3 # Copyright (c) 2012-2015. The SimGrid Team.
6 # This program is free software; you can redistribute it and/or modify it
7 # under the terms of the license (GNU LGPL) which comes with this package.
8 eval 'exec perl -S $0 ${1+"$@"}'
9 if $running_under_some_shell;
11 # If you change this file, please stick to the formatting you got with:
12 # perltidy --backup-and-modify-in-place --maximum-line-length=180 --output-line-ending=unix --cuddled-else
22 B<tesh> [I<options>]... I<testsuite>
26 Tesh is the testing shell, a specialized shell for running tests. It
27 provides the specified input to the tested commands, and check that
28 they produce the expected output and return the expected value.
32 --cd some/directory : ask tesh to switch the working directory before
34 --setenv var=value : set a specific environment variable
35 --cfg arg : add parameter --cfg=arg to each command line
36 --enable-coverage : ignore output lines starting with "profiling:"
38 =head1 TEST SUITE FILE SYTAX
40 A test suite is composed of one or several I<command blocks> separated
41 by empty lines, each of them being composed of a command to run, its
42 input text and the expected output.
44 The first char of each line specifies the type of line according to
45 the following list. The second char of each line is ignored.
47 `$' command to run in foreground
48 `&' command to run in background
50 `<' input to pass to the command
51 `>' output expected from the command
53 `!' metacommand, which can be one of:
54 `timeout' <integer>|no
55 `expect signal' <signal name>
56 `expect return' <integer>
57 `output' <ignore|display>
60 `p' an informative message to print
62 If the expected output do not match the produced output, or if the
63 command did not end as expected, Tesh provides an error message (see
64 the OUTPUT section below) and stops.
66 =head2 Command blocks examples
68 In a given command block, you can declare the command, its input and
69 its expected output in the order that you see fit.
83 You can group several commands together, provided that they don't have
89 =head2 Enforcing the command return code
91 By default, Tesh enforces that the tested command returns 0. If not,
92 it fails with an appropriate message and returns I<code+40> itself.
94 You specify that a given command block is expected to return another
97 # This command MUST return 42
101 The I<expect return> construct applies only to the next command block.
103 =head2 Commands that are expected to raise signals
105 By default, Tesh detects when the command is killed by a signal (such
106 as SEGV on segfaults). This is usually unexpected and unfortunate. But
107 if not, you can specify that a given command block is expected to fail
108 with a signal as follows:
110 # This command MUST raise a segfault
111 ! expect signal SIGSEGV
112 $ ./some_failing_code
114 The I<expect signal> construct applies only to the next command block.
118 By default, no command is allowed to run more than 5 seconds. You can
119 change this value as follows:
121 # Allow some more time to the command
123 $ ./some_longer_command
125 You can also disable the timeout completely by passing "no" as a value:
127 # This command will never timeout
129 $ ./some_very_long_but_safe_command
131 =head2 Setting environment variables
133 You can modify the environment of the tested commands as follows:
138 =head2 Not enforcing the expected output
140 By default, the commands output is matched against the one expected,
141 and an error is raised on discrepancy. Metacommands to change this:
147 The output is completely discarded.
151 The output is displayed, but no error is issued if it differs from the
156 The output is sorted before comparison (see next section).
160 =head2 Sorting output
162 If the order of the command output changes between runs, you want to
163 sort it before enforcing that it is exactly what you expect. In
164 SimGrid for example, this happens when parallel execution is
165 activated: User processes are run in parallel at each timestamp, and
166 the output is not reproducible anymore. Until you sort the lines.
168 You can sort the command output as follows:
171 $ ./some_multithreaded_command
173 Sorting lines this ways often makes the tesh output very intricate,
174 complicating the error analysis: the process logical order is defeated
175 by the lexicographical sort.
177 The solution is to prefix each line of your output with temporal
178 information so that lines can be grouped by timestamps. The
179 lexicographical sort then only applies to lines that occured at the
180 same timestamp. Here is a SimGrid example:
182 # Sort only lines depending on the first 19 chars
184 $ ./some_simgrid_simulator --log=root.fmt:[%10.6r]%e(%i:%P@%h)%e%m%n
186 This approach may seem surprizing at the first glance but it does its job:
190 =item Every timestamps remain separated, as it should;
192 =item In each timestamp, the output order of processes become
193 reproducible: that's the lexicographical order of their name;
195 =item For each process, the order of its execution is preserved: its
196 messages within a given timestamp are not reordered.
200 That way, tesh can do its job (no false positive, no false negative)
201 despite the unpredictable order of executions of processes within a
202 timestamp, and reported errors remain easy to analyze (execution of a
203 given process preserved).
205 This example is very SimGrid oriented, but the feature could even be
206 usable by others, who knows?
209 =head1 BUILTIN COMMANDS
211 =head2 mkfile: creating a file
213 This command creates a file of the name provided as argument, and adds
214 the content it gets as input.
220 It is not possible to use the cat command, as one would expect,
221 because stream redirections are currently not implemented in Tesh.
223 =head1 BUGS, LIMITATIONS AND POSSIBLE IMPROVEMENTS
225 The main limitation is the lack of stream redirections in the commands
226 (">", "<" and "|" shell constructs and friends). The B<mkfile> builtin
227 command makes this situation bearable.
229 It would be nice if we could replace the tesh file completely with
230 command line flags when the output is not to be verified.
235 # Disabling IPC::Run::Debug saves tons of useless calls.
236 $ENV{'IPCRUNDEBUG'} = 'none'
237 unless exists $ENV{'IPCRUNDEBUG'};
240 my $enable_coverage = 0;
242 my $diff_tool_tmp_fh = 0;
243 my $diff_tool_tmp_filename = 0;
244 my $sort_prefix = -1;
251 $SIG{'PIPE'} = 'IGNORE';
257 use lib "@CMAKE_BINARY_DIR@/bin";
259 use Diff qw(diff); # postpone a bit to have time to change INC
261 use Getopt::Long qw(GetOptions);
263 use Text::ParseWords;
264 use IPC::Run qw(start run timeout finish);
269 #### Portability bits for windows
272 use constant RUNNING_ON_WINDOWS => ( $OSNAME =~ /^(?:mswin|dos|os2)/oi );
273 use POSIX qw(:sys_wait_h WIFEXITED WIFSIGNALED WIFSTOPPED WEXITSTATUS WTERMSIG WSTOPSIG
274 :signal_h SIGINT SIGTERM SIGKILL SIGABRT SIGSEGV);
277 if (RUNNING_ON_WINDOWS) { # Missing on windows
278 *WIFEXITED = sub { not $_[0] & 127 };
279 *WEXITSTATUS = sub { $_[0] >> 8 };
280 *WIFSIGNALED = sub { ( $_[0] & 127 ) && ( $_[0] & 127 != 127 ) };
281 *WTERMSIG = sub { $_[0] & 127 };
283 # used on the command lines
284 $environ{'EXEEXT'} = ".exe";
290 #### Command line option handling
293 my %opts = ( "debug" => 0,
294 "timeout" => 5, # No command should run any longer than 5 seconds by default
297 Getopt::Long::config( 'bundling', 'no_getopt_compat', 'no_auto_abbrev' );
299 'debug|d' => \$opts{"debug"},
301 'difftool=s' => \$diff_tool,
303 'cd=s' => sub { cd_cmd( $_[1] ) },
304 'timeout=s' => \$opts{'timeout'},
305 'setenv=s' => sub { setenv_cmd( $_[1] ) },
306 'cfg=s' => sub { $opts{'cfg'} .= " --cfg=$_[1]" },
307 'enable-coverage+' => \$enable_coverage,
310 $tesh_file = pop @ARGV;
311 $tesh_name = $tesh_file;
312 $tesh_name =~ s|^.*?/([^/]*)$|$1|;
314 print "Enable coverage\n" if ($enable_coverage);
317 use File::Temp qw/ tempfile /;
318 ( $diff_tool_tmp_fh, $diff_tool_tmp_filename ) = tempfile();
319 print "New tesh: $diff_tool_tmp_filename\n";
322 if ( $tesh_file =~ m/(.*)\.tesh/ ) {
323 print "Test suite `$tesh_file'\n";
325 $tesh_name = "(stdin)";
326 print "Test suite from stdin\n";
329 ###########################################################################
332 my %cmd = %{ $_[0] };
333 if ( $opts{'debug'} ) {
334 map { print "IN: $_\n" } @{ $cmd{'in'} };
335 map { print "OUT: $_\n" } @{ $cmd{'out'} };
336 print "CMD: $cmd{'cmd'}\n";
339 # substitute environment variables
340 foreach my $key ( keys %environ ) {
341 $cmd{'cmd'} = var_subst( $cmd{'cmd'}, $key, $environ{$key} );
344 # substitute remaining variables, if any
345 while ( $cmd{'cmd'} =~ /\${(\w+)(?::[=-][^}]*)?}/ ) {
346 $cmd{'cmd'} = var_subst( $cmd{'cmd'}, $1, "" );
348 while ( $cmd{'cmd'} =~ /\$(\w+)/ ) {
349 $cmd{'cmd'} = var_subst( $cmd{'cmd'}, $1, "" );
353 $cmd{'cmd'} .= " $opts{'cfg'}"
354 if ( defined( $opts{'cfg'} ) && length( $opts{'cfg'} ) );
356 # finally trim any remaining space chars
357 $cmd{'cmd'} =~ s/^\s+//;
358 $cmd{'cmd'} =~ s/\s+$//;
360 print "[$tesh_name:$cmd{'line'}] $cmd{'cmd'}\n";
362 $cmd{'return'} ||= 0;
363 $cmd{'timeout'} ||= $opts{'timeout'};
367 # exec the command line
369 my @cmdline = quotewords( '\s+', 0, $cmd{'cmd'} );
370 my $input = defined($cmd{'in'})? join("\n",@{$cmd{'in'}}) : "";
371 my $output = " " x 10240; $output = ""; # Preallocate 10kB, and reset length to 0
372 $cmd{'got'} = \$output;
373 $cmd{'job'} = start \@cmdline, '<', \$input, '>&', \$output,
374 ($cmd{'timeout'} eq 'no' ? () : timeout($cmd{'timeout'}));
376 if ( $cmd{'background'} ) {
377 # Just enqueue the job. It will be dealed with at the end
378 push @bg_cmds, \%cmd;
380 # Deal with its ending conditions right away
381 analyze_result( \%cmd );
386 my %cmd = %{ $_[0] };
387 $cmd{'timeouted'} = 0; # initialization
389 # Wait for the end of the child process
392 finish( $cmd{'job'} );
394 if ($@) { # deal with the errors that occured in the child process
395 if ($@ =~ /timeout/) {
396 $cmd{'job'}->kill_kill;
397 $cmd{'timeouted'} = 1;
398 } elsif ($@ =~ /^ack / and $@ =~ /pipe/) { # IPC::Run is not very expressive about the pipes that it gets :(
399 print STDERR "Tesh: Broken pipe (ignored).\n";
401 die $@; # Don't know what it is, so let it go.
408 # pop all output from executing child
410 map { print "GOT: $_\n" } ${$cmd{'got'}} if $opts{'debug'};
411 foreach my $got ( split("\n", ${$cmd{'got'}}) ) {
414 print $diff_tool_tmp_fh "> $got\n" if ($diff_tool);
416 unless ( $enable_coverage and $got =~ /^profiling:/ ) {
421 # How did the child process terminate?
423 $cmd{'gotret'} = "Unparsable status. Please report this tesh bug.";
424 if ( $cmd{'timeouted'} ) {
425 $cmd{'gotret'} = "timeout after $cmd{'timeout'} sec";
428 } elsif ( WIFEXITED($status) ) {
429 $exitcode = WEXITSTATUS($status) + 40;
430 $cmd{'gotret'} = "returned code " . WEXITSTATUS($status);
431 } elsif ( WIFSIGNALED($status) ) {
433 if ( WTERMSIG($status) == SIGINT ) { $code = "SIGINT"; }
434 elsif ( WTERMSIG($status) == SIGTERM ) { $code = "SIGTERM"; }
435 elsif ( WTERMSIG($status) == SIGKILL ) { $code = "SIGKILL"; }
436 elsif ( WTERMSIG($status) == SIGABRT ) { $code = "SIGABRT"; }
437 elsif ( WTERMSIG($status) == SIGSEGV ) { $code = "SIGSEGV"; }
438 $exitcode = WTERMSIG($status) + 4;
439 $cmd{'gotret'} = "got signal $code";
442 # How was it supposed to terminate?
444 if ( defined( $cmd{'expect'} ) and ( $cmd{'expect'} ne "" ) ) {
445 $wantret = "got signal $cmd{'expect'}";
447 $wantret = "returned code " . ( defined( $cmd{'return'} ) ? $cmd{'return'} : 0 );
450 # Enforce the outcome
453 # Did it end as expected?
454 if ( $cmd{'gotret'} ne $wantret ) {
456 my $msg = "Test suite `$tesh_name': NOK (<$tesh_name:$cmd{'line'}> $cmd{'gotret'})\n";
458 $msg = $msg . "Output of <$tesh_name:$cmd{'line'}> so far:\n";
459 map { $msg .= "|| $_\n" } @got;
461 $msg .= "<$tesh_name:$cmd{'line'}> No output so far.\n";
466 # Does the output match?
467 if ( $cmd{'sort'} ) {
469 substr( $a, 0, $sort_prefix ) cmp substr( $b, 0, $sort_prefix );
472 if ( $sort_prefix > 0 ) {
473 @got = sort mysort @got;
477 while ( @got and $got[0] eq "" ) {
481 # Sort the expected output too, to make tesh files easier to write for humans
482 if ( defined( $cmd{'out'} ) ) {
483 if ( $sort_prefix > 0 ) {
484 @{ $cmd{'out'} } = sort mysort @{ $cmd{'out'} };
486 @{ $cmd{'out'} } = sort @{ $cmd{'out'} };
488 while ( @{ $cmd{'out'} } and ${ $cmd{'out'} }[0] eq "" ) {
489 shift @{ $cmd{'out'} };
494 # Report the output if asked so or if it differs
495 if ( defined( $cmd{'output display'} ) ) {
496 print "[Tesh/INFO] Here is the (ignored) command output:\n";
497 map { print "||$_\n" } @got;
498 } elsif ( defined( $cmd{'output ignore'} ) ) {
499 print "(ignoring the output of <$tesh_name:$cmd{'line'}> as requested)\n";
501 my $diff = build_diff( \@{ $cmd{'out'} }, \@got );
503 if ( length $diff ) {
504 print "Output of <$tesh_name:$cmd{'line'}> mismatch" . ( $cmd{'sort'} ? " (even after sorting)" : "" ) . ":\n";
505 map { print "$_\n" } split( /\n/, $diff );
506 if ( $cmd{'sort'} ) {
507 print "WARNING: Both the observed output and expected output were sorted as requested.\n";
508 print "WARNING: Output were only sorted using the $sort_prefix first chars.\n"
509 if ( $sort_prefix > 0 );
510 print "WARNING: Use <! output sort 19> to sort by simulated date and process ID only.\n";
512 # print "----8<--------------- Begin of unprocessed observed output (as it should appear in file):\n";
513 # map {print "> $_\n"} @{$cmd{'unsorted got'}};
514 # print "--------------->8---- End of the unprocessed observed output.\n";
517 print "Test suite `$tesh_name': NOK (<$tesh_name:$cmd{'line'}> output mismatch)\n";
524 my $infh; # The file descriptor from which we should read the teshfile
525 if ( $tesh_name eq "(stdin)" ) {
528 open $infh, $tesh_file
529 or die "[Tesh/CRITICAL] Unable to open $tesh_file: $!\n";
532 my %cmd; # everything about the next command to run
534 LINE: while ( not $error and defined( my $line = <$infh> )) {
539 print "[TESH/debug] $line_num: $line\n" if $opts{'debug'};
541 # deal with line continuations
542 while ( $line =~ /^(.*?)\\$/ ) {
544 die "[TESH/CRITICAL] Continued line at end of file\n"
545 unless defined($next);
548 print "[TESH/debug] $line_num: $next\n" if $opts{'debug'};
552 # If the line is empty, run any previously defined block and proceed to next line
553 unless ( $line =~ m/^(.)(.*)$/ ) {
554 if ( defined( $cmd{'cmd'} ) ) {
558 print $diff_tool_tmp_fh "$line\n" if ($diff_tool);
562 my ( $cmd, $arg ) = ( $1, $2 );
563 print $diff_tool_tmp_fh "$line\n" if ( $diff_tool and $cmd ne '>' );
568 # Deal with the lines that can contribute to the current command block
569 if ( $cmd =~ /^#/ ) { # comment
571 } elsif ( $cmd eq '>' ) { # expected result line
572 print "[TESH/debug] push expected result\n" if $opts{'debug'};
573 push @{ $cmd{'out'} }, $arg;
576 } elsif ( $cmd eq '<' ) { # provided input
577 print "[TESH/debug] push provided input\n" if $opts{'debug'};
578 push @{ $cmd{'in'} }, $arg;
581 } elsif ( $cmd eq 'p' ) { # comment
582 print "[$tesh_name:$line_num] $arg\n";
587 # We dealt with all sort of lines that can contribute to a command block, so we have something else here.
588 # If we have something buffered, run it now and start a new block
589 if ( defined( $cmd{'cmd'} ) ) {
594 # Deal with the lines that must be placed before a command block
595 if ( $cmd eq '$' ) { # Command
596 if ( $arg =~ /^mkfile / ) { # "mkfile" command line
597 die "[TESH/CRITICAL] Output expected from mkfile command!\n"
598 if scalar @{ cmd { 'out' } };
601 $cmd{'arg'} =~ s/mkfile //;
605 } elsif ( $arg =~ /^\s*cd / ) {
606 die "[TESH/CRITICAL] Input provided to cd command!\n"
607 if scalar @{ cmd { 'in' } };
608 die "[TESH/CRITICAL] Output expected from cd command!\n"
609 if scalar @{ cmd { 'out' } };
615 } else { # regular command
617 $cmd{'line'} = $line_num;
620 } elsif ( $cmd eq '&' ) { # background command line
621 die "[TESH/CRITICAL] mkfile cannot be run in background\n"
622 if ($arg =~ /^mkfile/);
623 die "[TESH/CRITICAL] cd cannot be run in background\n"
626 $cmd{'background'} = 1;
628 $cmd{'line'} = $line_num;
630 # Deal with the meta-commands
631 } elsif ( $line =~ /^! (.*)/) {
634 if ( $line =~ /^output sort/ ) {
636 if ( $line =~ /^output sort\s+(\d+)/ ) {
639 } elsif ($line =~ /^output ignore/ ) {
640 $cmd{'output ignore'} = 1;
641 } elsif ( $line =~ /^output display/ ) {
642 $cmd{'output display'} = 1;
643 } elsif ( $line =~ /^expect signal (\w*)/ ) {
645 } elsif ( $line =~ /^expect return/ ) {
646 $line =~ s/^expect return //g;
648 $cmd{'return'} = $line;
649 } elsif ( $line =~ /^setenv/ ) {
650 $line =~ s/^setenv //g;
653 } elsif ( $line =~ /^timeout/ ) {
654 $line =~ s/^timeout //;
656 $cmd{'timeout'} = $line;
659 die "[TESH/CRITICAL] parse error: $line\n";
663 # We are done reading the input file
664 close $infh unless ( $tesh_name eq "(stdin)" );
666 # Deal with last command, if any
667 if ( defined( $cmd{'cmd'} ) ) {
674 analyze_result( \%test );
678 close $diff_tool_tmp_fh;
679 system("$diff_tool $diff_tool_tmp_filename $tesh_file");
680 unlink $diff_tool_tmp_filename;
685 } elsif ( $tesh_name eq "(stdin)" ) {
686 print "Test suite from stdin OK\n";
688 print "Test suite `$tesh_name' OK\n";
694 #### Helper functions
699 my $diff = Diff->new(@_);
701 $diff->Base(1); # Return line numbers, not indices
702 my $chunk_count = $diff->Next(-1); # Compute the amount of chuncks
703 return "" if ( $chunk_count == 1 && $diff->Same() );
705 while ( $diff->Next() ) {
706 my @same = $diff->Same();
707 if ( $diff->Same() ) {
708 if ( $diff->Next(0) > 1 ) { # not first chunk: print 2 first lines
709 $res .= ' ' . $same[0] . "\n";
710 $res .= ' ' . $same[1] . "\n" if ( scalar @same > 1 );
712 $res .= "...\n" if ( scalar @same > 2 );
714 # $res .= $diff->Next(0)."/$chunk_count\n";
715 if ( $diff->Next(0) < $chunk_count ) { # not last chunk: print 2 last lines
716 $res .= ' ' . $same[ scalar @same - 2 ] . "\n"
717 if ( scalar @same > 1 );
718 $res .= ' ' . $same[ scalar @same - 1 ] . "\n";
721 next if $diff->Same();
722 map { $res .= "- $_\n" } $diff->Items(1);
723 map { $res .= "+ $_\n" } $diff->Items(2);
728 # Helper function replacing any occurence of variable '$name' by its '$value'
729 # As in Bash, ${$value:=BLABLA} is rewritten to $value if set or to BLABLA if $value is not set
731 my ( $text, $name, $value ) = @_;
733 $text =~ s/\${$name(?::[=-][^}]*)?}/$value/g;
734 $text =~ s/\$$name(\W|$)/$value$1/g;
736 $text =~ s/\${$name:=([^}]*)}/$1/g;
737 $text =~ s/\${$name}//g;
738 $text =~ s/\$$name(\W|$)/$1/g;
743 ################################ The possible commands ################################
746 my %cmd = %{ $_[0] };
747 my $file = $cmd{'arg'};
748 print STDERR "[Tesh/INFO] mkfile $file. Ctn: >>".join( '\n', @{ $cmd{'in'} })."<<\n"
752 open( FILE, ">$file" )
753 or die "[Tesh/CRITICAL] Unable to create file $file: $!\n";
754 print FILE join( "\n", @{ $cmd{'in'} } );
755 print FILE "\n" if ( scalar @{ $cmd{'in'} } > 0 );
759 # Command CD. Just change to the provided directory
761 my $directory = shift;
763 if ( -e $directory && -d $directory ) {
765 print "[Tesh/INFO] change directory to $directory\n";
767 } elsif ( -e $directory ) {
768 print "Cannot change directory to '$directory': it is not a directory\n";
770 print "Chdir to $directory failed: No such file or directory\n";
772 if ( $failure == 1 ) {
773 print "Test suite `$tesh_name': NOK (system error)\n";
778 # Command setenv. Gets "variable=content", and update the environment accordingly
781 if ( $arg =~ /^(.*)=(.*)$/ ) {
782 my ( $var, $ctn ) = ( $1, $2 );
783 print "[Tesh/INFO] setenv $var=$ctn\n";
784 $environ{$var} = $ctn;
787 die "[Tesh/CRITICAL] Malformed argument to setenv: expected 'name=value' but got '$arg'\n";