || { echo "integration_test_setup.sh not found!" >&2; exit 1; } |
| |
| function strip_lines_from_bazel_cc() { |
| # sed can't redirect back to its input file (it'll only generate an empty |
| # file). In newer versions of gnu sed there is a -i option to edit in place. |
| |
| # Ignore common warnings caused by the environment on our CI workers. |
| clean_log=$(\ |
| sed \ |
| -e '/^WARNING: ignoring JAVA_TOOL_OPTIONS in environment.$/d' \ |
| -e '/^WARNING: The following rc files are no longer being read, please transfer their contents or import their path into one of the standard rc files:$/d' \ |
| -e '/^\/etc\/bazel.bazelrc$/d' \ |
| $TEST_log) |
| |
| echo "$clean_log" > $TEST_log |
| } |
| |
| #### TESTS ############################################################# |
| |
| function test_client_debug() { |
| # Test that --client_debug sends log statements to stderr. |
| bazel --client_debug version >&$TEST_log || fail "'bazel version' failed" |
| expect_log "Debug logging requested" |
| bazel --client_debug --batch version >&$TEST_log || fail "'bazel version' failed" |
| expect_log "Debug logging requested" |
| |
| # Test that --client_debug can be disabled |
| bazel --noclient_debug version >&$TEST_log || fail "'bazel version' failed" |
| expect_not_log "Debug logging requested" |
| bazel --noclient_debug --batch version >&$TEST_log || fail "'bazel version' failed" |
| expect_not_log "Debug logging requested" |
| |
| # Test that --client_debug is off by default. |
| bazel --ignore_all_rc_files version >&$TEST_log || fail "'bazel version' failed" |
| expect_not_log "Debug logging requested" |
| bazel --ignore_all_rc_files --batch version >&$TEST_log || fail "'bazel version' failed" |
| expect_not_log "Debug logging requested" |
| } |
| |
| function test_client_debug_change_does_not_restart_server() { |
| local server_pid1=$(bazel --client_debug info server_pid 2>$TEST_log) |
| local server_pid2=$(bazel --noclient_debug info server_pid 2>$TEST_log) |
| assert_equals "$server_pid1" "$server_pid2" |
| expect_not_log "WARNING.* Running B\\(azel\\|laze\\) server needs to be killed" |
| } |
| |
| function test_server_restart_due_to_startup_options() { |
| local server_pid1=$(bazel --write_command_log info server_pid 2>$TEST_log) |
| local server_pid2=$(bazel --nowrite_command_log info server_pid 2>$TEST_log) |
| assert_not_equals "$server_pid1" "$server_pid2" # pid changed. |
| expect_log "WARNING.* Running B\\(azel\\|laze\\) server needs to be killed" |
| } |
| |
| function test_multiple_requests_same_server() { |
| local server_pid1=$(bazel info server_pid 2>$TEST_log) |
| local server_pid2=$(bazel info server_pid 2>$TEST_log) |
| assert_equals "$server_pid1" "$server_pid2" |
| expect_not_log "WARNING.* Running B\\(azel\\|laze\\) server needs to be killed" |
| } |
| |
| function test_no_server_restart_if_options_order_changes() { |
| local server_pid1=$(bazel \ |
| --host_jvm_args=-Dfoo \ |
| --host_jvm_args=-Dfoo \ |
| --host_jvm_args=-Dbar \ |
| --client_debug info server_pid 2>$TEST_log) |
| local server_pid2=$(bazel \ |
| --host_jvm_args=-Dfoo \ |
| --host_jvm_args=-Dbar \ |
| --host_jvm_args=-Dfoo \ |
| --client_debug info server_pid 2>$TEST_log) |
| assert_equals "$server_pid1" "$server_pid2" |
| expect_not_log "WARNING.* Running B\\(azel\\|laze\\) server needs to be killed" |
| } |
| |
| function test_server_restart_if_number_of_option_instances_increases() { |
| local server_pid1=$(bazel \ |
| --host_jvm_args=-Dfoo \ |
| --client_debug info server_pid 2>$TEST_log) |
| local server_pid2=$(bazel \ |
| --host_jvm_args -Dfoo \ |
| --host_jvm_args -Dfoo \ |
| --client_debug info server_pid 2>$TEST_log) |
| assert_not_equals "$server_pid1" "$server_pid2" |
| expect_log "\\[WARNING .*\\] Running B\\(azel\\|laze\\) server needs to be killed" |
| expect_log "\\[INFO .*\\] Args from the current request that were not included when creating the server:" |
| expect_log "\\[INFO .*\\] --host_jvm_args=-Dfoo" |
| } |
| |
| function test_server_restart_if_number_of_option_instances_decreases() { |
| local server_pid1=$(bazel \ |
| --host_jvm_args=-Dfoo \ |
| --host_jvm_args -Dfoo \ |
| --client_debug info server_pid 2>$TEST_log) |
| local server_pid2=$(bazel \ |
| --host_jvm_args -Dfoo \ |
| --client_debug info server_pid 2>$TEST_log) |
| assert_not_equals "$server_pid1" "$server_pid2" |
| expect_log "\\[WARNING .*\\] Running B\\(azel\\|laze\\) server needs to be killed" |
| expect_log "\\[INFO .*\\] Args from the running server that are not included in the current request:" |
| expect_log "\\[INFO .*\\] --host_jvm_args=-Dfoo" |
| } |
| |
| function test_server_not_restarted_when_only_TMPDIR_changes() { |
| mkdir tmp1 |
| mkdir tmp2 |
| tmp1path="$(pwd)/tmp1" |
| tmp2path="$(pwd)/tmp2" |
| PID1=$(export TMPDIR="$tmp1path" && bazel info server_pid 2>$TEST_log) \ |
| || fail "bazel info failed" |
| PID2=$(export TMPDIR="$tmp2path" && bazel info server_pid 2>$TEST_log) \ |
| || fail "bazel info failed" |
| expect_not_log "Running B\\(azel\\|laze\\) server needs to be killed." |
| |
| assert_equals "$PID1" "$PID2" |
| } |
| |
| function test_server_restarted_on_explicit_heap_dump_path_change() { |
| mkdir tmp1 |
| mkdir tmp2 |
| tmp1path="$(pwd)/tmp1" |
| tmp2path="$(pwd)/tmp2" |
| PID1=$(bazel --host_jvm_args=-XX:HeapDumpPath="$tmp1path" info server_pid \ |
| 2>$TEST_log) || fail "bazel info failed" |
| PID2=$(bazel --host_jvm_args=-XX:HeapDumpPath="$tmp2path" info server_pid \ |
| 2>$TEST_log) || fail "bazel info failed" |
| expect_log "Running B\\(azel\\|laze\\) server needs to be killed." |
| |
| assert_not_equals "$PID1" "$PID2" |
| } |
| |
| function test_shutdown() { |
| local server_pid1=$(bazel info server_pid 2>$TEST_log) |
| bazel shutdown >& $TEST_log || fail "Expected success" |
| local server_pid2=$(bazel info server_pid 2>$TEST_log) |
| assert_not_equals "$server_pid1" "$server_pid2" |
| expect_not_log "WARNING.* Running B\\(azel\\|laze\\) server needs to be killed" |
| expect_log "Starting local B\\(azel\\|laze\\) server (.*) and connecting to it" |
| } |
| |
| function test_shutdown_different_options() { |
| bazel --host_jvm_args=-Di.am.a=teapot info >& $TEST_log || fail "Expected success" |
| bazel shutdown >& $TEST_log || fail "Expected success" |
| expect_log "WARNING.* Running B\\(azel\\|laze\\) server needs to be killed" |
| expect_not_log "Starting local B\\(azel\\|laze\\) server (.*) and connecting to it" |
| } |
| |
| function test_server_restart_due_to_startup_options_with_client_debug_information() { |
| # Using --write_command_log for no particular reason, if that flag is removed, another startup |
| # option will do just fine. |
| local server_pid1=$(bazel --client_debug --write_command_log info server_pid 2>$TEST_log) |
| local server_pid2=$(bazel --client_debug --nowrite_command_log info server_pid 2>$TEST_log) |
| assert_not_equals "$server_pid1" "$server_pid2" # pid changed. |
| expect_log "\\[WARNING .*\\] Running B\\(azel\\|laze\\) server needs to be killed" |
| expect_log "\\[INFO .*\\] Args from the running server that are not included in the current request:" |
| expect_log "\\[INFO .*\\] --write_command_log" |
| expect_log "\\[INFO .*\\] Args from the current request that were not included when creating the server:" |
| expect_log "\\[INFO .*\\] --nowrite_command_log" |
| } |
| |
| function test_exit_code() { |
| bazel query not_a_query >/dev/null &>$TEST_log && |
| fail "bazel query: expected nonzero exit" |
| expect_log "'not_a_query'" |
| } |
| |
| function test_output_base() { |
| out=$(bazel --output_base=$TEST_TMPDIR/output info output_base 2>$TEST_log) |
| assert_equals $TEST_TMPDIR/output "$out" |
| } |
| |
| function test_output_base_is_file() { |
| bazel --output_base=/dev/null &>$TEST_log && fail "Expected non-zero exit" |
| expect_log "FATAL.* Output base directory '/dev/null' could not be created.*exists" |
| } |
| |
| function test_cannot_create_output_base() { |
| bazel --output_base=/foo &>$TEST_log && fail "Expected non-zero exit" |
| expect_log "FATAL.* Output base directory '/foo' could not be created" |
| } |
| |
| function test_nonwritable_output_base() { |
| bazel --output_base=/ &>$TEST_log && fail "Expected non-zero exit" |
| expect_log "FATAL.* Output base directory '/' must be readable and writable." |
| } |
| |
| function test_install_base_races_dont_leave_temp_files() { |
| declare -a client_pids |
| for i in {1..3}; do |
| bazel --install_base="$TEST_TMPDIR/race/install" \ |
| --output_base="$TEST_TMPDIR/out$i" info install_base & |
| client_pids+=($!) |
| done |
| for pid in "${client_pids[@]}"; do |
| wait $pid |
| done |
| # Expect the "race" directory to contain only "install" and "install.lock". |
| for filename in $(ls "$TEST_TMPDIR/race/"); do |
| assert_one_of install install.lock "$filename" |
| done |
| } |
| |
| # Regression test for b/1295038. |
| function test_install_base_corrupted_by_deleted_file() { |
| local -r install_base="$TEST_TMPDIR/corrupted_install_base" |
| |
| bazel --install_base="$install_base" shutdown || fail "Expected success" |
| |
| rm "$install_base/process-wrapper" |
| |
| bazel --install_base="$install_base" >& $TEST_log && fail "Expected failure" |
| expect_log "FATAL.* corrupt installation: file '.*process-wrapper' is missing or modified" |
| expect_log "Please remove '$install_base' and try again." # uh-oh. |
| |
| rm -rf "$install_base" |
| bazel --install_base="$install_base" >& $TEST_log || fail "Expected success" |
| expect_log "Usage: $PRODUCT_NAME" # phew |
| } |
| |
| function test_install_base_corrupted_by_touched_file() { |
| local -r install_base="$TEST_TMPDIR/corrupted_install_base" |
| |
| bazel --install_base="$install_base" shutdown || fail "Expected success" |
| |
| touch "$install_base/process-wrapper" |
| |
| bazel --install_base="$install_base" >&$TEST_log && fail "Expected failure" |
| expect_log "FATAL.* corrupt installation: file '.*process-wrapper' is missing or modified" |
| expect_log "Please remove '$install_base' and try again." # uh-oh. |
| |
| rm -rf "$install_base" |
| bazel --install_base="$install_base" >&$TEST_log || fail "Expected success" |
| expect_log "Usage: $PRODUCT_NAME" # phew |
| } |
| |
| # Regression test for b/380443969. |
| function test_readonly_install_base() { |
| local -r install_base_parent="$TEST_TMPDIR/install_base_parent" |
| local -r install_base="$install_base_parent/install_base" |
| |
| mkdir -p "$install_base_parent" || fail "mkdir failed" |
| |
| # First install ourselves. |
| bazel --install_base="$install_base" shutdown || fail "Expected success" |
| |
| # Make the install base deeply read-only, including its parent directory. |
| chmod -R a-w "$install_base_parent" || fail "chmod failed" |
| |
| # Check that we're still able to run. |
| bazel --install_base="$install_base" shutdown || fail "Expected success" |
| } |
| |
| function test_output_user_root() { |
| # Test absolute path |
| bazel --output_user_root=$TEST_TMPDIR/user info output_base >& $TEST_log \ |
| || fail "Expected success" |
| expect_log "$TEST_TMPDIR/user/[0-9a-f]\{32\}" |
| |
| # Test relative path |
| bazel --output_user_root=../user info output_base >& $TEST_log \ |
| || fail "Expected success" |
| expect_log "$(cd .. && pwd)/user/[0-9a-f]\{32\}" |
| } |
| |
| function test_multiple_commands_same_output_base() { |
| # This test verifies that competing Bazel commands for the same output base |
| # will run sequentially. It also verifies that the messages printed by Bazel |
| # to tell the user that other processes are waited for are correct. |
| # |
| # This is a complex test because it deals with non-determinism: once we have |
| # started Bazel commands in parallel, we can't tell how they will |
| # finish... nor how they'll even start. To address this, we force each |
| # invocation to get "stuck" within a genrule and unblock it in a controlled |
| # manner. This lets us capture the order in which the commands complete so |
| # that we can later make assertions on them. |
| |
| mkdir pkg |
| |
| local -r invocations=3 # Number of concurrent invocations. |
| |
| declare -a ready # Files created by the wait[i] genrules when entered. |
| declare -a lock # Files on which the wait[i] genrules wait for. |
| for i in $(seq ${invocations}); do |
| ready[$i]="${TEST_TMPDIR}/ready.$i" |
| lock[$i]="${TEST_TMPDIR}/fifo.$i"; mkfifo "${lock[$i]}" |
| |
| # Make sure the actions run locally, even if Bazel is configured to run them |
| # remotely (which is the case when this test is run at Google), because we |
| # must be able to synchronize with them through the fifos. |
| cat >>pkg/BUILD <<EOF |
| genrule(name='wait$i', local = True, outs=['out.$i'], |
| cmd='touch ${ready[$i]}; sleep 5; cat ${lock[$i]} >\$@') |
| EOF |
| done |
| |
| declare -a log # Paths to the outputs of the Bazel invocations. |
| declare -a pid # PIDs of the Bazel invocations. |
| for i in $(seq ${invocations}); do |
| log[$i]="${TEST_TMPDIR}/log.$i" |
| bazel build "//pkg:wait$i" >>"${log[$i]}" 2>&1 & |
| pid[$i]="${!}" |
| done |
| |
| # The various Bazel invocations are are now competing to run. Wait for them |
| # to start, in any order, and then allow them to proceed, recording the order |
| # in which they actually started running the command. |
| declare -a order |
| local position=1 |
| while [ ${position} -le ${invocations} ]; do |
| for i in $(seq ${invocations}); do |
| if [ -e "${ready[$i]}" ]; then |
| order[$i]=${position}; position=$((position + 1)) |
| echo unlock >"${lock[$i]}" |
| rm "${ready[$i]}" |
| fi |
| done |
| sleep 1 |
| done |
| wait # We unblocked all genrules so wait for actual terminations. |
| |
| # Dump outputs to the test log for debugging in case of test failure. |
| for i in $(seq ${invocations}); do |
| sed "s,^,bazel $i: ," "${log[$i]}" >>"${TEST_log}" |
| done |
| |
| # Reorder invocations in the order they ran their commands so we can make |
| # assertions more easily. |
| declare -a orderedlog orderedpid |
| for i in $(seq ${invocations}); do |
| echo "bazel ${i} finished in position ${order[$i]} with PID ${pid[$i]}" \ |
| >>"${TEST_log}" |
| orderedlog[${order[$i]}]="${log[$i]}" |
| orderedpid[${order[$i]}]="${pid[$i]}" |
| done |
| |
| # Helper function to check if the ith Bazel log contains the a regexp. |
| expect_ith_log() { |
| local i="${1}"; shift |
| local re="${1}"; shift |
| if ! grep -qE "${re}" "${orderedlog[$i]}"; then |
| fail "$*: cannot find '${re}' in ${orderedlog[$i]}" |
| fi |
| } |
| |
| # Helper function to check if the ith Bazel log does not contain a regexp. |
| not_expect_ith_log() { |
| local i="${1}"; shift |
| local re="${1}"; shift |
| if grep -qE "${re}" "${orderedlog[$i]}"; then |
| fail "$*: found '${re}' in ${orderedlog[$i]}" |
| fi |
| } |
| |
| # Expectations for the first invocation to run are easy: we know that it |
| # didn't have to wait for anything. |
| not_expect_ith_log 1 "Another command.*is running" \ |
| "first invocation waited but should not have" |
| |
| # Expectations for all other invocations are... tricky. We can't tell how |
| # ran, because of how locking works: sometimes we wait on the client and |
| # sometimes we wait on the server. Furthermore, the loop that used to check |
| # if the PIDs have changed are time-based, so whether we detect a PID change |
| # or not in the logs is also subject to timing. Better to not even try. |
| for i in $(seq 2 ${invocations}); do |
| expect_ith_log $i \ |
| "\\(pid=${orderedpid[1]}\\).*on the (client|server)..." \ |
| "invocation $i did not wait for first one with pid ${orderedpid[1]}" |
| |
| # Make sure the trailing messages added to the wait lines are never seen on |
| # their own. |
| not_expect_ith_log $i "^ *lock taken by another server" \ |
| "lock taken message did not follow waiting message" |
| done |
| } |
| |
| function test_multiple_commands_different_output_base() { |
| # This test verifies that competing Bazel commands for different output bases |
| # will run in parallel. |
| # |
| # It works by having two concurrent commands, each running a non-hermetic test |
| # that synchronizes with the other through a FIFO. If the tests time out, it |
| # likely means one command was stuck waiting for the other. |
| |
| # The output bases for the two commands. |
| local -r output_base_1="$TEST_TMPDIR/output_base_1" |
| local -r output_base_2="$TEST_TMPDIR/output_base_2" |
| |
| # The FIFO used by the test to communicate. |
| local -r fifo="$TEST_TMPDIR/fifo" |
| mkfifo "$fifo" || fail "couldn't create fifo" |
| |
| # The test target. |
| # Make sure it runs locally even if Bazel is configured to run actions |
| # remotely (which is the case when this test is run at Google), because it |
| # must be able to synchronize through the FIFO. |
| mkdir -p x |
| cat > x/BUILD <<'EOF' |
| sh_test(name = "x", srcs = ["x.sh"], local = True) |
| EOF |
| |
| # The test script, whose arguments are: |
| # - one of "read" or "write" |
| # - the path to the fifo |
| cat > x/x.sh <<'EOF' |
| #!/bin/bash |
| if [[ "$1" == "read" ]]; then |
| cat "$2" > /dev/null |
| elif [[ "$1" == "write" ]]; then |
| echo 1 > "$2" |
| else |
| echo "invalid argument: $1" |
| exit 1 |
| fi |
| EOF |
| chmod +x x/x.sh |
| |
| # Launch two commands concurrently. |
| |
| bazel --output_base="$output_base_1" test //x:x \ |
| --test_arg=read --test_arg="$fifo" --test_timeout=10 \ |
| &> "$TEST_log-1" & |
| local -r pid_1="$!" |
| |
| bazel --output_base="$output_base_2" test //x:x \ |
| --test_arg=write --test_arg="$fifo" --test_timeout=10 \ |
| &> "$TEST_log-2" & |
| local -r pid_2="$!" |
| |
| # Wait for both commands to complete. |
| |
| if ! wait "$pid_1"; then |
| cat "$TEST_log-1" >> "$TEST_log" |
| fail "first command failed" |
| fi |
| |
| if ! wait "$pid_2"; then |
| cat "$TEST_log-2" >> "$TEST_log" |
| fail "second command failed" |
| fi |
| } |
| |
| function test_noblock_for_lock_reuse_server() { |
| # Use a FIFO to spoonfeed the Bazel server. |
| mkdir -p a && mkfifo a/BUILD || fail "couldn't create fifo a" |
| mkdir -p b && mkfifo b/BUILD || fail "couldn't create fifo b" |
| bazel --client_debug build --nobuild //a:a &> "$TEST_log" & |
| local -r subshell_pid="$!" |
| |
| # Wait until Bazel reads a/BUILD. After that, it will block on b/BUILD. |
| echo "filegroup(name='a', srcs=['//b:b'])" > a/BUILD |
| |
| # Get the client pid from the log. This isn't necessarily the subshell pid |
| # because there might be wrapper scripts in between. |
| local -r client_pid="$(cat "$TEST_log" | scrape_client_pid)" |
| |
| # Run another command in the same workspace but different startup options, |
| # which requires a server restart. Since the server is currently running the |
| # first command, it cannot restart immediately. |
| local exit_code=0 |
| bazel --client_debug --noblock_for_lock info &> "$TEST_log-2" || exit_code=$? |
| |
| # Unstick the first server *before* checking expectations, otherwise the test |
| # suite will hang. |
| echo "filegroup(name='b', visibility=['//visibility:public'])" > b/BUILD |
| wait "$subshell_pid" || fail "Couldn't wait" |
| rm -rf a b |
| |
| assert_equals 9 "$exit_code" # LOCK_HELD_NOBLOCK_FOR_LOCK |
| |
| cat "$TEST_log-2" >> "$TEST_log" |
| expect_log \ |
| "Another command (pid=$client_pid) is running. Exiting immediately." |
| } |
| |
| function test_noblock_for_lock_new_server() { |
| # Use a FIFO to spoonfeed the Bazel server. |
| mkdir -p a && mkfifo a/BUILD || fail "couldn't create fifo a" |
| mkdir -p b && mkfifo b/BUILD || fail "couldn't create fifo b" |
| bazel --client_debug build --nobuild //a:a &> "$TEST_log" & |
| local -r subshell_pid="$!" |
| |
| # Wait until Bazel reads a/BUILD. After that, it will block on b/BUILD. |
| echo "filegroup(name='a', srcs=['//b:b'])" > a/BUILD |
| |
| # Run another command in the same workspace with the same startup options, so |
| # that the server can be reused. Since the server is currently running the |
| # first command, the second command cannot be immediately run. |
| local exit_code=0 |
| bazel --client_debug --noblock_for_lock --host_jvm_args=-Dchampagne.supernova=1 info \ |
| &> "$TEST_log-2" || exit_code=$? |
| |
| # Unstick the first server *before* checking expectations, otherwise the test |
| # suite will hang. |
| echo "filegroup(name='b', visibility=['//visibility:public'])" > b/BUILD |
| wait "$subshell_pid" || fail "Couldn't wait" |
| rm -rf a b |
| |
| assert_equals 9 "$exit_code" # LOCK_HELD_NOBLOCK_FOR_LOCK |
| |
| cat "$TEST_log-2" >> "$TEST_log" |
| # Note: In this case, the user does not get a "pid=#" description of the |
| # client blocking the call. See the todo in KillRunningServer for a |
| # potential fix. |
| expect_log \ |
| "Exiting because the lock is held and --noblock_for_lock was given" |
| } |
| |
| function test_noblock_for_lock_with_batch() { |
| # Use a FIFO to spoonfeed the Bazel server. |
| mkdir -p a && mkfifo a/BUILD || fail "couldn't create fifo a" |
| mkdir -p b && mkfifo b/BUILD || fail "couldn't create fifo b" |
| bazel --client_debug --batch build --nobuild //a:a &>"$TEST_log" & |
| local -r subshell_pid="$!" |
| |
| # Wait until Bazel reads a/BUILD. After that, it will block on b/BUILD. |
| echo "filegroup(name='a', srcs=['//b:b'])" > a/BUILD |
| |
| # Get the client pid from the log. This isn't necessarily the subshell pid |
| # because there might be wrapper scripts in between. |
| local -r client_pid="$(cat "$TEST_log" | scrape_client_pid)" |
| |
| local exit_code=0 |
| bazel --client_debug --batch --noblock_for_lock info &>"$TEST_log-2" || exit_code=$? |
| |
| # Unstick the first server *before* checking expectations, otherwise the test |
| # suite will hang. |
| echo "filegroup(name='b', visibility=['//visibility:public'])" > b/BUILD |
| wait "$subshell_pid" || fail "Couldn't wait" |
| rm -rf a b |
| |
| assert_equals 9 "$exit_code" # LOCK_HELD_NOBLOCK_FOR_LOCK |
| |
| cat "$TEST_log-2" >> "$TEST_log" |
| expect_log "Another command holds the output base lock" |
| expect_log "pid=$client_pid" |
| expect_log \ |
| "Exiting because the output base lock is held and --noblock_for_lock was given" |
| } |
| |
| function test_no_arguments() { |
| bazel >&$TEST_log || fail "Expected zero exit" |
| expect_log "Usage: b\\(laze\\|azel\\)" |
| } |
| |
| function test_empty_command() { |
| bazel '' >&$TEST_log && fail "Expected non-zero exit" |
| expect_log "Command cannot be the empty string." |
| } |
| |
| function test_local_startup_timeout() { |
| local output_base=$(bazel info output_base 2>"$TEST_log") || |
| fail "bazel info failed" |
| |
| # --host-jvm_debug will cause the server to block, forcing the client |
| # into the timeout condition. |
| bazel --host_jvm_args="-agentlib:jdwp=transport=dt_socket,server=y,suspend=y,address=localhost:41687" \ |
| --local_startup_timeout_secs=1 2>"$TEST_log" & |
| local timeout=20 |
| while true; do |
| local jobs_output=$(jobs) |
| [[ $jobs_output =~ Exit ]] && break |
| [[ $jobs_output =~ Done ]] && fail "bazel should have exited non-zero" |
| |
| timeout="$(( ${timeout} - 1 ))" |
| [[ "${timeout}" -gt 0 ]] || { |
| kill -9 %1 |
| wait %1 |
| fail "--local_startup_timeout_secs was not respected" |
| } |
| # Wait for the client to exit. |
| sleep 1 |
| done |
| |
| expect_log "Starting local.*server (.*) and connecting to it" |
| expect_log "FATAL: couldn't connect to server" |
| } |
| |
| function test_max_idle_secs() { |
| # TODO(https://github.com/bazelbuild/bazel/issues/6773): Remove when fixed. |
| bazel shutdown |
| |
| local options=( --max_idle_secs=1 ) |
| |
| local output_base |
| output_base="$(bazel "${options[@]}" info output_base 2>"$TEST_log")" \ |
| || fail "bazel info failed" |
| local timeout=60 # Lower than the default --max_idle_secs. |
| while [[ -f "${output_base}/server/server.pid.txt" ]]; do |
| timeout="$(( ${timeout} - 1 ))" |
| [[ "${timeout}" -gt 0 ]] || fail "--max_idle_secs was not respected" |
| |
| # Wait for the server to go away. |
| sleep 1 |
| done |
| |
| bazel "${options[@]}" info >"$TEST_log" 2>&1 || fail "bazel info failed" |
| expect_log "Starting local.*server (.*) and connecting to it" |
| # Ensure the restart was not triggered by different startup options. |
| expect_not_log "WARNING: Running B\\(azel\\|laze\\) server needs to be killed" |
| } |
| |
| function test_dashdash_before_command() { |
| bazel -- info &>$TEST_log && "Expected failure" |
| exitcode=$? |
| assert_equals 2 $exitcode |
| expect_log "\\[FATAL .*\\] Unknown startup option: '--'." |
| } |
| |
| function test_dashdash_after_command() { |
| bazel info -- &>$TEST_log || fail "info -- failed" |
| } |
| |
| function test_nobatch() { |
| local pid1=$(bazel --batch --nobatch info server_pid 2> $TEST_log) |
| local pid2=$(bazel --batch --nobatch info server_pid 2> $TEST_log) |
| assert_equals "$pid1" "$pid2" |
| expect_not_log "WARNING.* Running B\\(azel\\|laze\\) server needs to be killed" |
| } |
| |
| # Regression test for #1875189, "bazel client should pass through '--help' like |
| # a command". |
| function test_bazel_dash_dash_help_is_passed_through() { |
| bazel --help >&$TEST_log |
| expect_log "Usage: b\\(azel\\|laze\\) <command> <options> ..." |
| expect_not_log "Unknown startup option: '--help'." |
| } |
| |
| function test_bazel_dash_help() { |
| bazel -help >&$TEST_log |
| expect_log "Usage: b\\(azel\\|laze\\) <command> <options> ..." |
| } |
| |
| function test_bazel_dash_h() { |
| bazel -h >&$TEST_log |
| expect_log "Usage: b\\(azel\\|laze\\) <command> <options> ..." |
| } |
| |
| function test_bazel_dash_s_is_not_parsed() { |
| bazel -s --help >&$TEST_log && fail "Expected failure" |
| expect_log "Unknown startup option: '-s'." |
| } |
| |
| function test_batch() { |
| local pid1=$(bazel info server_pid 2> $TEST_log) |
| local pid2=$(bazel --batch info server_pid 2> $TEST_log) |
| assert_not_equals "$pid1" "$pid2" |
| expect_log "WARNING.* Running B\\(azel\\|laze\\) server needs to be killed" |
| } |
| |
| function test_cmdline_not_written_in_batch_mode() { |
| OUTPUT_BASE=$(bazel --batch info output_base 2> $TEST_log) |
| rm -f $OUTPUT_BASE/server/cmdline |
| OUTPUT_BASE2=$(bazel --batch info output_base 2> $TEST_log) |
| assert_equals "$OUTPUT_BASE" "$OUTPUT_BASE2" |
| [[ ! -e $OUTPUT_BASE/server/cmdline ]] || fail "Command line file written." |
| } |
| |
| function test_bad_command_batch() { |
| bazel --batch notacommand &> $TEST_log && "Expected failure" |
| exitcode=$? |
| assert_equals 2 "$exitcode" |
| expect_log "Command 'notacommand' not found." |
| } |
| |
| function test_bad_command_nobatch() { |
| bazel --nobatch notacommand &> $TEST_log && "Expected failure" |
| exitcode=$? |
| assert_equals 2 "$exitcode" |
| expect_log "Command 'notacommand' not found." |
| } |
| |
| function get_pid_environment() { |
| local pid="$1" |
| case "$(uname -s)" in |
| Linux) |
| cat "/proc/${pid}/environ" | tr '\0' '\n' |
| ;; |
| Darwin) |
| if ! ps > /dev/null; then |
| echo "Cannot use ps command, probably due to sandboxing." >&2 |
| return 1 |
| fi |
| ps eww -o command "${pid}" | tr ' ' '\n' |
| ;; |
| *) |
| false |
| ;; |
| esac |
| } |
| |
| function test_proxy_settings() { |
| # We expect that proxy settings are propagated from the client to the server |
| # process, but are _not_ used for client-server communication. |
| |
| bazel shutdown # We are changing the server process's environment variables. |
| |
| local example_no_proxy='foo.example.com' |
| # A known-invalid http*_proxy value which, if not ignored, would be expected |
| # to cause the client-server gRPC channel to time out or otherwise fail. |
| local invalid_proxy='http://localhost:0' |
| local server_pid |
| server_pid="$(http_proxy="${invalid_proxy}" HTTP_PROXY="${invalid_proxy}" \ |
| https_proxy="${invalid_proxy}" HTTPS_PROXY="${invalid_proxy}" \ |
| no_proxy="${example_no_proxy}" NO_PROXY="${example_no_proxy}" \ |
| bazel info server_pid 2> $TEST_log)" \ |
| || fail "http*_proxy env variables not ignored by client-server channel." |
| |
| # Check that the server uses the *_proxy env variables set by the client. |
| if get_pid_environment "${server_pid}" > "${TEST_TMPDIR}/server_env"; then |
| local var |
| for var in http{,s}_proxy HTTP{,S}_PROXY; do |
| assert_contains "^${var}=${invalid_proxy}\$" "${TEST_TMPDIR}/server_env" |
| done |
| for var in no_proxy NO_PROXY; do |
| assert_contains "^${var}=${example_no_proxy}\$" \ |
| "${TEST_TMPDIR}/server_env" |
| done |
| else |
| echo "cannot test server process environment on this platform" |
| fi |
| } |
| |
| function test_macos_qos_class() { |
| for class in utility background; do |
| bazel --macos_qos_class="${class}" info >"${TEST_log}" 2>&1 \ |
| || fail "Unknown QoS class ${class}" |
| # On macOS it'd be nice to verify that the server is indeed running at the |
| # desired class... but this is very hard to do. Common utilities do not |
| # print the QoS level, and powermetrics (which requires root privileges) |
| # only reports it under load -- so an "info" command is insufficient and the |
| # real thing would be quite expensive. |
| done |
| |
| for class in user-interactive user-initiated default ; do |
| bazel --macos_qos_class="${class}" >"${TEST_log}" 2>&1 \ |
| && fail "Expected failure with invalid QoS class name" |
| expect_log "Invalid argument.*qos_class.*${class}" |
| done |
| } |
| |
| function test_ignores_jdk_option_environment_variables() { |
| bazel shutdown # Environment variables are only checked on server startup |
| _JAVA_OPTIONS=--wat1 JDK_JAVA_OPTIONS=--wat2 JAVA_TOOL_OPTIONS=--wat3 \ |
| bazel version >&$TEST_log || fail "_JAVA_OPTIONS not ignored" |
| |
| expect_log ".*ignoring _JAVA_OPTIONS" |
| expect_log ".*ignoring JDK_JAVA_OPTIONS" |
| expect_log ".*ignoring JAVA_TOOL_OPTIONS" |
| } |
| |
| # Demonstrates that the client program prints exactly what we expect to stderr |
| # and stdout. Notably by default (--client_debug=false) there should be no debug |
| # log statements from our own codebase (or even from libraries we use!) printed |
| # to stderr. |
| function test_client_is_quiet_by_default() { |
| local capitalized_product_name="$(echo "$PRODUCT_NAME" | awk '{print toupper(substr($0,1,1)) substr($0,2)}')" |
| # Ensure we don't have a server running. Also ensure we've already extracted |
| # the installation (that way we don't expect an informational message about |
| # that). |
| bazel shutdown &> /dev/null |
| |
| bazel info server_pid > stdout 2> stderr || fail "bazel info failed" |
| cp stderr $TEST_log || fail "cp failed" |
| |
| strip_lines_from_bazel_cc |
| |
| assert_equals 2 $(cat $TEST_log | wc -l) |
| expect_log "^\$TEST_TMPDIR defined, some defaults will be overridden" |
| expect_log "^Starting local $capitalized_product_name server (.*) and connecting to it...$" |
| cp stdout $TEST_log || fail "cp failed" |
| |
| strip_lines_from_bazel_cc |
| |
| assert_equals 1 $(cat $TEST_log | wc -l) |
| expect_log "^[0-9]\+$" |
| |
| rm stderr stdout || fail "rm failed" |
| bazel info server_pid > stdout 2> stderr || fail "bazel info failed" |
| cp stderr $TEST_log || fail "cp failed" |
| |
| strip_lines_from_bazel_cc |
| |
| assert_equals 1 $(cat $TEST_log | wc -l) |
| expect_log "^\$TEST_TMPDIR defined, some defaults will be overridden" |
| cp stdout $TEST_log || fail "cp failed" |
| |
| strip_lines_from_bazel_cc |
| |
| assert_equals 1 $(cat $TEST_log | wc -l) |
| expect_log "^[0-9]\+$" |
| } |
| |
| function test_sigquit() { |
| # Use a FIFO to spoonfeed the Bazel server. |
| mkdir -p a && mkfifo a/BUILD || fail "couldn't create fifo a" |
| mkdir -p b && mkfifo b/BUILD || fail "couldn't create fifo b" |
| bazel --client_debug build --nobuild //a:a &> "$TEST_log" & |
| local -r subshell_pid="$!" |
| |
| # Wait until Bazel reads a/BUILD. After that, it will block on b/BUILD. |
| echo "filegroup(name='a', srcs=['//b:b'])" > a/BUILD |
| |
| # Get the client pid from the log. This isn't necessarily the subshell pid |
| # because there might be wrapper scripts in between. |
| local -r client_pid="$(cat "$TEST_log" | scrape_client_pid)" |
| |
| # Send a SIGQUIT to the client. |
| kill -SIGQUIT "$client_pid" |
| |
| # Unstick the server *before* checking expectations, otherwise the test suite |
| # will hang. |
| echo "filegroup(name='b', visibility=['//visibility:public'])" > b/BUILD |
| wait "$subshell_pid" || fail "Couldn't wait" |
| rm -rf a b |
| |
| # Get the jvm.out location. |
| local -r jvm_out="$(bazel --client_debug info output_base)/server/jvm.out" |
| |
| # Look for a distinctive string indicating the presence of a thread dump. |
| assert_contains "Full thread dump" "$jvm_out" |
| } |
| |
| function scrape_client_pid() { |
| sed -nr 's/.*Running \(pid=([0-9]+)\)/\1/p' |
| } |
| |
| run_suite "Tests of the bazel client." |