| #!/bin/bash |
| # |
| # Copyright 2016 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. |
| # |
| # An end-to-end test that Bazel's UI produces reasonable output. |
| |
| # --- begin runfiles.bash initialization --- |
| set -euo pipefail |
| if [[ ! -d "${RUNFILES_DIR:-/dev/null}" && ! -f "${RUNFILES_MANIFEST_FILE:-/dev/null}" ]]; then |
| if [[ -f "$0.runfiles_manifest" ]]; then |
| export RUNFILES_MANIFEST_FILE="$0.runfiles_manifest" |
| elif [[ -f "$0.runfiles/MANIFEST" ]]; then |
| export RUNFILES_MANIFEST_FILE="$0.runfiles/MANIFEST" |
| elif [[ -f "$0.runfiles/bazel_tools/tools/bash/runfiles/runfiles.bash" ]]; then |
| export RUNFILES_DIR="$0.runfiles" |
| fi |
| fi |
| if [[ -f "${RUNFILES_DIR:-/dev/null}/bazel_tools/tools/bash/runfiles/runfiles.bash" ]]; then |
| source "${RUNFILES_DIR}/bazel_tools/tools/bash/runfiles/runfiles.bash" |
| elif [[ -f "${RUNFILES_MANIFEST_FILE:-/dev/null}" ]]; then |
| source "$(grep -m1 "^bazel_tools/tools/bash/runfiles/runfiles.bash " \ |
| "$RUNFILES_MANIFEST_FILE" | cut -d ' ' -f 2-)" |
| else |
| echo >&2 "ERROR: cannot find @bazel_tools//tools/bash/runfiles:runfiles.bash" |
| exit 1 |
| fi |
| # --- end runfiles.bash initialization --- |
| |
| source "$(rlocation "io_bazel/src/test/shell/integration_test_setup.sh")" \ |
| || { echo "integration_test_setup.sh not found!" >&2; exit 1; } |
| |
| case "$(uname -s | tr [:upper:] [:lower:])" in |
| msys*|mingw*|cygwin*) |
| declare -r is_windows=true |
| ;; |
| *) |
| declare -r is_windows=false |
| ;; |
| esac |
| |
| if "$is_windows"; then |
| export MSYS_NO_PATHCONV=1 |
| export MSYS2_ARG_CONV_EXCL="*" |
| fi |
| |
| #### SETUP ############################################################# |
| |
| add_to_bazelrc "build --genrule_strategy=local" |
| add_to_bazelrc "test --test_strategy=standalone" |
| |
| function set_up() { |
| if [[ -d pkg ]]; then |
| # All tests share these scratch packages. No need to recreate them if they |
| # already exist. |
| return |
| fi |
| |
| mkdir -p pkg |
| touch remote_file |
| cat > pkg/true.sh <<EOF |
| #!/bin/sh |
| exit 0 |
| EOF |
| chmod 755 pkg/true.sh |
| cat > pkg/slow.sh <<EOF |
| #!/bin/sh |
| sleep 3 |
| exit 0 |
| EOF |
| chmod 755 pkg/slow.sh |
| cat > pkg/false.sh <<EOF |
| #!/bin/sh |
| exit 1 |
| EOF |
| chmod 755 pkg/false.sh |
| cat > pkg/output.sh <<EOF |
| #!/bin/sh |
| `which echo` -n foo |
| sleep 1 |
| `which echo` -n bar |
| exit 0 |
| EOF |
| chmod 755 pkg/output.sh |
| cat > pkg/do_output.sh <<EOF |
| #!/bin/sh |
| echo Beginning \$1 |
| for _ in \`seq 1 10240\` |
| do echo '1234567890' |
| done |
| echo Ending \$1 |
| EOF |
| chmod 755 pkg/do_output.sh |
| cat > pkg/BUILD <<EOF |
| sh_test( |
| name = "true", |
| srcs = ["true.sh"], |
| ) |
| sh_test( |
| name = "slow", |
| srcs = ["slow.sh"], |
| ) |
| sh_test( |
| name = "false", |
| srcs = ["false.sh"], |
| ) |
| sh_test( |
| name = "output", |
| srcs = ["output.sh"], |
| ) |
| genrule( |
| name = "gentext", |
| outs = ["gentext.txt"], |
| cmd = "echo here be dragons > \"\$@\"" |
| ) |
| genrule( |
| name = "withOutputA", |
| outs = ["a"], |
| tools = [":do_output.sh"], |
| cmd = "\$(location :do_output.sh) A && touch \"\$@\"", |
| ) |
| genrule( |
| name = "withOutputB", |
| outs = ["b"], |
| tools = [":do_output.sh"], |
| cmd = "\$(location :do_output.sh) B && touch \$@", |
| ) |
| sh_library( |
| name = "outputlib", |
| data = [":withOutputA", ":withOutputB"], |
| ) |
| sh_test( |
| name = "truedependingonoutput", |
| srcs = ["true.sh"], |
| deps = [":outputlib"], |
| ) |
| EOF |
| mkdir -p error |
| cat > error/BUILD <<'EOF' |
| genrule( |
| name = "failwitherror", |
| outs = ["fail.txt"], |
| cmd = "echo Here is the error message; exit 1", |
| ) |
| EOF |
| mkdir -p pkg/errorAfterWarning |
| cat > pkg/errorAfterWarning/BUILD <<'EOF' |
| RANGE = range(500) |
| |
| [ genrule( |
| name = "true%s_c" % i, |
| outs = ["true%s.c" % i], |
| cmd = "echo Build Warning...; echo 'int main(int argc, char **argv) { return 0; }' > $@", |
| ) for i in RANGE] |
| |
| [ cc_binary( |
| name = "true_%s" % i, |
| srcs = ["true%s.c" % i], |
| ) for i in RANGE] |
| |
| genrule( |
| name = "failing", |
| outs = ["failing.txt"], |
| srcs = ["true_%s" % i for i in RANGE], |
| cmd = "echo This is the error message; false", |
| ) |
| EOF |
| chmod -w pkg/* # prevent accidental editing |
| # keep directories writable though, so that test clean up can work |
| chmod 755 error |
| chmod 755 pkg/errorAfterWarning |
| mkdir -p pkg/debugMessages |
| cat > pkg/debugMessages/rule.bzl <<'EOF' |
| def _impl(ctx): |
| print("static debug message") |
| ctx.actions.write(ctx.outputs.out, "Hello World") |
| |
| withdebug = rule( |
| implementation = _impl, |
| attrs = {}, |
| outputs = {"out" : "%{name}.txt"}, |
| ) |
| EOF |
| cat > pkg/debugMessages/BUILD <<'EOF' |
| load("//pkg/debugMessages:rule.bzl", "withdebug") |
| |
| [ withdebug(name = "target%d" % (i,)) for i in range(50) ] |
| EOF |
| mkdir -p bzl |
| touch bzl/BUILD |
| cat > bzl/bzl.bzl <<'EOF' |
| x = invalidsyntax |
| EOF |
| mkdir -p pkgloadingerror |
| cat > pkgloadingerror/BUILD <<'EOF' |
| load("//bzl:bzl.bzl", "x") |
| EOF |
| mkdir -p fancyOutput |
| cat > fancyOutput/BUILD <<'EOF' |
| genrule( |
| name = "withFancyOutput", |
| outs = ["out.txt"], |
| cmd = "echo $$'\\xF0\\x9F\\x8D\\x83'; echo Hello World > $@", |
| ) |
| EOF |
| } |
| |
| function create_pkg() { |
| local -r pkg=$1 |
| mkdir -p $pkg |
| cat > $pkg/true.sh <<EOF |
| #!/bin/sh |
| exit 0 |
| EOF |
| chmod 755 $pkg/true.sh |
| cat > $pkg/BUILD <<EOF |
| sh_test( |
| name = "true", |
| srcs = ["true.sh"], |
| ) |
| EOF |
| } |
| |
| #### TESTS ############################################################# |
| |
| function test_basic_progress() { |
| bazel test --curses=yes --color=yes pkg:true 2>$TEST_log \ |
| || fail "${PRODUCT_NAME} test failed" |
| # some progress indicator is shown |
| expect_log '\[[0-9,]* / [0-9,]*\]' |
| # curses are used to delete at least one line |
| expect_log $'\x1b\[1A\x1b\[K' |
| # As precisely one target is specified, it should be reported during |
| # analysis phase. |
| expect_log 'Analy.*pkg:true' |
| } |
| |
| function test_line_wrapping() { |
| local -r pkg=$FUNCNAME |
| create_pkg $pkg |
| bazel test --curses=yes --color=yes --terminal_columns=5 $pkg:true 2>$TEST_log || fail "bazel test failed" |
| # curses are used to delete at least one line |
| expect_log $'\x1b\[1A\x1b\[K' |
| # something is written in green |
| expect_log $'\x1b\[32m' |
| # lines are wrapped, hence at least one line should end with backslash |
| expect_log '\\'$'\r''$\|\\$' |
| } |
| |
| function test_noshow_progress() { |
| bazel test --noshow_progress --curses=yes --color=yes \ |
| pkg:true 2>$TEST_log || fail "${PRODUCT_NAME} test failed" |
| # Info messages should still go through |
| expect_log 'Elapsed time' |
| # no progress indicator is shown |
| expect_not_log '\[[0-9,]* / [0-9,]*\]' |
| } |
| |
| function test_basic_progress_no_curses() { |
| bazel test --curses=no --color=yes pkg:true 2>$TEST_log \ |
| || fail "${PRODUCT_NAME} test failed" |
| # some progress indicator is shown |
| expect_log '\[[0-9,]* / [0-9,]*\]' |
| # cursor is not moved up |
| expect_not_log $'\x1b\[1A' |
| # no line is deleted |
| expect_not_log $'\x1b\[K' |
| # but some green color is used |
| expect_log $'\x1b\[32m' |
| } |
| |
| function test_no_curses_no_linebreak() { |
| bazel test --curses=no --color=yes --terminal_columns=9 \ |
| pkg:true 2>$TEST_log || fail "${PRODUCT_NAME} test failed" |
| # expect a long-ish status line |
| expect_log '\[[0-9,]* / [0-9,]*\]......' |
| } |
| |
| function test_pass() { |
| bazel test --curses=yes --color=yes pkg:true >$TEST_log \ |
| || fail "${PRODUCT_NAME} test failed" |
| # PASS is written in green on the same line as the test target |
| expect_log 'pkg:true.*'$'\x1b\[32m''.*PASS' |
| } |
| |
| function test_fail() { |
| bazel test --curses=yes --color=yes pkg:false >$TEST_log \ |
| && fail "expected failure" |
| # FAIL is written in red bold on the same line as the test target |
| expect_log 'pkg:false.*'$'\x1b\[31m\x1b\[1m''.*FAIL' |
| } |
| |
| function test_timestamp() { |
| bazel test --show_timestamps pkg:true 2>$TEST_log \ |
| || fail "${PRODUCT_NAME} test failed" |
| # expect something that looks like HH:mm:ss |
| expect_log '[0-2][0-9]:[0-5][0-9]:[0-6][0-9]' |
| } |
| |
| function test_info_spacing() { |
| # Verify that the output of "bazel info" is suitable for backtick escapes, |
| # in particular free carriage-return characters. |
| BAZEL_INFO_OUTPUT=XXX`bazel info workspace`XXX |
| echo "$BAZEL_INFO_OUTPUT" | grep -q 'XXX[^'$'\r'']*XXX' \ |
| || fail "${PRODUCT_NAME} info output spaced as $BAZEL_INFO_OUTPUT" |
| } |
| |
| function test_query_spacing() { |
| # Verify that the output of "bazel query" is suitable for consumption by |
| # other tools, i.e., contains only result lines, separated only by newlines. |
| BAZEL_QUERY_OUTPUT=`bazel query 'deps(//pkg:true)'` |
| echo "$BAZEL_QUERY_OUTPUT" | grep -q -v '^[@/]' \ |
| && fail "bazel query output is >$BAZEL_QUERY_OUTPUT<" || true |
| if ! $is_windows; then |
| echo "$BAZEL_QUERY_OUTPUT" | grep -q $'\r' \ |
| && fail "bazel query output is >$BAZEL_QUERY_OUTPUT<" || true |
| fi |
| } |
| |
| function test_query_progress() { |
| # Verify that some form of progress is reported during bazel query |
| bazel query 'deps(//pkg:true)' 2> "${TEST_log}" |
| expect_log 'Loading:.*packages loaded' |
| } |
| |
| function test_clean_nobuild { |
| bazel clean 2>$TEST_log \ |
| || fail "bazel shutdown failed" |
| expect_not_log "actions running" |
| expect_not_log "Building" |
| } |
| |
| function test_clean_color_nobuild { |
| bazel clean --color=yes 2>$TEST_log \ |
| || fail "bazel shutdown failed" |
| expect_not_log "actions running" |
| expect_not_log "Building" |
| } |
| |
| function test_help_nobuild { |
| bazel help 2>$TEST_log \ |
| || fail "bazel help failed" |
| expect_not_log "actions running" |
| expect_not_log "Building" |
| } |
| |
| function test_help_color_nobuild { |
| bazel help --color=yes 2>$TEST_log \ |
| || fail "bazel help failed" |
| expect_not_log "actions running" |
| expect_not_log "Building" |
| } |
| |
| function test_version_nobuild { |
| bazel version --curses=yes 2>$TEST_log \ |
| || fail "bazel version failed" |
| expect_not_log "action" |
| expect_not_log "Building" |
| } |
| |
| function test_version_nobuild_announce_rc { |
| bazel version --curses=yes --announce_rc 2>$TEST_log \ |
| || fail "bazel version failed" |
| expect_not_log "action" |
| expect_not_log "Building" |
| } |
| |
| function test_subcommand { |
| bazel clean || fail "${PRODUCT_NAME} clean failed" |
| bazel build -s pkg:gentext 2>$TEST_log \ |
| || fail "bazel build failed" |
| expect_log "here be dragons" |
| } |
| |
| function test_subcommand_notdefault { |
| bazel clean || fail "${PRODUCT_NAME} clean failed" |
| bazel build pkg:gentext 2>$TEST_log \ |
| || fail "bazel build failed" |
| expect_not_log "dragons" |
| } |
| |
| function test_loading_progress { |
| bazel clean || fail "${PRODUCT_NAME} clean failed" |
| bazel test pkg:true 2>$TEST_log \ |
| || fail "${PRODUCT_NAME} test failed" |
| # some progress indicator is shown during loading |
| expect_log 'Loading.*[0-9,]* packages' |
| } |
| |
| function test_failure_scrollback_buffer_curses { |
| bazel clean || fail "${PRODUCT_NAME} clean failed" |
| bazel test --curses=yes --color=yes \ |
| --nocache_test_results pkg:false pkg:slow 2>$TEST_log \ |
| && fail "expected failure" |
| # Some line starts with FAIL in red bold. |
| expect_log '^'$'\(.*\x1b\[K\)*\x1b\[31m\x1b\[1mFAIL:' |
| } |
| |
| function test_terminal_title { |
| bazel test --curses=yes \ |
| --progress_in_terminal_title pkg:true \ |
| 2>$TEST_log || fail "${PRODUCT_NAME} test failed" |
| # The terminal title is changed |
| expect_log $'\x1b\]0;.*\x07' |
| } |
| |
| function test_failure_scrollback_buffer { |
| bazel clean || fail "${PRODUCT_NAME} clean failed" |
| bazel test --curses=no --color=yes \ |
| --nocache_test_results pkg:false pkg:slow 2>$TEST_log \ |
| && fail "expected failure" |
| # Some line starts with FAIL in red bold. |
| expect_log '^'$'\x1b\[31m\x1b\[1mFAIL:' |
| } |
| |
| function test_streamed { |
| bazel test --curses=yes --color=yes \ |
| --nocache_test_results --test_output=streamed pkg:output >$TEST_log \ |
| || fail "expected success" |
| expect_log 'foobar' |
| } |
| |
| function test_stdout_bundled { |
| # Verify that the error message is part of the error event |
| bazel build --experimental_ui_debug_all_events \ |
| error:failwitherror > "${TEST_log}" 2>&1 \ |
| && fail "expected failure" || : |
| grep -A1 '^ERROR' "${TEST_log}" \ |
| | grep -q "with STDOUT: Here is the error message" \ |
| || fail "Error message not bundled" |
| } |
| |
| function test_experimental_ui_attempt_to_print_relative_paths_failing_action() { |
| # On the BazelCI Windows environment, `pwd` returns a string that uses a |
| # lowercase drive letter and unix-style path separators (e.g. '/c/') with |
| # a lowercase drive letter. But internally in Bazel, Path#toString |
| # unconditionally uses an uppercase drive letter (see |
| # WindowsOsPathPolicy#normalize). I want these tests to check for exact |
| # string contents (that's the entire goal of the flag being tested), but I |
| # don't want them to be brittle across different Windows enviromments, so |
| # I've disabled them for now. |
| # TODO(nharmata): Fix this. |
| [[ "$is_windows" == "true" ]] && return 0 |
| |
| bazel clean || fail "${PRODUCT_NAME} clean failed" |
| |
| bazel build --attempt_to_print_relative_paths=false \ |
| error:failwitherror > "${TEST_log}" 2>&1 && fail "expected failure" |
| expect_log "^ERROR: $(pwd)/error/BUILD:1:8: Executing genrule //error:failwitherror failed: " |
| |
| bazel build --attempt_to_print_relative_paths=true \ |
| error:failwitherror > "${TEST_log}" 2>&1 && fail "expected failure" |
| expect_log "^ERROR: error/BUILD:1:8: Executing genrule //error:failwitherror failed: " |
| expect_not_log "$(pwd)/error/BUILD" |
| } |
| |
| function test_experimental_ui_attempt_to_print_relative_paths_pkg_error() { |
| # See the note in the test case above for why this is disabled on Windows. |
| # TODO(nharmata): Fix this. |
| [[ "$is_windows" == "true" ]] && return 0 |
| |
| bazel clean || fail "${PRODUCT_NAME} clean failed" |
| |
| bazel build --attempt_to_print_relative_paths=false \ |
| pkgloadingerror:all > "${TEST_log}" 2>&1 && fail "expected failure" |
| expect_log "^ERROR: $(pwd)/bzl/bzl.bzl:1:5: name 'invalidsyntax' is not defined" |
| |
| bazel build --attempt_to_print_relative_paths=true \ |
| pkgloadingerror:all > "${TEST_log}" 2>&1 && fail "expected failure" |
| expect_log "^ERROR: bzl/bzl.bzl:1:5: name 'invalidsyntax' is not defined" |
| expect_not_log "$(pwd)/bzl/bzl.bzl" |
| } |
| |
| function test_fancy_symbol_encoding() { |
| bazel build //fancyOutput:withFancyOutput > "${TEST_log}" 2>&1 \ |
| || fail "expected success" |
| expect_log $'\xF0\x9F\x8D\x83' |
| } |
| |
| function test_ui_events_filters() { |
| bazel clean || fail "${PRODUCT_NAME} clean failed" |
| |
| bazel build pkgloadingerror:all > "${TEST_log}" 2>&1 && fail "expected failure" |
| expect_log "^ERROR: .*/bzl/bzl.bzl:1:5: name 'invalidsyntax' is not defined" |
| expect_log "^WARNING: Target pattern parsing failed." |
| expect_log "^INFO: Elapsed time" |
| |
| bazel build --ui_event_filters=-error pkgloadingerror:all > "${TEST_log}" 2>&1 && fail "expected failure" |
| expect_not_log "^ERROR: .*bzl/bzl.bzl:1:5: name 'invalidsyntax' is not defined" |
| expect_log "^WARNING: Target pattern parsing failed." |
| expect_log "^INFO: Elapsed time" |
| |
| bazel build --ui_event_filters=info pkgloadingerror:all > "${TEST_log}" 2>&1 && fail "expected failure" |
| expect_not_log "^ERROR: .*/bzl/bzl.bzl:1:5: name 'invalidsyntax' is not defined" |
| expect_not_log "^WARNING: Target pattern parsing failed." |
| expect_log "^INFO: Elapsed time" |
| |
| bazel build --ui_event_filters= pkgloadingerror:all > "${TEST_log}" 2>&1 && fail "expected failure" |
| expect_not_log "^ERROR: .*/bzl/bzl.bzl:1:5: name 'invalidsyntax' is not defined" |
| expect_not_log "^WARNING: Target pattern parsing failed." |
| expect_not_log "^INFO: Elapsed time" |
| } |
| |
| function test_max_stdouterr_bytes_capping_behavior() { |
| mkdir -p outs |
| cat >outs/BUILD <<'EOF' |
| genrule( |
| name = "short-stdout", |
| outs = ["short-stdout.txt"], |
| cmd = "echo 'abc'; touch $@", |
| ) |
| genrule( |
| name = "short-stderr", |
| outs = ["short-stderr.txt"], |
| cmd = "echo 'abc' 1>&2; touch $@", |
| ) |
| genrule( |
| name = "long-stdout", |
| outs = ["long-stdout.txt"], |
| cmd = "echo 'long line of text'; touch $@", |
| ) |
| genrule( |
| name = "long-stderr", |
| outs = ["long-stderr.txt"], |
| cmd = "echo 'long line of text' 1>&2; touch $@", |
| ) |
| genrule( |
| name = "short-stdout-long-stderr", |
| outs = ["short-stdout-long-stderr.txt"], |
| cmd = "echo 'abc'; echo 'long line of text' 1>&2; touch $@", |
| ) |
| genrule( |
| name = "long-stdout-short-stderr", |
| outs = ["long-stdout-short-stderr.txt"], |
| cmd = "echo 'abc' 1>&2; echo 'long line of text'; touch $@", |
| ) |
| EOF |
| |
| for n in stdout stderr; do |
| bazel build --experimental_ui_max_stdouterr_bytes=5 "//outs:short-$n" \ |
| >"${TEST_log}" 2>&1 || fail "build failed" |
| expect_log 'abc' |
| expect_not_log 'exceeds maximum size' |
| |
| bazel build --experimental_ui_max_stdouterr_bytes=5 "//outs:long-$n" \ |
| >"${TEST_log}" 2>&1 || fail "build failed" |
| expect_not_log 'long line of text' |
| expect_log 'exceeds maximum size' |
| done |
| |
| bazel build --experimental_ui_max_stdouterr_bytes=5 \ |
| //outs:short-stdout-long-stderr \ |
| >"${TEST_log}" 2>&1 || fail "build failed" |
| expect_log 'abc' |
| expect_log 'stderr .*/actions/stderr-.* exceeds maximum size' |
| |
| bazel build --experimental_ui_max_stdouterr_bytes=5 \ |
| //outs:long-stdout-short-stderr \ |
| >"${TEST_log}" 2>&1 || fail "build failed" |
| expect_log 'abc' |
| expect_log 'stdout .*/actions/stdout-.* exceeds maximum size' |
| } |
| |
| function test_max_stdouterr_bytes_is_for_individual_outputs() { |
| mkdir -p outs |
| cat >>outs/BUILD <<'EOF' |
| [genrule( |
| name = "out-%d" % i, |
| outs = [("out-%d.txt") % i], |
| cmd = ("echo '>>>%d<<<'; touch $@") % i, |
| ) for i in range(1000)] |
| EOF |
| # Set --experimental_ui_max_stdouterr_bytes to a low-enough number that |
| # tolerates a single stdout and run many concurrent jobs at once to ensure |
| # that this limit applies to individual outputs even if we are trying to read |
| # all files concurrently. |
| bazel build --experimental_ui_max_stdouterr_bytes=20 --jobs=200 //outs/... \ |
| >"${TEST_log}" 2>&1 || fail "build failed" |
| for i in $(seq 0 999); do |
| expect_log ">>>${i}<<<" |
| done |
| } |
| |
| function test_interleaved_errors_and_progress() { |
| # Background process necessary to interrupt Bazel doesn't go well on Windows. |
| [[ "$is_windows" == true ]] && return |
| mkdir -p foo |
| cat > foo/BUILD <<'EOF' |
| genrule(name = 'sleep', outs = ['sleep.out'], cmd = 'sleep 10000') |
| genrule(name = 'fail', |
| outs = ['fail.out'], |
| srcs = [':multiline.sh'], |
| cmd = '$(location :multiline.sh)' |
| ) |
| EOF |
| cat > foo/multiline.sh <<'EOF' |
| echo "This |
| is |
| a |
| multiline error message |
| before |
| failure" |
| false |
| EOF |
| chmod +x foo/multiline.sh |
| bazel build -k //foo:all --curses=yes >& "$TEST_log" & |
| pid="$!" |
| while ! grep -q "multiline error message" "$TEST_log" ; do |
| sleep 1 |
| done |
| while ! grep -q '\[2 / 3\] Executing genrule //foo:sleep' "$TEST_log" ; do |
| sleep 1 |
| done |
| kill -SIGINT "$pid" |
| wait "$pid" || exit_code="$?" |
| [[ "$exit_code" == 8 ]] || fail "Should have been interrupted: $exit_code" |
| tr -s <"$TEST_log" '\n' '@' | |
| grep -q 'Executing genrule //foo:fail failed:[^@]*@This@is@a@multiline error message@before@failure@\[2 / 3\] Executing genrule //foo:sleep;' \ |
| || fail "Unified genrule error message not found" |
| # Make sure server is still usable. |
| bazel info server_pid >& "$TEST_log" || fail "Couldn't use server" |
| } |
| |
| function test_progress_bar_after_stderr() { |
| mkdir -p foo |
| cat > foo/BUILD <<'EOF' |
| genrule(name = 'fail', outs = ['fail.out'], cmd = 'false') |
| sh_test(name = 'foo', data = [':fail'], srcs = ['foo.sh']) |
| EOF |
| touch foo/foo.sh |
| chmod +x foo/foo.sh |
| # Build event file needed so UI considers build to continue after failure. |
| ! bazel test --build_event_json_file=bep.json --curses=yes --color=yes \ |
| //foo:foo &> "$TEST_log" || fail "Expected failure" |
| # Expect to see a failure message with an "erase line" control code prepended. |
| expect_log $'\e'"\[K"$'\e'"\[31m"$'\e'"\[1mFAILED:"$'\e'"\[0m Build did NOT complete successfully" |
| # We should not see a build failure message without an "erase line" to start. |
| # TODO(janakr): Fix the excessive printing of this failure message. |
| expect_log_n "^"$'\e'"\[31m"$'\e'"\[1mFAILED:"$'\e'"\[0m Build did NOT complete successfully" 4 |
| } |
| |
| run_suite "Integration tests for ${PRODUCT_NAME}'s UI" |