2 # vim:ts=4:sw=4:expandtab
3 # © 2010-2012 Michael Stapelberg and contributors
9 # the following are modules which ship with Perl (>= 5.10):
12 use File::Temp qw(tempfile tempdir);
17 use TAP::Parser::Aggregator;
18 use Time::HiRes qw(time);
20 # these are shipped with the testsuite
25 # the following modules are not shipped with Perl
29 use AnyEvent::I3 qw(:all);
30 use X11::XCB::Connection;
31 use JSON::XS; # AnyEvent::I3 depends on it, too.
33 binmode STDOUT, ':utf8';
34 binmode STDERR, ':utf8';
36 # Close superfluous file descriptors which were passed by running in a VIM
37 # subshell or situations like that.
38 AnyEvent::Util::close_all_fds_except(0, 1, 2);
40 # convinience wrapper to write to the log file
42 sub Log { say $log "@_" }
46 # Number of tests to run in parallel. Important to know how many Xephyr
47 # instances we need to start (unless @displays are given). Defaults to
58 my $keep_xserver_output = 0;
60 my $result = GetOptions(
61 "coverage-testing" => \$options{coverage},
62 "keep-xserver-output" => \$keep_xserver_output,
63 "valgrind" => \$options{valgrind},
64 "strace" => \$options{strace},
65 "xtrace" => \$options{xtrace},
66 "display=s" => \@displays,
67 "parallel=i" => \$parallel,
71 pod2usage(-verbose => 2, -exitcode => 0) if $help;
73 # Check for missing executables
77 ../i3-config-wizard/i3-config-wizard
78 ../i3-dump-log/i3-dump-log
81 ../i3-nagbar/i3-nagbar
84 foreach my $binary (@binaries) {
85 die "$binary executable not found, did you run “make”?" unless -e $binary;
86 die "$binary is not an executable" unless -x $binary;
89 if ($options{coverage}) {
90 qx(command -v lcov &> /dev/null);
91 die "Cannot find lcov needed for coverage testing." if $?;
92 qx(command -v genhtml &> /dev/null);
93 die "Cannot find genhtml needed for coverage testing." if $?;
95 # clean out the counters that may be left over from previous tests.
96 qx(lcov -d ../ --zerocounters &> /dev/null);
99 qx(Xephyr -help 2>&1);
100 die "Xephyr was not found in your path. Please install Xephyr (xserver-xephyr on Debian)." if $?;
102 @displays = split(/,/, join(',', @displays));
103 @displays = map { s/ //g; $_ } @displays;
105 # 2: get a list of all testcases
106 my @testfiles = @ARGV;
108 # if no files were passed on command line, run all tests from t/
109 @testfiles = <t/*.t> if @testfiles == 0;
111 my $numtests = scalar @testfiles;
113 # No displays specified, let’s start some Xephyr instances.
114 if (@displays == 0) {
115 @displays = start_xserver($parallel, $numtests, $keep_xserver_output);
118 # 1: create an output directory for this test-run
119 my $outdir = "testsuite-";
120 $outdir .= POSIX::strftime("%Y-%m-%d-%H-%M-%S-", localtime());
121 $outdir .= `git describe --tags`;
123 mkdir($outdir) or die "Could not create $outdir";
124 unlink("latest") if -l "latest";
125 symlink("$outdir", "latest") or die "Could not symlink latest to $outdir";
128 # connect to all displays for two reasons:
129 # 1: check if the display actually works
130 # 2: keep the connection open so that i3 is not the only client. this prevents
131 # the X server from exiting
133 for my $display (@displays) {
135 my $x = X11::XCB::Connection->new(display => $display);
137 die "Could not connect to display $display\n";
139 # start a TestWorker for each display
140 push @single_worker, worker($display, $x, $outdir, \%options);
144 # Read previous timing information, if available. We will be able to roughly
145 # predict the test duration and schedule a good order for the tests.
146 my $timingsjson = StartXServer::slurp('.last_run_timings.json');
147 %timings = %{decode_json($timingsjson)} if length($timingsjson) > 0;
149 # Re-order the files so that those which took the longest time in the previous
150 # run will be started at the beginning to not delay the whole run longer than
152 @testfiles = map { $_->[0] }
153 sort { $b->[1] <=> $a->[1] }
154 map { [$_, $timings{$_} // 999] } @testfiles;
156 # Run 000-load-deps.t first to bail out early when dependencies are missing.
157 my $loadtest = "t/000-load-deps.t";
158 if ((scalar grep { $_ eq $loadtest } @testfiles) > 0) {
159 @testfiles = ($loadtest, grep { $_ ne $loadtest } @testfiles);
162 printf("\nRough time estimate for this run: %.2f seconds\n\n", $timings{GLOBAL})
163 if exists($timings{GLOBAL});
165 # Forget the old timings, we don’t necessarily run the same set of tests as
166 # before. Otherwise we would end up with left-overs.
167 %timings = (GLOBAL => time());
169 my $logfile = "$outdir/complete-run.log";
170 open $log, '>', $logfile or die "Could not create '$logfile': $!";
172 say "Writing logfile to '$logfile'...";
176 my $num = @testfiles;
177 my $harness = TAP::Harness->new({ });
179 my $aggregator = TAP::Parser::Aggregator->new();
180 $aggregator->start();
182 status_init(displays => \@displays, tests => $num);
184 my $single_cv = AE::cv;
186 # We start tests concurrently: For each display, one test gets started. Every
187 # test starts another test after completing.
188 for (@single_worker) {
190 take_job($_, $single_cv, \@testfiles);
197 # print empty lines to seperate failed tests from statuslines
201 my ($test, $output) = @$_;
202 say "no output for $test" unless $output;
203 Log "output for $test:";
205 # print error messages of failed tests
206 say for $output =~ /^not ok.+\n+((?:^#.+\n)+)/mg
210 $harness->summary($aggregator);
214 # 5: Save the timings for better scheduling/prediction next run.
215 $timings{GLOBAL} = time() - $timings{GLOBAL};
216 open(my $fh, '>', '.last_run_timings.json');
217 print $fh encode_json(\%timings);
220 # 6: Print the slowest test files.
221 my @slowest = map { $_->[0] }
222 sort { $b->[1] <=> $a->[1] }
223 map { [$_, $timings{$_}] }
224 grep { !/^GLOBAL$/ } keys %timings;
226 say 'The slowest tests are:';
227 printf("\t%s with %.2f seconds\n", $_, $timings{$_})
228 for @slowest[0..($#slowest > 4 ? 4 : $#slowest)];
230 # When we are running precisely one test, print the output. Makes developing
231 # with a single testcase easier.
232 if ($numtests == 1) {
235 say StartXServer::slurp($logfile);
240 if ($options{coverage}) {
241 print("\nGenerating test coverage report...\n");
242 qx(lcov -d ../ -b ../ --capture -o latest/i3-coverage.info);
243 qx(genhtml -o latest/i3-coverage latest/i3-coverage.info);
245 print("Could not generate test coverage html. Did you compile i3 with test coverage support?\n");
247 print("Test coverage report generated in latest/i3-coverage\n");
251 exit ($aggregator->failed > 0);
254 # Takes a test from the beginning of @testfiles and runs it.
256 # The TAP::Parser (which reads the test output) will get called as soon as
257 # there is some activity on the stdout file descriptor of the test process
258 # (using an AnyEvent->io watcher).
260 # When a test completes and @done contains $num entries, the $cv condvar gets
261 # triggered to finish testing.
264 my ($worker, $cv, $tests) = @_;
266 my $test = shift @$tests
269 my $display = $worker->{display};
271 Log status($display, "$test: starting");
272 $timings{$test} = time();
273 worker_next($worker, $test);
275 # create a TAP::Parser with an in-memory fh
277 my $parser = TAP::Parser->new({
278 source => do { open(my $fh, '<', \$output); $fh },
281 my $ipc = $worker->{ipc};
288 state $tests_completed = 0;
291 sysread($ipc, my $buf, 4096) or die "sysread: $!";
294 $buf = $partial . $buf;
298 # make sure we feed TAP::Parser complete lines so it doesn't blow up
299 if (substr($buf, -1, 1) ne "\n") {
300 my $nl = rindex($buf, "\n");
306 # strip partial from buffer
307 $partial = substr($buf, $nl + 1, '');
310 # count lines before stripping eof-marker otherwise we might
311 # end up with for (1 .. 0) { } which would effectivly skip the loop
312 my $lines = $buf =~ tr/\n//;
313 my $t_eof = $buf =~ s/^$TestWorker::EOF$//m;
318 my $result = $parser->next;
319 next unless defined($result);
320 if ($result->is_test) {
322 status($display, "$test: [$tests_completed/??] ");
323 } elsif ($result->is_bailout) {
324 Log status($display, "$test: BAILOUT");
325 status_completed(scalar @done);
327 say "test $test bailed out: " . $result->explanation;
332 return unless $t_eof;
334 Log status($display, "$test: finished");
335 $timings{$test} = time() - $timings{$test};
336 status_completed(scalar @done);
338 $aggregator->add($test, $parser);
339 push @done, [ $test, $output ];
342 take_job($worker, $cv, $tests);
349 $_->() for our @CLEANUP;
353 # must be in a begin block because we C<exit 0> above
356 require Carp; Carp::cluck("Caught SIG$_[0]\n");
358 } for qw(INT TERM QUIT KILL PIPE)
365 complete-run.pl - Run the i3 testsuite
369 complete-run.pl [files...]
373 To run the whole testsuite on a reasonable number of Xephyr instances (your
374 running X11 will not be touched), run:
377 To run only a specific test (useful when developing a new feature), run:
378 ./complete-run t/100-fullscreen.t
386 Specifies which X11 display should be used. Can be specified multiple times and
387 will parallelize the tests:
389 # Run tests on the second X server
390 ./complete-run.pl -d :1
392 # Run four tests in parallel on some Xephyr servers
393 ./complete-run.pl -d :1,:2,:3,:4
395 Note that it is not necessary to specify this anymore. If omitted,
396 complete-run.pl will start (num_cores * 2) Xephyr instances.
400 Runs i3 under valgrind to find memory problems. The output will be available in
401 C<latest/valgrind-for-$test.log>.
405 Runs i3 under strace to trace system calls. The output will be available in
406 C<latest/strace-for-$test.log>.
410 Runs i3 under xtrace to trace X11 requests/replies. The output will be
411 available in C<latest/xtrace-for-$test.log>.
413 =item B<--coverage-testing>
415 Generates a test coverage report at C<latest/i3-coverage>. Exits i3 cleanly
416 during tests (instead of kill -9) to make coverage testing work properly.
420 Number of Xephyr instances to start (if you don't want to start num_cores * 2
421 instances for some reason).
423 # Run all tests on a single Xephyr instance
424 ./complete-run.pl -p 1