#!/bin/sh # -*-Perl-*- (for Emacs) vim:set filetype=perl: (for vim) #======================================================================# # Run the right perl version: if [ -x /usr/local/bin/perl ]; then perl=/usr/local/bin/perl elif [ -x /usr/bin/perl ]; then perl=/usr/bin/perl else perl=`which perl| sed 's/.*aliased to *//'` fi exec $perl -x -S $0 "$@" # -x: start from the following line #======================================================================# #! /Good_Path/perl -w # line 17 # Name: auto-test # Author: wd (Wolfgang.Dobler@kis.uni-freiburg.de) # Date: 12-Aug-2002 # SVN: $Id$ # Description: # Test compilation and results on a set of sample directories to # validate the pencil code. Uses Perl because I started to do this with # the Test/Test::Harness module and because we need to do pattern # matching to extract the relevant output lines. # To Do: # - Add an option to exclude those tests that use MPI # - Remove directory from /tmp when encountering Ctl-c # - Treat `svn update' as separate stage that can be `OK' or not BEGIN { # Make sure ${PENCIL_HOME}/lib/perl is in the Perl path if (-d "$ENV{PENCIL_HOME}/lib/perl") { unshift @INC, "$ENV{PENCIL_HOME}/lib/perl"; } else { if ($0 =~ m!(.*[/\\])!) { unshift @INC, "$1../lib/perl"; } } } use Pencil::Util; Pencil::Util::use_pencil_perl_modules('Pencil::ReadmeReader') or die; use strict; use constant NOQUIET => 0; use constant QUIET => 1; use constant LOCK_FAILED => 128; # return status if we can't get lock use Cwd; use POSIX qw(floor); use Getopt::Long; use vars qw{ %failed $test_status }; # do we really need global vars here? use IPC::Open2; # ---------------------------------------------------------------------- # # Test the following subdirs of samples/ . # Tests are grouped by `level': # 0 is the absolute minimum, # 1 are the most important tests, # 2 is the standard program for nightly tests, # 3 adds the remaining tests under samples (most of which will curently # fail). # [Note: until we get closer to have all test running, we will use more # levels to distinguish the different degrees of failures] # Within a level, tests should be roughly sorted by total run time, such # that the slowest tests are run last. my $default_level = 2; my %tests = ( 0 => [qw( no-modules most-modules )], 1 => [qw( backwards-compatible 2d-tests/selfgravitating-shearwave 2d-tests/chemistry_GrayScott 0d-tests/heating_ionize dust-vortex )], 2 => [qw( 0d-tests/heating_noionize 1d-tests/H2_flamespeed chiral-diffusion cosmicray 1d-tests/sod_10 1d-tests/ambipolar_diffusion helical-MHDturb testfield_z random_uu_particles 1d-tests/solar-atmosphere-temperature kin-dynamo meissner conv-slab-noequi conv-slab 2d-tests/globaldisc 2d-tests/cylindrical_gdisk 0d-tests/solar_system mdwarf sedimentation MRI-turbulence_hyper dust_turb_globaldisk geodynamo 2d-tests/cylinder_deposition 2d-tests/selfgravdisk-deadzone-dust 2d-tests/spherical_viscous_ring 2d-tests/field-loop-fargo baroclinic 1d-tests/sedimentation-noneq-blocks 1d-tests/alphadisk 2d-tests/torque_migration_noniso )], # Level 3 for now contains all non-canonical tests that run # successfully 3 => [qw( 0d-tests/heating_ionize_planck 1d-tests/jeans-drag-dustpar-x 1d-tests/jeans-x 1d-tests/sod_10s 1d-tests/toomre-x 2d-tests/bubble 2d-tests/conv-slab-MLT supernova-driven-turbulence/1Dhd_init_equidistant/ supernova-driven-turbulence/1Dhd_init_non-equidistant/ supernova-driven-turbulence/3Dhd_equidistant/ supernova-driven-turbulence/3Dhd_non-equidistant/ supernova-driven-turbulence/sedov/ supernova-driven-turbulence/snowplough/ interlocked-fluxrings conv-slab-flat corona coronae_heatflux_boris spherical-convection testfield testfield_xz testfield_nonlin_z )], # Level 4 for now contains overlong tests (e.g. test that # compile, start and then run for more than 5 minutes on # Frenesi -- this is at least as good as the tests in level # 5, but not practical). # # Note: The following _canonical_ tests fail to finish within 5 # minutes: # random_uu_particles # interstellar # geodynamo # 0d-tests/heating_ionize 4 => [qw( forced-boundary sound-spherical-bufferzone 2d-tests/shearwave-dust-par 2d-tests/selfgrav-shearwave-dust-par )], # Level 5 for now contains all non-canonical tests that run, # but produce wrong results 5 => [qw( interstellar hexagon MHD-ABCforcing taylor-couette potential_field 1d-tests/sod_100 1d-tests/sod_10y geodynamo-noequi 1d-tests/sod_1000 1d-tests/sod_100s 1d-tests/sod_10sy 1d-tests/sod_10sz 2d-tests/GMSW1976 1d-tests/sod_1000s 1d-tests/sod_10sho taylor-couette-cyl 0d-tests/chemistry_aped 2d-tests/GMSW1976_cpne1 0d-tests/coag_kernel_lin 2d-tests/dynamical_alpha 2d-tests/potential_field 2d-tests/spherical_gdisk solar-atmosphere-magnetic MRI-turbulence_hyper_strict 0d-tests/heating_ionize_temp 1d-tests/solar-atmosphere )], # Level 6 for now contains all non-canonical tests that # start, but don't run successfully 6 => [qw( 0d-tests/coagulation-fragmentation )], # Level 7 for now contains all non-canonical tests that # compile, but don't start sucessfully 7 => [qw( interstellar_ion spherical-globaldisk turb-cylindrical-disk 2d-tests/A3+chi11+Ra1e5 )], # Level 8 for now contains all non-canonical tests that don't # compile sucessfully 8 => [qw( multigrid )], ); # ---------------------------------------------------------------------- # my $ntests = 0; # total number of tests run my $failed = 0; my $created_datadir = 0; my $remove_lock = 0; my $failure_message = ''; my $reference_out = 'reference.out'; # automatically changed to # reference.out.double below for # double precision runs my $user = $ENV{USER} || $ENV{LOGNAME} || 'unknown'; my $tmpdir = "/tmp/pencil-tmp-$user-$$"; my $lockdir = '/tmp'; # put / check for lock file here my $lockfile = "$lockdir/pencil-auto-test-$user.pid"; my $headerpat = '^-*\s*(-+\s*[A-Za-z0-9_]+)+-*\s*$'; # regexp for header line my $noheader = "Couldn't find header line"; my $cfloat = '([+-]?)(?=\d|\.\d)\d*(\.\d*)?([Ee]([+-]?\d+))?'; # regexp # for C float my $ieee_float = "(?:$cfloat|[+-]?(?:NaN|Inf))"; # C float | ±NaN | ±Inf my $res; my ($t_comp,$t_start,$t_run,$t_res) = (1,1,1,1); # test everything by default my (%opts); # variables written by GetOptions (my $cmdname = $0) =~ s{.*/}{}; my $help = 0; my $itest = 1; my $indi_fmt = '%-17s'; # allow that much space for indicators like # `[double]' or `[KEEP_RELOADING]' my $usage = "Usage: $cmdname [options] [dir1 [dir2 [..]]] Test compilation and results on a set of sample directories (or on the list of directories given in the command line) to validate the pencil code. Uses Perl because we need to do pattern matching to extract the relevant output lines. If you don't have a data directory, $cmdname creates a link to $tmpdir -- both link and directory are removed after each run. Options: -h, --help \tShow usage overview -v, --version \tPrint version number -C, --clean \tClean up (make cleann) before compiling -c, --compile-only \tTest compilation only -n, --norun \tTest compilation and starting only -f, --results-from-file \tDon't run, read results from data/timeseries.dat -r, --reload \tKeep reloading parameters (using KEEP_RELOADING) -l, --list \tList test directories (don't run any tests) -L, --list-all \tList even test dirs lacking reference.out -t, --time \tPrint CPU time used -T, --summarize-times \tPrint timings line for pencil-test and statistics -i, --infofile= \tAfter last test, print content of -D, --pencil-home= \tSet PENCIL_HOME directory to -N, --nice= \tRun tests with nice value -j, --jobs= \tPass through -j option to make -d, --datadir \tUse mkdir if data/ is needed, don't ln -s /tmp -s, --short \tPrint short directory names only -p, --postproc=PP \tApply postprocessor PP to output from make -a, --all-dirs \tRecurse through all subdirectories of samples/ --level=LEV \tRun only tests from level LEV --max-level=LEV \tRun all tests below with level <= LEV (default: 2) --time-limit=limit \tUse reaper to limit time for {start,run}.x --list-maintainers \tPrint maintainer emails for failed tests --rsh=RSH-PROG \tUse the specified program as to get a remote shell --nodelist=NODE1,.. \tUse a forked remote shell to parallelize --no-summary \tDo not summarize at the end --no-lock \tIgnore and don't write lock file --debug \tPrint lots of debugging output Examples: auto-test # run auto-test standard program, recycling .o files auto-test -C # run auto-test standard program, compiling from scratch auto-test . # run auto-test in current directory auto-test -l # list dirs in standard program auto-test --time-limit=2m # run standard program, killing (for each test) # any remaining start.x or run.x processes # after 2 minutes. # Mainly useful for run.x, but we also limit # start.x o avoid hour-long runaway processes auto-test -a # run auto-test in all subdirs of samples/ with a # referece.out{,.double} file auto-test --level=1 # run all level 1 tests auto-test -l --max-level=1 # list test dirs in levels 0 and 1 auto-test -al # list all dirs 'auto-test -a' would run in auto-test -aL # list all run dirs under samples/, even those without # reference data "; ## Process command line options eval { Getopt::Long::config("bundling"); # make single-letter opts. case-sensitive }; GetOptions(\%opts, qw( -h --help -v --version --debug -C --clean -c --compile-only -n --norun -f --results-from-file -r --reload -l --list -L --list-all -t --time -T --summarize-times -i=s --infofile=s -D=s --pencil-home=s -N=s --nice=s -j=n --jobs=n -d --datadir -s --short -p=s --postproc=s -a --all-dirs --level=s --max-level=s --time-limit=s --list-maintainers --rsh=s --nodelist=s --no-summary --nolock --no-lock )) or $help=1, die "Aborting.\n"; if ($opts{'h'} || $opts{'help'}) { $help=1; die "$usage\n"; } if ($opts{'v'} || $opts{'version'}) { $help=1; die '$Id$ ' . "\n"; } if ($opts{'c'} || $opts{'compile-only'}) { $t_start=$t_run=$t_res=0 }; if ($opts{'n'} || $opts{'norun'}) { $t_run=0 }; my $res_from_file = ($opts{'f'} || $opts{'res_from_file'} || 0 ); if ($res_from_file) { $t_comp=$t_start=$t_run=0 }; my $debug = ( $opts{'debug'} || 0 ); my $clean = ($opts{'C'} || $opts{'clean'} || 0 ); my $reload = ($opts{'r'} || $opts{'reload'} || 0 ); my $list_only = ($opts{'l'} || $opts{'list'} || 0 ); my $list_all = ($opts{'L'} || $opts{'list-all'} || 0 ); my $time = ($opts{'t'} || $opts{'time'} || 0 ); my $summarize_t = ($opts{'T'} || $opts{'summarize-times'} || 0 ); my $infofile = ($opts{'i'} || $opts{'infofile'} || '' ); my $pencil_home = ($opts{'D'} || $opts{'pencil-home'} || "$ENV{PENCIL_HOME}" || ''); my $niceval = ($opts{'N'} || $opts{'nice'} || 0 ); my $makejobs = ($opts{'j'} || $opts{'jobs'} || 0 ); my $mkdatadir = ($opts{'d'} || $opts{'datadir'} || 0 ); my $short = ($opts{'s'} || $opts{'short'} || 0 ); my $postproc = ($opts{'p'} || $opts{'postproc'} || '' ); my $deep = ($opts{'a'} || $opts{'all-dirs'} || 0 ); my $level = ( $opts{'level'} ); my $max_level = ( $opts{'max-level'} ); my $time_limit = ( $opts{'time-limit'} || '' ); my $list_maint = ( $opts{'list-maintainers'} || 0 ); my $rsh = ( $opts{'rsh'} || 'ssh'); my $nodelist = ( $opts{'nodelist'} || 0 ); my $nosummary = ( $opts{'no-summary'} || 0 ); my $nolock = ( $opts{'nolock'} || $opts{'no-lock'} || 0 ); my @exit_status=('setting up directory', 'cleaning up', 'compilation', 'starting', 'running', 'results'); $list_only = 1 if ($list_all); my $makeopts = ''; if ($makejobs) { $makeopts = "-j $makejobs"; } my $host = `hostname`; chomp($host); my $nice = ""; if ($niceval != 0) { $nice = "\\nice -n $niceval" }; # Note: Csh's builtin `nice' cmd has a different calling syntax, so we # need to use ther real nice cmd here. my @testdirs; # full names of all dirs to run tests in my %time_summary; # hash to accumulate timing summary # within each run ## Debugging output: if ($debug) { eval("use Data::Dumper"); # Need eval here to calm down compiler print STDERR Dumper(\%opts); print STDERR "(\$t_comp,\$t_start,\$t_run,\$t_res)" . " = ($t_comp,$t_start,$t_run,$t_res)\n"; print STDERR "(\$debug,\$help,\$short) " . " = ($debug,$help,$short)\n"; } # $t_comp .. $t_res form a hierarchy: We can't test running without # compiling and starting $t_start &&= $t_comp; $t_run &&= $t_start; $t_res &&= ($t_run || $res_from_file); # Check for lock file and write one handle_lock_file() or exit(LOCK_FAILED) unless ($list_only); # Remember current directory my $cwd = `pwd`; chomp($cwd); # pc_update_autotest system("$ENV{PENCIL_HOME}/bin/pc_update_autotest.sh &"); # Make sure we are in the top directory and have the right PATH die "Need to set environment variable PENCIL_HOME\n" unless (defined($pencil_home)); my $topdir = "$pencil_home"; chdir $topdir; $ENV{PATH} .= ":$pencil_home/bin"; # Don't follow these boring subdirectories when recursing through whole tree: my $prunedirs = '^(data|src|CVS|RCS|_darcs|.hg|.svn)$'; # Autoflush stdout: $| = 1; @ARGV = grep { /./ } @ARGV; # Eliminate empty @ARGV list (needed under # OSF1, no clue why) if ($list_only && ! $short) { print "Test directories:\n" } if ($summarize_t) { write_timing_summary_header(); } my %t_global = ( 'compile' => 0, 'start+run' => 0, ); my $t0_global = get_time(); # remember start time if (@ARGV) { ## Process dirs given in cmd line @ARGV = map { s{^(?!/)}{$cwd/}; $_ } @ARGV; # Make `.' and similar work $ntests = @ARGV; for my $d (@ARGV) { test_rundir("$d",$makeopts); } if ((scalar @ARGV)==1) { if (%failed) { # Do some bizarre chicken sacrifice voodoo # [wd 29-Jan-2007:] Tried to sanitize this without knowing what # it was supposed to do # Assuming that the exit code is supposed to indicate the # worst phase that failed (i.e., 2 means setting up directory # failed, ..., 7 means results are wrong, 8 means something # unexpected failed, see @exit_status above): my $exit_code=1; # Loop through exit codes, starting with most serious one: foreach my $status (@exit_status) { # Exit if any of the exit codes stored in %failed matches # $status: if (grep /^$status$/, values %failed) { exit $exit_code; } else { $exit_code++; } } exit $exit_code; # Still here? Then exit now. } } } else { ## No dirs in cmd line -> do standard program if ($deep) { # Scan samples/ tree recursively find_test_dirs("$topdir/samples"); # populates @testdirs } else { # Use predefined directories my @sampdirs; if (defined($level)) { @sampdirs = get_sampdirs($level); } else { $max_level = $default_level unless defined($max_level); @sampdirs = get_all_sampdirs($max_level); } @testdirs = ( map { "$topdir/samples/$_" } @sampdirs, ); } $ntests = @testdirs; if ($nodelist) { # Run tests in parallel my @nodes=split /,/ , $nodelist; my $passthru_opts = ''; if ($clean) { $passthru_opts .= '-C '; } if ($makejobs) { $passthru_opts .= " -j $makejobs"; } if ($opts{'c'} || $opts{'compile-only'}) { $passthru_opts .= ' -c'; } do_tests_in_parallel(\@testdirs,\@nodes,$passthru_opts); } else { # Run tests consecutively for my $d (@testdirs) { test_rundir("$d",$makeopts); } } } # ---------------------------------------------------------------------- # sub get_all_sampdirs { # Return a flat list of all sample dirs that belong to a level <= $level . my $level = shift; my @dirs; my @levels = grep { $_ <= $level } keys %tests; foreach my $lev (sort {$a <=> $b} @levels) { push @dirs, @{$tests{$lev}}; } return @dirs; } # ---------------------------------------------------------------------- # sub get_sampdirs { # Return list of sample dirs that belong to $level . my $level = shift; my @dirs; my @levels = grep { $_ == $level } keys %tests; foreach my $lev (sort {$a <=> $b} @levels) { push @dirs, @{$tests{$lev}}; } return @dirs; } # ---------------------------------------------------------------------- # sub handle_lock_file { # Check for existing lock file and exit (if there is one) or write one. return(1) if ($nolock); # skip it my $got_lock = 0; # need to deserve this first.. if (-e $lockfile) { # lock file exists -> get and check pid open(LOCK,"< $lockfile") or die "Cannot read <$lockfile>"; my $pid = ; close LOCK; $pid =~ s/^\s*([0-9]+)\s*$/$1/ or die "$lockfile contains garbage instead of pid: <$pid>\n"; my $pid_cmd = `ps -p $pid -o args=`; # Who runs with pid from lockfile? if ($pid_cmd =~ /^\S*$/) { # no process for this pid print "Removing stale lock file (no process with pid $pid)\n"; unlink $lockfile or die "Cannot unlink <$lockfile>\n"; $got_lock = 1; } elsif ($pid_cmd =~ /auto-test\s*$/) { # pid belongs to auto-test print "Failed getting lock (lock file is $lockfile):\n" . "Another auto-test is running as pid $pid\n" . "You can use `auto-test --nolock' to ignore the lock file\n"; $got_lock = 0; } else { # pid belongs to other process print "Failed getting lock (lock file is <$lockfile>):\n" . "Another process runs as pid $pid (this is weird...)\n" . "You can use `auto-test --nolock' to ignore the lock file\n"; $got_lock = 0; } } else { print STDERR "handle_lock_file: Found no lock file\n" if ($debug); $got_lock = 1; } # Write our own lock file if ($got_lock) { print STDERR "Writing <$$> to lock file $lockfile\n" if ($debug); write_to_file($lockfile, "$$", 0, 1); $remove_lock = 1; # clean up lock at END } print STDERR "handle_lock_file: \$got_lock = $got_lock\n" if ($debug); unless ($got_lock) { $failed++; $failure_message .= "Couldn't get lock file\n"; } return ($got_lock); } # ---------------------------------------------------------------------- # sub do_tests_in_parallel { # Distribute auto-tests over nodes contained in $nodelist my $jobs = shift; my $nodelist = shift; my $passthru_opts = shift; my @queue=@$jobs; my %nodes; for my $node (@$nodelist) { $nodes{$node}{'busy'} = 0; $nodes{$node}{'pid'} = 0; } my $running = 0; my $queued = scalar @queue; while ((scalar @queue) || $running) { if (scalar @queue) { foreach my $node (keys %nodes) { next if $nodes{$node}{'busy'}; my $job = shift @queue; $nodes{$node}{'busy'}=$job; $running++; if (! ($nodes{$node}{'pid'} = fork)) { my $nodeb=$node; exec "$rsh $node auto-test --no-summary $passthru_opts $job | sed 's/^/$nodeb : /'"; # didn't get here } } } if ($running) { my $pid = wait; if ($pid != -1) { my $result = $?; $running--; foreach my $node (keys %nodes) { if ($nodes{$node}{'pid'} == $pid) { my $exit_code = ($result >> 8); $failed{$nodes{$node}{'busy'}} = $exit_status[$exit_code-1] if ($exit_code); $nodes{$node}{'busy'} = 0; $nodes{$node}{'pid'} = 0; last; } } } } } } # ---------------------------------------------------------------------- # sub my_ok { # Similar to Test's and Test::Harness' ok() function: consider success if # first two argumets are equal, otherwise report a problem and print the # third argument (which should normally contain the output from the shell # calls that we are normally testing here). # Args #4 and #5 are the current run directory and the phase (compilation, # starting, running) we are in. my $arg1 = shift; my $arg2 = shift; my $mesg = (shift || ""); chomp($mesg); my $dir = shift; my $phase = shift; my $quiet = (shift || 0); my $dt = shift; my $timestr; if (defined($dt) && $time) { $timestr = s_to_hms($dt,7); } else { $timestr = ''; } # Allow for calls like `ok(0)' or `ok(1)': if (!defined($arg2)) { $arg2 = $arg1 ? $arg1 : 1; } if ($arg1 eq $arg2) { print " ok $timestr\n" unless ($quiet); } else { print " not ok: $timestr\n$mesg\n"; $test_status = 1; # report only first failure: $failed{$dir} = $phase unless defined($failed{$dir}); } # Record time usage in an easily greppable format if ($summarize_t) { record_time($phase, $dt); } } # ---------------------------------------------------------------------- # sub record_time { # Store a time value so we can summarize times when done with this # directory my ($phase, $time) = @_; if (defined($time)) { $time_summary{$phase} = $time; } } # ---------------------------------------------------------------------- # sub find_test_dirs { # Traverse the samples/ directory tree and find all run directories eval { require File::Find }; if ($@) { die "Could not load File::Find (required for the --deep option):\n" . "$@\n"; } import File::Find qw( find ); my @rootdirs = @_; find( { wanted => \&FF_wanted, follow => 1, follow_skip => 2, }, @rootdirs); # $File::Find::prune = 1; } # ---------------------------------------------------------------------- # sub FF_wanted { # `wanted' function for File::Find, i.e. callback function that is called # from each file or directory found. my $name = $File::Find::name; # Suppress spurious `used only once' warning: my $dummy = "$File::Find::prune $File::Find::name"; if (-d $name) { if ($_ =~ /$prunedirs/) { # Don't recurse into well-known boring dirs: $File::Find::prune = 1; } else { # Directory, but not boring: if ( $list_all || -r "$name/reference.out" || -r "$name/reference.out.double" ) { push @testdirs, $name; } } } } # ---------------------------------------------------------------------- # sub test_shell_cmd { # As the name says... my $cmd = shift; my $dir = (shift || ""); my $phase = (shift || ""); my $quiet = (shift || 0); # flag for suppressing the `ok' my $t_ref = shift; print STDERR "\nRunning `$nice $cmd`" if ($debug); my $res = (`$nice $cmd 2>&1` || "Very BAD: `$cmd` returned undef"); my $cmdstatus = $?; print STDERR " ..done \n" if ($debug); chomp($res); if (($phase eq 'compilation') && ($postproc ne '')) { print STDERR "Postprocessing output..\n" if ($debug); $res = postprocess($res); } my $dt; $dt = get_time()-$t_ref if (defined($t_ref)); # otherwise leave undefined my_ok($cmdstatus, 0, $res, $dir, $phase, $quiet, $dt); # $? = 0 means success return $res; # Return output so we can analyze it if desired } # ---------------------------------------------------------------------- # sub test_timelimited_shell_cmd { # Test shell command, but run it through reaper to limit total time it is # allowed to take my $cmd = shift; my $maxtime = shift; my $procname = shift; my @rest = @_; return test_shell_cmd("reaper -t '$maxtime,$procname' '$cmd'", @rest); } # ---------------------------------------------------------------------- # sub test_compile { # Test compilation only if ($t_comp) { my $dir = shift; my @makeopts = (shift); # gets one string ("" or "-j N") my $t_ref = shift; # start time print " Compiling.. "; # First pc_setupsrc, then make clean (if requested). Must be done # in this order as pc_setupsrc links the Makefile to the run directory. test_shell_cmd("pc_setupsrc --pencil-home $pencil_home", $dir, "setting up directory", QUIET); return 'ERROR calling pc_setupsrc' if ($test_status); if ($clean) { test_shell_cmd('make cleann', $dir, "cleaning up", QUIET); return 'ERROR running make cleann' if ($test_status); } # Compile in double precision if we have data for that if (-r 'print.in.double' && -r 'reference.out.double') { push @makeopts, 'REAL_PRECISION=double'; $reference_out = 'reference.out.double'; printf "$indi_fmt ", '[double]'; } else { $reference_out = 'reference.out'; printf "$indi_fmt ", ''; } print STDERR "\@makeopts = @makeopts\n" if ($debug); test_shell_cmd("make @makeopts", $dir, "compilation", NOQUIET, $t_ref); } } # ---------------------------------------------------------------------- # sub test_start { # Test starting only my $dir = shift; # not used... my $t_ref = shift; # start time if ($t_start) { print " Starting.. "; printf "$indi_fmt ", ''; if ($time_limit) { test_timelimited_shell_cmd( "start.csh", $time_limit, 'start.x', $dir, "starting", NOQUIET, $t_ref ); } else { test_shell_cmd( "start.csh", $dir, "starting", NOQUIET, $t_ref ); } } } # ---------------------------------------------------------------------- # sub test_run { # Test running only (does not analyze output, but returns it) my $dir = shift; my $t_ref = shift; # start time if ($t_run) { print " Running.. "; # Handle KEEP_RELOADING file (our -r/--reload option) my $reload_file = "$dir/KEEP_RELOADING"; if (-e $reload_file) { warn "Found file $reload_file -- deleting\n"; unlink $reload_file or die "Cannot unlink $reload_file!\n"; } if ($reload) { write_to_file($reload_file, ''); printf "$indi_fmt ", '[KEEP_RELOADING]'; } else { printf "$indi_fmt ", ''; } # Run the Code if ($time_limit) { $res = test_timelimited_shell_cmd( "run.csh", $time_limit, 'run.x', $dir, "running", NOQUIET, $t_ref ); } else { $res = test_shell_cmd( "run.csh", $dir, "running", NOQUIET, $t_ref ); } # Clean up if (-e $reload_file) { unlink $reload_file or die "Cannot unlink $reload_file!\n"; } } } # ---------------------------------------------------------------------- # sub test_results { # Analyze results from code my $dir = shift; if ($t_res) { my ($rdmsg,$diagn); print " Validating results.. "; printf "$indi_fmt ", ''; my @output; if ($res_from_file) { @output = read_lines('data/time_series.dat', $rdmsg); if ($rdmsg) { my_ok(0,1,"[$rdmsg]",$dir,"results"); return; } } else { @output = split(/[\n\r]+/,$res); } my_ok(0,0,"",$dir,"results","quiet"); my @foutput = extract_results(\@output); # filtered output if ($foutput[0] ne "$noheader") { my @ref_output = read_lines($reference_out,$rdmsg); if ($rdmsg) { print " [$rdmsg]\n"; } else { my $comp = compare_results(\@ref_output,\@foutput,$diagn); my_ok($comp,1,$diagn,$dir,"results"); } } else { # No header line found warn " No header line found\n"; my_ok(0,1,join("\n",@output),$dir,"results"); } } } # ---------------------------------------------------------------------- # sub extract_results { # Extract result lines (header line and numerical data) from the output of # the code my $lineptr = shift; my @lines = @$lineptr; my @results = (); my @labels = (); my $body = 0; # true after header line my $ncols; line: foreach my $line (@lines) { unless ($body) { # Before or at header line next line unless ($line =~ /$headerpat/); push @results, $line; $body = 1; @labels = grep { /\S/ } split(/-+/,$line); # ignore empty labels $ncols = @labels; } else { # After header line next unless ($line =~ /^(\s+$ieee_float){$ncols}\s*$/i); push @results, $line; } } if ($body) { @results; } else { ($noheader); } } # ---------------------------------------------------------------------- # sub read_lines { # Read file an return hash of non-empty lines my $file = shift; my @lines = (); my $msg = ""; { local $/ = undef; # read in whole file if (open (REF, "< $file")) { @lines = grep { /\S/ } split(/[\n\r]+/,); # Remove leading comment sign from header line: $lines[0] =~ s/^(\s*)#/$1 /; } else { $msg = "Couldn't open $file"; } } $_[0] = $msg; @lines; } # ---------------------------------------------------------------------- # sub compare_results { # Compare two arrays of lines linewise; if all lines are the same, return # 1; if not, return 0 and write report to third argument my $arr1 = shift; my $arr2 = shift; my @pref = ("< ", "> ", " "); # prefixes to mark differing lines my $n1 = $#$arr1; my $n2 = $#$arr2; my $N = ($n2>$n1 ? $n2 : $n1); my $diagn = ""; my $equal = 1; for my $i (0..$N) { my $line1 = ($$arr1[$i] || ""); chomp $line1; my $line2 = ($$arr2[$i] || ""); chomp $line2; unless (compare_lines_fuzzily($line1,$line2)) { $diagn .= $pref[0] . $line1 . "\n" if ($line1); $diagn .= $pref[1] . $line2 . "\n" if ($line2); $equal = 0; } else { if ($i == 0) { # Keep header line for easier reading $diagn .= $pref[2] . $line1 . "\n"; } } } $_[0] = $diagn; # the remaining argument (was the third one) $equal; } # ---------------------------------------------------------------------- # sub compare_lines_fuzzily { # Compare the numerical values in two lines fuzzily. Return 1 if numbers # are approximately equal (differ by 1 or less in the last decimal), zero # otherwise. my $xline = shift; my $yline = shift; my $equal = 1; if ($xline =~ /$headerpat/) { # header lines --> string comparison $equal = 0 unless ($xline eq $yline); } else { # diagnostic lines --> fuzzy comparison my @x = split(/\s+/,$xline); my @y = split(/\s+/,$yline); @x = grep(/([0-9]|NaN|Inf)/i, @x); # weed out empty entries @y = grep(/([0-9]|NaN|Inf)/i, @y); return 0 unless ($#x == $#y); # lengths must match foreach my $i (0..$#x) { my ($x,$y) = ($x[$i], $y[$i]); $equal = 0 unless (compare_numbers_fuzzily($x,$y)); print STDERR "compare_lines_fuzzily: \$equal = $equal for <$x>, <$y>\n" if ($debug); } } return $equal; } # ---------------------------------------------------------------------- # sub compare_numbers_fuzzily { # Compare two numbers and return true if they differ only by one in the # last digit, else false. my $x = shift; my $y = shift; my $debug = 0; # Short-circuit NaN/Inf -- they will be taken care of as they will make # any comparison fail return 1 if ( ($x =~ /^([+-]?(NaN|Inf))$/i) || ($y =~ /^([+-]?(NaN|Inf))$/i) ); # Regexp for general float, groups into pre-exponent and # post-exponent part my $numeric = '((?:[+-]?)(?=\d|\.\d)\d*(?:\.\d*)?)((?:[EeDd](?:[+-]?\d+))?)'; my ($x1,$x2) = ($x =~ /^\s*$numeric\s*$/); $x2 = '' unless defined($x2); print STDERR "(\$x1,\$x2) = ($x1,$x2)\n" if ($debug); my ($y1,$y2) = ($y =~ /^\s*$numeric\s*$/); $y2 = '' unless defined($y2); print STDERR "(\$y1,\$y2) = ($y1,$y2)\n" if ($debug); # Are $x, $y really numeric? unless(defined($x1) && defined($y1)) { warn "Not a numerical value: <$x> or <$y>\n"; return 0; } # Short-circuit identical numbers (so Perl's rounding errors to double # precision can't screw them up) return 1 if ("$x1#$x2" eq "$y1#$y2"); # Choose the longer string for decrementing/incrementing, so we # correctly compare stuff like [1.3599, 1.36] (succeeds) and [1.3598, # 1.36] (fails) in both directions. if (length($x1) < length($y1)) { ($x,$x1,$x2, $y,$y1,$y2) = ($y,$y1,$y2, $x,$x1,$x2); } # Do comparison $x1 =~ /([+-]?[0-9]*)\.?([0-9]*)/; my ($xint,$xfrac) = ($1, $2); die "Cannot split <$x1> into int/fract parts (should not happen)" unless defined($xint); my $delta; if (length($xfrac) > 0) { $delta = '0.' . '0'x(length($xfrac)-1) . '1'; } else { $delta = 1; } my $x1_p = $x1 + $delta; my $x1_m = $x1 - $delta; ($x1_m, $x1_p) = sort { $a <=> $b } $x1_m, $x1_p; my ($x_m, $x_p) = ("${x1_m}${x2}", "${x1_p}${x2}"); # Are the constructed numbers $x_m, $x_p really numerical? eval { my $dummy = 1 + $x_m + $x_p }; if ($@) { warn "$@"; return 0; } printf STDERR "%8s -> %8s -> <%6s>.<%6s> -> %8s -> [%8s, %8s]\n", $x, $x1, $xint, $xfrac, $delta, $x_m, $x_p if ($debug); print STDERR "$y >= $x_m = ", ($y >= $x_m), "\n" if ($debug); print STDERR "$y <= $x_p = ", ($y <= $x_p), "\n" if ($debug); print STDERR "Ergo: ", (($y >= $x_m) && ($y <= $x_p)), "\n" if ($debug); if (($y >= $x_m) && ($y <= $x_p)) { return 1; } else { return 0; } } # ---------------------------------------------------------------------- # sub test_rundir { # Full program for one run directory: Test compilation, starting, running # and results my $dir = shift; my $makeopts = shift; my $t0 = get_time(); $test_status = 0; # so far, everything is OK %time_summary =(); # Indicate current run directory in process name (for ps) my $shortdir = $dir; $shortdir =~ s{^$topdir/}{}; # remove common directory prefix $shortdir =~ s{^samples/}{}; # remove leading 'samples/' $0 =~ s/^\s*(\[.*?\])?\s*/[$shortdir] /; # Go to directory and identify it if (! -d $dir) { print STDERR "No such directory: $dir\n"; return; } chdir $dir; my $cwd = cwd(); if ($short) { $cwd =~ s{.*pencil[^/]*/}{}i }; if ($list_only) { print "$cwd\n"; return; } print "\n$cwd:"; print " ($itest/$ntests)" if ($ntests>1); print "\n"; $itest++; # Make sure we have everything we need if (! defined(-e 'src/run.f90')) { # has `pc_setupsrc' been run yet? my $res = `pc_setupsrc --pencil-home $pencil_home 2>&1`; if ($?) { print " Problems running pc_setupsrc:\n", $res; } } # 1. Test compilation test_compile($dir,$makeopts,$t0); return if ($test_status); my $t1 = get_time(); $t_global{'compile'} += ($t1-$t0); # 2. Ensure we have a data directory and test starting... if ($t_start) { # no need for a data directory unless we start # the code $created_datadir = check_for_datadir(); test_start($dir,$t1); goto cleanup if ($test_status); } # 3 ..and running (we only get here if compiling and starting was OK) my $t2 = get_time(); test_run($dir,$t2); return if ($test_status); my $t3 = get_time(); $t_global{'start+run'} += ($t3-$t1); # 4. Check the output test_results($dir); # Clean up if necessary cleanup: if ($created_datadir) { system("rm", "data"); system("rm", "-r", "$tmpdir"); } # Summarize timings in human-readable form my $t4 = get_time(); if ($time) { print " Time used: ", s_to_hms(time_diff($t0,$t3), 44), " = ", s_to_hms(time_diff($t0,$t1)), " + ", s_to_hms(time_diff($t1,$t3)), "\n"; } # Write timings in greppable form if (%time_summary) { record_time('total', time_diff($t0,$t3)); for my $phase qw( compilation starting running total ) { write_timing_summary($dir, $phase, \%time_summary); } } } # ---------------------------------------------------------------------- # sub write_timing_summary_header { # Write comment to make timing summary files more comprehensible printf "%s %s\n", '#TIMING ', '# --------------------------------------------------'; printf "%s %s\n", '#TIMING ', '# Date: ' . localtime() . ", host: $host"; printf "%s %-45s %-12s %-12s %-7s\n", '#TIMING ', '# directory', 'phase', 'time-stamp', 'seconds'; } # ---------------------------------------------------------------------- # sub write_timing_summary { # Write timings summary in greppable form my ($dir, $phase, $summary_ref) = @_; my $time = $summary_ref->{$phase}; if (defined($phase) && defined($time)) { printf "%s %-45s %-12s %12.1f %7.3f\n", '#TIMING ', compactify(quote($dir)), ($phase), time_stamp(), $time; } } # ---------------------------------------------------------------------- # sub check_for_datadir { # Make sure we have an appropriate data directory my $created_datadir = 0; if (! -d 'data') { if ($mkdatadir) { if (-l 'data') { print " data is file or empty link;\n" . " removing and generating data directory \n"; system("rm", "data"); } else { print " No data directory; generating data directory\n"; } system("mkdir -p data"); } else { if (-l 'data') { print " data is file or empty link;\n" . " removing and generating data -> $tmpdir\n"; system("rm", "data"); } else { print " No data directory; generating data -> $tmpdir\n"; } system("mkdir -p $tmpdir; ln -s $tmpdir data"); $created_datadir = 1; } } return $created_datadir; } # ---------------------------------------------------------------------- # sub time_diff { # Return difference of times if both args are defined, undef otherwise my $t1 = shift; my $t2 = shift; if (defined($t1) && defined($t2)) { return $t2-$t1; } else { return undef; } } # ---------------------------------------------------------------------- # sub s_to_hms { # Convert no. of seconds to [ddd][hh:]mm:ss string my $secs = shift; my $width = (shift || 0); my $string; # Not much to do if arg is undef: if (! defined($secs)) { $string = 'undef'; } else { my $ss = $secs % 60; my $mm = floor($secs/60) % 60; my $hh = floor($secs/3600) % 24; my $dd = floor($secs/86400); $string = sprintf("%02d:%02d", $mm,$ss); if ($hh) { $string = sprintf("%02d:", $hh) . $string }; if ($dd) { $string = sprintf("%dd", $dd) . $string }; } if (length($string) < $width) { $string = (" " x ($width-length($string))) . $string; }; return $string; } # ---------------------------------------------------------------------- # sub quote { # Quote most non-word characters in a string, but not a foew common ones my ($text) = @_; $text = "\Q$text\E"; $text =~ s{\\(/|-|_)}{$1}g; # unquote some chars return $text; } # ---------------------------------------------------------------------- # sub compactify { # Remove uninteresting parts from a rund directory's file name my ($text) = @_; $text =~ s{/+\\\.$}{}; # drop trailing `/.' from path names $text =~ s|.*$ENV{PENCIL_HOME}/||; $text =~ s{.*/pencil-(code|auto-test)/}{}; return $text; } # ---------------------------------------------------------------------- # sub time_stamp { # Return a time stamp based on unix time (epoch 1970/1/1) return get_time() - get_unix_epoch_start(); } # ---------------------------------------------------------------------- # sub get_unix_epoch_start { # Get the time value for 1970/01/01 0:00 UTC -- even on operating systems # that use a different epoch. # On Unix, this function returns 0. my $t0; { require POSIX; my $old_TZ = $ENV{TZ}; $ENV{TZ} = 'UCT'; POSIX::tzset(); $t0 = POSIX::mktime(0, 0, 0, 1, 0, 70); if (defined $old_TZ) { $ENV{TZ} = $old_TZ; } else { delete $ENV{TZ}; } } return $t0; } # ---------------------------------------------------------------------- # sub get_time { # Wrapper around time() that uses Time::HiRes if possible my $time = 0; eval { require Time::HiRes; $time = Time::HiRes::time(); }; return $time || time(); } # ---------------------------------------------------------------------- # sub postprocess { # Clean compiler output with postprocessor my $text = shift; print STDERR " PP: Cleaning compiler output with <$postproc>\n" if ($debug); # open read/write pipe to postprocessor (see `man perlipc') my $pid = open2(\*POSTPROCD,\*UNPROCD, "$postproc 2>&1"); die "Couldn't start $postproc" unless defined($pid); # fork() a child to avoid deadlock if $text is large my $child_id; if ($child_id=fork()) { # parent --> read output close UNPROCD; # important local $/ = undef; # get it all at once $text = ; close POSTPROCD; # not necessary, I guess waitpid($child_id,0); # wait for child to finish to avoid zombie wait(); # bizarre enough there is another zombie.. } else { # child --> write to postproc die "Cannot fork: $!" unless defined($child_id); $nosummary = 1; # don't waffle when exiting close POSTPROCD; # apparently not necessary print UNPROCD $text; close UNPROCD; exit 1; # irrelevant } print STDERR " PP: done postprocessing\n" if ($debug); $text; } # ---------------------------------------------------------------------- # sub write_to_file { # Write TEXT to file FILE, aborting if file exists (unless OVERWRITE is # set), and write a helpful header (unless NOHEADER is set) my $file = shift; my $text = shift || ''; my $overwrite = shift || 0; my $noheader = shift || 0; my $header = !$noheader; # Avoid iterated negation (=spinning head) if (-e $file && !$overwrite) { die "File $file exists!\n"; } else { open(FILE,"> $file") or die "Cannot open file $file for writing\n"; if ($header) { my $shortfile = $file; $shortfile =~ s{/\./}{/}g; # remove `./' path components $shortfile =~ s{^.*?([^/]*/?[^/]*)$}{$1}g; # keep only last path component print FILE "$shortfile\n"; print FILE "-" x length($shortfile), "\n"; print FILE "Created: ", scalar localtime(), "\n\n"; print FILE "This file was automatically generated by auto-test," . " so think twice before\nyou modify it.\n\n"; } chomp($text); print FILE "$text\n" if (length($text)); close FILE or die "Cannot close file $file\n"; } } # ---------------------------------------------------------------------- # sub get_maintainer_list { # Return comma-separated list of maintaners of directories listed in # %failed, avoiding duplicate entries. my %maintainers; foreach my $dir (keys %failed) { foreach my $maint (get_maintainers("$dir/README")) { $maintainers{$maint}++; } } return join(',', grep /./, keys %maintainers); } # ---------------------------------------------------------------------- # sub get_maintainers { # Given a run directory, return list of maintainers extracted from the # README file. my $readme = shift; unless (-r $readme) { warn "Cannot open $readme\n"; return (); } my @emails = (); my $reader = Pencil::ReadmeReader->new($readme); foreach my $maint ($reader->get_content('Maintainer')) { $maint =~ s{(\[at\]|\@)}{/}; # obfuscate `@' (may end up on the web) $maint =~ s{(\[dot\]|\.)}{:}g; # obfuscate `.' $maint =~ s{^\s*(.*?)\s*$}{$1}; # strip surrounding whitespace if ($maint =~ /<([^>]*)>/) { $maint = $1; } if ($maint =~ m{[-.:a-zA-Z_0-9]+/[-:a-zA-Z_0-9]+}) { push @emails, $maint; } } return @emails; } # ---------------------------------------------------------------------- # ## Summarize results END { unless ($help || $nosummary || $list_only) { print "\n" . "-" x 70 . "\n"; # Print failure header that can be identified by pencil-test if ($failed || %failed) { print "### auto-test failed ###\n"; } # Failed outside individual tests (e.g. other auto-test is running) if ($failed) { print "$failure_message"; } # Failed during some of the individual tests if (%failed) { print "Failed ", scalar(keys %failed), " test(s) out of $ntests:\n"; while (my ($dir,$phase) = each %failed) { print " $dir ($phase)\n"; } } else { if ($ntests == 1) { print "Test succeeded.\n"; } elsif ($ntests > 1) { print "All $ntests tests succeeded.\n"; # pc_update_validated system("$ENV{PENCIL_HOME}/bin/pc_update_validated.sh &"); } elsif (($ntests < 1) && ! $failed) { print "There was no test to run???\n"; } } # Print timing numbers if ($time) { my @t = times(); print "\nCPU time (including compilation): ", s_to_hms($t[2], 7) . "u ", s_to_hms($t[3] ) . "s\n"; my $t1_global = time(); # end time print "Total wall-clock time: ", s_to_hms(time_diff($t0_global,$t1_global), 7), " = ", s_to_hms($t_global{'compile'}), " + ", s_to_hms($t_global{'start+run'}), "\n"; } # Print info from file if (-r $infofile) { print "\n------------ Other info: ------------\n"; open(INFO,"< $infofile") or warn "Couldn't open $infofile\n"; print while (); close(INFO); } # List maintainers of failed tests if ($list_maint && %failed) { print 'Maintainers of failed tests: ', get_maintainer_list(), "\n"; } } # Clean up generated directories if ($created_datadir) { if (-d $tmpdir) { system("rm", "-r", "$tmpdir"); } } # Remove lock file if ($remove_lock) { unlink $lockfile or warn "Couldn't remove lockfile <$lockfile>\n"; } } # End of file auto-test