# # t/test.pl - most of Test::More functionality without the fuss # NOTE: # # Do not rely on features found only in more modern Perls here, as some CPAN # distributions copy this file and must operate on older Perls. Similarly, keep # things, simple as this may be run under fairly broken circumstances. For # example, increment ($x++) has a certain amount of cleverness for things like # # $x = 'zz'; # $x++; # $x eq 'aaa'; # # This stands more chance of breaking than just a simple # # $x = $x + 1 # # In this file, we use the latter "Baby Perl" approach, and increment # will be worked over by t/op/inc.t $| = 1; our $Level = 1; my $test = 1; my $planned; my $noplan; my $Perl; # Safer version of $^X set by which_perl() # This defines ASCII/UTF-8 vs EBCDIC/UTF-EBCDIC $::IS_ASCII = ord 'A' == 65; $::IS_EBCDIC = ord 'A' == 193; # This is 'our' to enable harness to account for TODO-ed tests in # overall grade of PASS or FAIL our $TODO = 0; our $NO_ENDING = 0; our $Tests_Are_Passing = 1; # Use this instead of print to avoid interference while testing globals. sub _print { local($\, $", $,) = (undef, ' ', ''); print STDOUT @_; } sub _print_stderr { local($\, $", $,) = (undef, ' ', ''); print STDERR @_; } sub plan { my $n; if (@_ == 1) { $n = shift; if ($n eq 'no_plan') { undef $n; $noplan = 1; } } else { my %plan = @_; $plan{skip_all} and skip_all($plan{skip_all}); $n = $plan{tests}; } _print "1..$n\n" unless $noplan; $planned = $n; } # Set the plan at the end. See Test::More::done_testing. sub done_testing { my $n = $test - 1; $n = shift if @_; _print "1..$n\n"; $planned = $n; } END { my $ran = $test - 1; if (!$NO_ENDING) { if (defined $planned && $planned != $ran) { _print_stderr "# Looks like you planned $planned tests but ran $ran.\n"; } elsif ($noplan) { _print "1..$ran\n"; } } } sub _diag { return unless @_; my @mess = _comment(@_); $TODO ? _print(@mess) : _print_stderr(@mess); } # Use this instead of "print STDERR" when outputting failure diagnostic # messages sub diag { _diag(@_); } # Use this instead of "print" when outputting informational messages sub note { return unless @_; _print( _comment(@_) ); } sub is_miniperl { return !defined &DynaLoader::boot_DynaLoader; } sub set_up_inc { # Don’t clobber @INC under miniperl @INC = () unless is_miniperl; unshift @INC, @_; } sub _comment { return map { /^#/ ? "$_\n" : "# $_\n" } map { split /\n/ } @_; } sub _have_dynamic_extension { my $extension = shift; unless (eval {require Config; 1}) { warn "test.pl had problems loading Config: $@"; return 1; } $extension =~ s!::!/!g; return 1 if ($Config::Config{extensions} =~ /\b$extension\b/); } sub skip_all { if (@_) { _print "1..0 # Skip @_\n"; } else { _print "1..0\n"; } exit(0); } sub skip_all_if_miniperl { skip_all(@_) if is_miniperl(); } sub skip_all_without_dynamic_extension { my ($extension) = @_; skip_all("no dynamic loading on miniperl, no $extension") if is_miniperl(); return if &_have_dynamic_extension; skip_all("$extension was not built"); } sub skip_all_without_perlio { skip_all('no PerlIO') unless PerlIO::Layer->find('perlio'); } sub skip_all_without_config { unless (eval {require Config; 1}) { warn "test.pl had problems loading Config: $@"; return; } foreach (@_) { next if $Config::Config{$_}; my $key = $_; # Need to copy, before trying to modify. $key =~ s/^use//; $key =~ s/^d_//; skip_all("no $key"); } } sub skip_all_without_unicode_tables { # (but only under miniperl) if (is_miniperl()) { skip_all_if_miniperl("Unicode tables not built yet") unless eval 'require "unicore/UCD.pl"'; } } sub find_git_or_skip { my ($source_dir, $reason); if ( $ENV{CONTINUOUS_INTEGRATION} && $ENV{WORKSPACE} ) { $source_dir = $ENV{WORKSPACE}; if ( -d "${source_dir}/.git" ) { $ENV{GIT_DIR} = "${source_dir}/.git"; return $source_dir; } } if (-d '.git') { $source_dir = '.'; } elsif (-l 'MANIFEST' && -l 'AUTHORS') { my $where = readlink 'MANIFEST'; die "Can't readlink MANIFEST: $!" unless defined $where; die "Confusing symlink target for MANIFEST, '$where'" unless $where =~ s!/MANIFEST\z!!; if (-d "$where/.git") { # Looks like we are in a symlink tree if (exists $ENV{GIT_DIR}) { diag("Found source tree at $where, but \$ENV{GIT_DIR} is $ENV{GIT_DIR}. Not changing it"); } else { note("Found source tree at $where, setting \$ENV{GIT_DIR}"); $ENV{GIT_DIR} = "$where/.git"; } $source_dir = $where; } } elsif (exists $ENV{GIT_DIR} || -f '.git') { my $commit = '8d063cd8450e59ea1c611a2f4f5a21059a2804f1'; my $out = `git rev-parse --verify --quiet '$commit^{commit}'`; chomp $out; if($out eq $commit) { $source_dir = '.' } } if ($ENV{'PERL_BUILD_PACKAGING'}) { $reason = 'PERL_BUILD_PACKAGING is set'; } elsif ($source_dir) { my $version_string = `git --version`; if (defined $version_string && $version_string =~ /\Agit version (\d+\.\d+\.\d+)(.*)/) { return $source_dir if eval "v$1 ge v1.5.0"; # If you have earlier than 1.5.0 and it works, change this test $reason = "in git checkout, but git version '$1$2' too old"; } else { $reason = "in git checkout, but cannot run git"; } } else { $reason = 'not being run from a git checkout'; } skip_all($reason) if $_[0] && $_[0] eq 'all'; skip($reason, @_); } sub BAIL_OUT { my ($reason) = @_; _print("Bail out! $reason\n"); exit 255; } sub _ok { my ($pass, $where, $name, @mess) = @_; # Do not try to microoptimize by factoring out the "not ". # VMS will avenge. my $out; if ($name) { # escape out '#' or it will interfere with '# skip' and such $name =~ s/#/\\#/g; $out = $pass ? "ok $test - $name" : "not ok $test - $name"; } else { $out = $pass ? "ok $test" : "not ok $test"; } if ($TODO) { $out = $out . " # TODO $TODO"; } else { $Tests_Are_Passing = 0 unless $pass; } _print "$out\n"; if ($pass) { note @mess; # Ensure that the message is properly escaped. } else { my $msg = "# Failed test $test - "; $msg.= "$name " if $name; $msg .= "$where\n"; _diag $msg; _diag @mess; } $test = $test + 1; # don't use ++ return $pass; } sub _where { my @caller = caller($Level); return "at $caller[1] line $caller[2]"; } # DON'T use this for matches. Use like() instead. sub ok ($@) { my ($pass, $name, @mess) = @_; _ok($pass, _where(), $name, @mess); } sub _q { my $x = shift; return 'undef' unless defined $x; my $q = $x; $q =~ s/\\/\\\\/g; $q =~ s/'/\\'/g; return "'$q'"; } sub _qq { my $x = shift; return defined $x ? '"' . display ($x) . '"' : 'undef'; }; # Support pre-5.10 Perls, for the benefit of CPAN dists that copy this file. # Note that chr(90) exists in both ASCII ("Z") and EBCDIC ("!"). my $chars_template = defined(eval { pack "W*", 90 }) ? "W*" : "U*"; eval 'sub re::is_regexp { ref($_[0]) eq "Regexp" }' if !defined &re::is_regexp; # keys are the codes \n etc map to, values are 2 char strings such as \n my %backslash_escape; foreach my $x (split //, 'enrtfa\\\'"') { $backslash_escape{ord eval "\"\\$x\""} = "\\$x"; } # A way to display scalars containing control characters and Unicode. # Trying to avoid setting $_, or relying on local $_ to work. sub display { my @result; foreach my $x (@_) { if (defined $x and not ref $x) { my $y = ''; foreach my $c (unpack($chars_template, $x)) { if ($c > 255) { $y = $y . sprintf "\\x{%x}", $c; } elsif ($backslash_escape{$c}) { $y = $y . $backslash_escape{$c}; } elsif ($c < ord " ") { # Use octal for characters with small ordinals that are # traditionally expressed as octal: the controls below # space, which on EBCDIC are almost all the controls, but # on ASCII don't include DEL nor the C1 controls. $y = $y . sprintf "\\%03o", $c; } elsif (chr $c =~ /[[:print:]]/a) { $y = $y . chr $c; } else { $y = $y . sprintf "\\x%02X", $c; } } $x = $y; } return $x unless wantarray; push @result, $x; } return @result; } sub is ($$@) { my ($got, $expected, $name, @mess) = @_; my $pass; if( !defined $got || !defined $expected ) { # undef only matches undef $pass = !defined $got && !defined $expected; } else { $pass = $got eq $expected; } unless ($pass) { unshift(@mess, "# got "._qq($got)."\n", "# expected "._qq($expected)."\n"); } _ok($pass, _where(), $name, @mess); } sub isnt ($$@) { my ($got, $isnt, $name, @mess) = @_; my $pass; if( !defined $got || !defined $isnt ) { # undef only matches undef $pass = defined $got || defined $isnt; } else { $pass = $got ne $isnt; } unless( $pass ) { unshift(@mess, "# it should not be "._qq($got)."\n", "# but it is.\n"); } _ok($pass, _where(), $name, @mess); } sub cmp_ok ($$$@) { my($got, $type, $expected, $name, @mess) = @_; my $pass; { local $^W = 0; local($@,$!); # don't interfere with $@ # eval() sometimes resets $! $pass = eval "\$got $type \$expected"; } unless ($pass) { # It seems Irix long doubles can have 2147483648 and 2147483648 # that stringify to the same thing but are actually numerically # different. Display the numbers if $type isn't a string operator, # and the numbers are stringwise the same. # (all string operators have alphabetic names, so tr/a-z// is true) # This will also show numbers for some unneeded cases, but will # definitely be helpful for things such as == and <= that fail if ($got eq $expected and $type !~ tr/a-z//) { unshift @mess, "# $got - $expected = " . ($got - $expected) . "\n"; } unshift(@mess, "# got "._qq($got)."\n", "# expected $type "._qq($expected)."\n"); } _ok($pass, _where(), $name, @mess); } # Check that $got is within $range of $expected # if $range is 0, then check it's exact # else if $expected is 0, then $range is an absolute value # otherwise $range is a fractional error. # Here $range must be numeric, >= 0 # Non numeric ranges might be a useful future extension. (eg %) sub within ($$$@) { my ($got, $expected, $range, $name, @mess) = @_; my $pass; if (!defined $got or !defined $expected or !defined $range) { # This is a fail, but doesn't need extra diagnostics } elsif ($got !~ tr/0-9// or $expected !~ tr/0-9// or $range !~ tr/0-9//) { # This is a fail unshift @mess, "# got, expected and range must be numeric\n"; } elsif ($range < 0) { # This is also a fail unshift @mess, "# range must not be negative\n"; } elsif ($range == 0) { # Within 0 is == $pass = $got == $expected; } elsif ($expected == 0) { # If expected is 0, treat range as absolute $pass = ($got <= $range) && ($got >= - $range); } else { my $diff = $got - $expected; $pass = abs ($diff / $expected) < $range; } unless ($pass) { if ($got eq $expected) { unshift @mess, "# $got - $expected = " . ($got - $expected) . "\n"; } unshift@mess, "# got "._qq($got)."\n", "# expected "._qq($expected)." (within "._qq($range).")\n"; } _ok($pass, _where(), $name, @mess); } # Note: this isn't quite as fancy as Test::More::like(). sub like ($$@) { like_yn (0,@_) }; # 0 for - sub unlike ($$@) { like_yn (1,@_) }; # 1 for un- sub like_yn ($$$@) { my ($flip, undef, $expected, $name, @mess) = @_; # We just accept like(..., qr/.../), not like(..., '...'), and # definitely not like(..., '/.../') like # Test::Builder::maybe_regex() does. unless (re::is_regexp($expected)) { die "PANIC: The value '$expected' isn't a regexp. The like() function needs a qr// pattern, not a string"; } my $pass; $pass = $_[1] =~ /$expected/ if !$flip; $pass = $_[1] !~ /$expected/ if $flip; my $display_got = $_[1]; $display_got = display($display_got); my $display_expected = $expected; $display_expected = display($display_expected); unless ($pass) { unshift(@mess, "# got '$display_got'\n", $flip ? "# expected !~ /$display_expected/\n" : "# expected /$display_expected/\n"); } local $Level = $Level + 1; _ok($pass, _where(), $name, @mess); } sub refcount_is { # Don't unpack first arg; access it directly via $_[0] to avoid creating # another reference and upsetting the refcount my (undef, $expected, $name, @mess) = @_; my $got = &Internals::SvREFCNT($_[0]) + 1; # +1 to account for the & calling style my $pass = $got == $expected; unless ($pass) { unshift @mess, "# got $got references\n" . "# expected $expected\n"; } _ok($pass, _where(), $name, @mess); } sub pass { _ok(1, '', @_); } sub fail { _ok(0, _where(), @_); } sub curr_test { $test = shift if @_; return $test; } sub next_test { my $retval = $test; $test = $test + 1; # don't use ++ $retval; } # Note: can't pass multipart messages since we try to # be compatible with Test::More::skip(). sub skip { my $why = shift; my $n = @_ ? shift : 1; my $bad_swap; my $both_zero; { local $^W = 0; $bad_swap = $why > 0 && $n == 0; $both_zero = $why == 0 && $n == 0; } if ($bad_swap || $both_zero || @_) { my $arg = "'$why', '$n'"; if (@_) { $arg .= join(", ", '', map { qq['$_'] } @_); } die qq[$0: expected skip(why, count), got skip($arg)\n]; } for (1..$n) { _print "ok $test # skip $why\n"; $test = $test + 1; } local $^W = 0; last SKIP; } sub skip_if_miniperl { skip(@_) if is_miniperl(); } sub skip_without_dynamic_extension { my $extension = shift; skip("no dynamic loading on miniperl, no extension $extension", @_) if is_miniperl(); return if &_have_dynamic_extension($extension); skip("extension $extension was not built", @_); } sub todo_skip { my $why = shift; my $n = @_ ? shift : 1; for (1..$n) { _print "not ok $test # TODO & SKIP $why\n"; $test = $test + 1; } local $^W = 0; last TODO; } sub eq_array { my ($ra, $rb) = @_; return 0 unless $#$ra == $#$rb; for my $i (0..$#$ra) { next if !defined $ra->[$i] && !defined $rb->[$i]; return 0 if !defined $ra->[$i]; return 0 if !defined $rb->[$i]; return 0 unless $ra->[$i] eq $rb->[$i]; } return 1; } sub eq_hash { my ($orig, $suspect) = @_; my $fail; while (my ($key, $value) = each %$suspect) { # Force a hash recompute if this perl's internals can cache the hash key. $key = "" . $key; if (exists $orig->{$key}) { if ( defined $orig->{$key} != defined $value || (defined $value && $orig->{$key} ne $value) ) { _print "# key ", _qq($key), " was ", _qq($orig->{$key}), " now ", _qq($value), "\n"; $fail = 1; } } else { _print "# key ", _qq($key), " is ", _qq($value), ", not in original.\n"; $fail = 1; } } foreach (keys %$orig) { # Force a hash recompute if this perl's internals can cache the hash key. $_ = "" . $_; next if (exists $suspect->{$_}); _print "# key ", _qq($_), " was ", _qq($orig->{$_}), " now missing.\n"; $fail = 1; } !$fail; } # We only provide a subset of the Test::More functionality. sub require_ok ($) { my ($require) = @_; if ($require =~ tr/[A-Za-z0-9:.]//c) { fail("Invalid character in \"$require\", passed to require_ok"); } else { eval < [ command-line switches ] # nolib => 1 # don't use -I../lib (included by default) # non_portable => Don't warn if a one liner contains quotes # prog => one-liner (avoid quotes) # progs => [ multi-liner (avoid quotes) ] # progfile => perl script # stdin => string to feed the stdin (or undef to redirect from /dev/null) # stderr => If 'devnull' suppresses stderr, if other TRUE value redirect # stderr to stdout # args => [ command-line arguments to the perl program ] # verbose => print the command line my $is_mswin = $^O eq 'MSWin32'; my $is_vms = $^O eq 'VMS'; my $is_cygwin = $^O eq 'cygwin'; sub _quote_args { my ($runperl, $args) = @_; foreach (@$args) { # In VMS protect with doublequotes because otherwise # DCL will lowercase -- unless already doublequoted. $_ = q(").$_.q(") if $is_vms && !/^\"/ && length($_) > 0; $runperl = $runperl . ' ' . $_; } return $runperl; } sub _create_runperl { # Create the string to qx in runperl(). my %args = @_; my $runperl = which_perl(); if ($runperl =~ m/\s/) { $runperl = qq{"$runperl"}; } #- this allows, for example, to set PERL_RUNPERL_DEBUG=/usr/bin/valgrind if ($ENV{PERL_RUNPERL_DEBUG}) { $runperl = "$ENV{PERL_RUNPERL_DEBUG} $runperl"; } unless ($args{nolib}) { $runperl = $runperl . ' "-I../lib" "-I." '; # doublequotes because of VMS } if ($args{switches}) { local $Level = 2; die "test.pl:runperl(): 'switches' must be an ARRAYREF " . _where() unless ref $args{switches} eq "ARRAY"; $runperl = _quote_args($runperl, $args{switches}); } if (defined $args{prog}) { die "test.pl:runperl(): both 'prog' and 'progs' cannot be used " . _where() if defined $args{progs}; $args{progs} = [split /\n/, $args{prog}, -1] } if (defined $args{progs}) { die "test.pl:runperl(): 'progs' must be an ARRAYREF " . _where() unless ref $args{progs} eq "ARRAY"; foreach my $prog (@{$args{progs}}) { if (!$args{non_portable}) { if ($prog =~ tr/'"//) { warn "quotes in prog >>$prog<< are not portable"; } if ($prog =~ /^([<>|]|2>)/) { warn "Initial $1 in prog >>$prog<< is not portable"; } if ($prog =~ /&\z/) { warn "Trailing & in prog >>$prog<< is not portable"; } } if ($is_mswin || $is_vms) { $runperl = $runperl . qq ( -e "$prog" ); } else { $runperl = $runperl . qq ( -e '$prog' ); } } } elsif (defined $args{progfile}) { $runperl = $runperl . qq( "$args{progfile}"); } else { # You probably didn't want to be sucking in from the upstream stdin die "test.pl:runperl(): none of prog, progs, progfile, args, " . " switches or stdin specified" unless defined $args{args} or defined $args{switches} or defined $args{stdin}; } if (defined $args{stdin}) { # so we don't try to put literal newlines and crs onto the # command line. $args{stdin} =~ s/\n/\\n/g; $args{stdin} =~ s/\r/\\r/g; if ($is_mswin || $is_vms) { $runperl = qq{$Perl -e "print qq(} . $args{stdin} . q{)" | } . $runperl; } else { $runperl = qq{$Perl -e 'print qq(} . $args{stdin} . q{)' | } . $runperl; } } elsif (exists $args{stdin}) { # Using the pipe construction above can cause fun on systems which use # ksh as /bin/sh, as ksh does pipes differently (with one less process) # With sh, for the command line 'perl -e 'print qq()' | perl -e ...' # the sh process forks two children, which use exec to start the two # perl processes. The parent shell process persists for the duration of # the pipeline, and the second perl process starts with no children. # With ksh (and zsh), the shell saves a process by forking a child for # just the first perl process, and execing itself to start the second. # This means that the second perl process starts with one child which # it didn't create. This causes "fun" when if the tests assume that # wait (or waitpid) will only return information about processes # started within the test. # They also cause fun on VMS, where the pipe implementation returns # the exit code of the process at the front of the pipeline, not the # end. This messes up any test using OPTION FATAL. # Hence it's useful to have a way to make STDIN be at eof without # needing a pipeline, so that the fork tests have a sane environment # without these surprises. # /dev/null appears to be surprisingly portable. $runperl = $runperl . ($is_mswin ? ' nul' : ' 2>/dev/null'); } elsif ($args{stderr}) { $runperl = $runperl . ' 2>&1'; } if ($args{verbose}) { my $runperldisplay = $runperl; $runperldisplay =~ s/\n/\n\#/g; _print_stderr "# $runperldisplay\n"; } return $runperl; } # usage: # $ENV{PATH} =~ /(.*)/s; # local $ENV{PATH} = untaint_path($1); sub untaint_path { my $path = shift; my $sep; if (! eval {require Config; 1}) { warn "test.pl had problems loading Config: $@"; $sep = ':'; } else { $sep = $Config::Config{path_sep}; } $path = join $sep, grep { $_ ne "" and $_ ne "." and -d $_ and ($is_mswin or $is_vms or !(stat && (stat _)[2]&0022)) } split quotemeta ($sep), $1; if ($is_cygwin) { # Must have /bin under Cygwin if (length $path) { $path = $path . $sep; } $path = $path . '/bin'; } elsif (!$is_vms and !length $path) { # empty PATH is the same as a path of "." on *nix so to prevent # tests from dieing under taint we need to return something # absolute. Perhaps "/" would be better? Anything absolute will do. $path = "/usr/bin"; } $path; } # sub run_perl {} is alias to below # Since this uses backticks to run, it is subject to the rules of the shell. # Locale settings may pose a problem, depending on the program being run. sub runperl { die "test.pl:runperl() does not take a hashref" if ref $_[0] and ref $_[0] eq 'HASH'; my $runperl = &_create_runperl; my $result; my $tainted = ${^TAINT}; my %args = @_; exists $args{switches} && grep m/^-T$/, @{$args{switches}} and $tainted = $tainted + 1; if ($tainted) { # We will assume that if you're running under -T, you really mean to # run a fresh perl, so we'll brute force launder everything for you my @keys = grep {exists $ENV{$_}} qw(CDPATH IFS ENV BASH_ENV); local @ENV{@keys} = (); # Untaint, plus take out . and empty string: local $ENV{'DCL$PATH'} = $1 if $is_vms && exists($ENV{'DCL$PATH'}) && ($ENV{'DCL$PATH'} =~ /(.*)/s); $ENV{PATH} =~ /(.*)/s; local $ENV{PATH} = untaint_path($1); $runperl =~ /(.*)/s; $runperl = $1; $result = `$runperl`; } else { $result = `$runperl`; } $result =~ s/\n\n/\n/g if $is_vms; # XXX pipes sometimes double these return $result; } # Nice alias *run_perl = *run_perl = \&runperl; # shut up "used only once" warning # Run perl with specified environment and arguments, return (STDOUT, STDERR) # set DEBUG_RUNENV=1 in the environment to debug. sub runperl_and_capture { my ($env, $args) = @_; my $STDOUT = tempfile(); my $STDERR = tempfile(); my $PERL = $^X; my $FAILURE_CODE = 119; local %ENV = %ENV; delete $ENV{PERLLIB}; delete $ENV{PERL5LIB}; delete $ENV{PERL5OPT}; delete $ENV{PERL_USE_UNSAFE_INC}; my $pid = fork; return (0, "Couldn't fork: $!") unless defined $pid; # failure if ($pid) { # parent waitpid $pid,0; my $exit_code = $? ? $? >> 8 : 0; my ($out, $err)= ("", ""); local $/; if (open my $stdout, '<', $STDOUT) { $out .= <$stdout>; } else { $err .= "Could not read STDOUT '$STDOUT' file: $!\n"; } if (open my $stderr, '<', $STDERR) { $err .= <$stderr>; } else { $err .= "Could not read STDERR '$STDERR' file: $!\n"; } if ($exit_code == $FAILURE_CODE) { $err .= "Something went wrong. Received FAILURE_CODE as exit code.\n"; } if ($ENV{DEBUG_RUNENV}) { print "OUT: $out\n"; print "ERR: $err\n"; } return ($out, $err); } elsif (defined $pid) { # child # Just in case the order we update the environment changes how # the environment is set up we sort the keys here for consistency. for my $k (sort keys %$env) { $ENV{$k} = $env->{$k}; } if ($ENV{DEBUG_RUNENV}) { print "Child Process $$ Executing:\n$PERL @$args\n"; } open STDOUT, '>', $STDOUT or do { print "Failed to dup STDOUT to '$STDOUT': $!"; exit $FAILURE_CODE; }; open STDERR, '>', $STDERR or do { print "Failed to dup STDERR to '$STDERR': $!"; exit $FAILURE_CODE; }; exec $PERL, @$args or print STDERR "Failed to exec: ", join(" ",map { "'$_'" } $^X, @$args), ": $!\n"; exit $FAILURE_CODE; } } sub DIE { _print_stderr "# @_\n"; exit 1; } # A somewhat safer version of the sometimes wrong $^X. sub which_perl { unless (defined $Perl) { $Perl = $^X; # VMS should have 'perl' aliased properly return $Perl if $is_vms; my $exe; if (! eval {require Config; 1}) { warn "test.pl had problems loading Config: $@"; $exe = ''; } else { $exe = $Config::Config{_exe}; } $exe = '' unless defined $exe; # This doesn't absolutize the path: beware of future chdirs(). # We could do File::Spec->abs2rel() but that does getcwd()s, # which is a bit heavyweight to do here. if ($Perl =~ /^perl\Q$exe\E$/i) { my $perl = "perl$exe"; if (! eval {require File::Spec; 1}) { warn "test.pl had problems loading File::Spec: $@"; $Perl = "./$perl"; } else { $Perl = File::Spec->catfile(File::Spec->curdir(), $perl); } } # Build up the name of the executable file from the name of # the command. if ($Perl !~ /\Q$exe\E$/i) { $Perl = $Perl . $exe; } warn "which_perl: cannot find $Perl from $^X" unless -f $Perl; # For subcommands to use. $ENV{PERLEXE} = $Perl; } return $Perl; } sub unlink_all { my $count = 0; foreach my $file (@_) { 1 while unlink $file; if( -f $file ){ _print_stderr "# Couldn't unlink '$file': $!\n"; }else{ $count = $count + 1; # don't use ++ } } $count; } # _num_to_alpha - Returns a string of letters representing a positive integer. # Arguments : # number to convert # maximum number of letters # returns undef if the number is negative # returns undef if the number of letters is greater than the maximum wanted # _num_to_alpha( 0) eq 'A'; # _num_to_alpha( 1) eq 'B'; # _num_to_alpha(25) eq 'Z'; # _num_to_alpha(26) eq 'AA'; # _num_to_alpha(27) eq 'AB'; my @letters = qw(A B C D E F G H I J K L M N O P Q R S T U V W X Y Z); # Avoid ++ -- ranges split negative numbers sub _num_to_alpha { my($num,$max_char) = @_; return unless $num >= 0; my $alpha = ''; my $char_count = 0; $max_char = 0 if !defined($max_char) or $max_char < 0; while( 1 ){ $alpha = $letters[ $num % 26 ] . $alpha; $num = int( $num / 26 ); last if $num == 0; $num = $num - 1; # char limit next unless $max_char; $char_count = $char_count + 1; return if $char_count == $max_char; } return $alpha; } my %tmpfiles; sub unlink_tempfiles { unlink_all keys %tmpfiles; %tempfiles = (); } END { unlink_tempfiles(); } # NOTE: tempfile() may be used as a module names in our tests # so the result must be restricted to only legal characters for a module # name. # A regexp that matches the tempfile names $::tempfile_regexp = 'tmp_[A-Z]+_[A-Z]+'; # Avoid ++, avoid ranges, avoid split // my $tempfile_count = 0; my $max_file_chars = 3; sub tempfile { # if you change the format returned by tempfile() you MUST change # the $::tempfile_regex define above. my $try_prefix = (-d "t" ? "t/" : "")."tmp_"._num_to_alpha($$); while (1) { my $alpha = _num_to_alpha($tempfile_count,$max_file_chars); last unless defined $alpha; my $try = $try_prefix . "_" . $alpha; $tempfile_count = $tempfile_count + 1; # Need to note all the file names we allocated, as a second request # may come before the first is created. Also we are avoiding ++ here # so we aren't using the normal idiom for this kind of test. if (!$tmpfiles{$try} && !-e $try) { # We have a winner $tmpfiles{$try} = 1; return $try; } } die sprintf 'panic: Too many tempfile()s with prefix "%s", limit of %d reached', $try_prefix, 26 ** $max_file_chars; } # register_tempfile - Adds a list of files to be removed at the end of the current test file # Arguments : # a list of files to be removed later # returns a count of how many file names were actually added # Reuses %tmpfiles so that tempfile() will also skip any files added here # even if the file doesn't exist yet. sub register_tempfile { my $count = 0; for( @_ ){ if( $tmpfiles{$_} ){ _print_stderr "# Temporary file '$_' already added\n"; }else{ $tmpfiles{$_} = 1; $count = $count + 1; } } return $count; } # This is the temporary file for fresh_perl my $tmpfile = tempfile(); sub fresh_perl { my($prog, $runperl_args) = @_; # Run 'runperl' with the complete perl program contained in '$prog', and # arguments in the hash referred to by '$runperl_args'. The results are # returned, with $? set to the exit code. Unless overridden, stderr is # redirected to stdout. # # Placing the program in a file bypasses various sh vagaries die sprintf "Second argument to fresh_perl_.* must be hashref of args to fresh_perl (or {})" unless !(defined $runperl_args) || ref($runperl_args) eq 'HASH'; # Given the choice of the mis-parsable {} # (we want an anon hash, but a borked lexer might think that it's a block) # or relying on taking a reference to a lexical # (\ might be mis-parsed, and the reference counting on the pad may go # awry) # it feels like the least-worse thing is to assume that auto-vivification # works. At least, this is only going to be a run-time failure, so won't # affect tests using this file but not this function. $runperl_args->{progfile} ||= $tmpfile; $runperl_args->{stderr} = 1 unless exists $runperl_args->{stderr}; open TEST, '>', $tmpfile or die "Cannot open $tmpfile: $!"; binmode TEST, ':utf8' if $runperl_args->{wide_chars}; print TEST $prog; close TEST or die "Cannot close $tmpfile: $!"; my $results = runperl(%$runperl_args); my $status = $?; # Not necessary to save this, but it makes it clear to # future maintainers. # Clean up the results into something a bit more predictable. $results =~ s/\n+$//; $results =~ s/at\s+$::tempfile_regexp\s+line/at - line/g; $results =~ s/of\s+$::tempfile_regexp\s+aborted/of - aborted/g; # bison says 'parse error' instead of 'syntax error', # various yaccs may or may not capitalize 'syntax'. $results =~ s/^(syntax|parse) error/syntax error/mig; if ($is_vms) { # some tests will trigger VMS messages that won't be expected $results =~ s/\n?%[A-Z]+-[SIWEF]-[A-Z]+,.*//; # pipes double these sometimes $results =~ s/\n\n/\n/g; } $? = $status; return $results; } sub _fresh_perl { my($prog, $action, $expect, $runperl_args, $name) = @_; my $results = fresh_perl($prog, $runperl_args); my $status = $?; # Use the first line of the program as a name if none was given unless( $name ) { (my $first_line, $name) = $prog =~ /^((.{1,50}).*)/; $name = $name . '...' if length $first_line > length $name; } # Historically this was implemented using a closure, but then that means # that the tests for closures avoid using this code. Given that there # are exactly two callers, doing exactly two things, the simpler approach # feels like a better trade off. my $pass; if ($action eq 'eq') { $pass = is($results, $expect, $name); } elsif ($action eq '=~') { $pass = like($results, $expect, $name); } else { die "_fresh_perl can't process action '$action'"; } unless ($pass) { _diag "# PROG: \n$prog\n"; _diag "# STATUS: $status\n"; } return $pass; } # # fresh_perl_is # # Combination of run_perl() and is(). # sub fresh_perl_is { my($prog, $expected, $runperl_args, $name) = @_; # _fresh_perl() is going to clip the trailing newlines off the result. # This will make it so the test author doesn't have to know that. $expected =~ s/\n+$//; local $Level = 2; _fresh_perl($prog, 'eq', $expected, $runperl_args, $name); } # # fresh_perl_like # # Combination of run_perl() and like(). # sub fresh_perl_like { my($prog, $expected, $runperl_args, $name) = @_; local $Level = 2; _fresh_perl($prog, '=~', $expected, $runperl_args, $name); } # Many tests use the same format in __DATA__ or external files to specify a # sequence of (fresh) tests to run, extra files they may temporarily need, and # what the expected output is. Putting it here allows common code to serve # these multiple tests. # # Each program is source code to run followed by an "EXPECT" line, followed # by the expected output. # # The first line of the code to run may be a command line switch such as -wE # or -0777 (alphanumerics only; only one cluster, beginning with a minus is # allowed). Later lines may contain (note the '# ' on each): # # TODO reason for todo # # SKIP reason for skip # # SKIP ?code to test if this should be skipped # # NAME name of the test (as with ok($ok, $name)) # # The expected output may contain: # OPTION list of options # OPTIONS list of options # # The possible options for OPTION may be: # regex - the expected output is a regular expression # random - all lines match but in any order # fatal - the code will fail fatally (croak, die) # nonfatal - the code is not expected to fail fatally # # If the actual output contains a line "SKIPPED" the test will be # skipped. # # If the actual output contains a line "PREFIX", any output starting with that # line will be ignored when comparing with the expected output # # If the global variable $FATAL is true then OPTION fatal is the # default. our $FATAL; sub _setup_one_file { my $fh = shift; # Store the filename as a program that started at line 0. # Real files count lines starting at line 1. my @these = (0, shift); my ($lineno, $current); while (<$fh>) { if ($_ eq "########\n") { if (defined $current) { push @these, $lineno, $current; } undef $current; } else { if (!defined $current) { $lineno = $.; } $current .= $_; } } if (defined $current) { push @these, $lineno, $current; } ((scalar @these) / 2 - 1, @these); } sub setup_multiple_progs { my ($tests, @prgs); foreach my $file (@_) { next if $file =~ /(?:~|\.orig|,v)$/; next if $file =~ /perlio$/ && !PerlIO::Layer->find('perlio'); next if -d $file; open my $fh, '<', $file or die "Cannot open $file: $!\n" ; my $found; while (<$fh>) { if (/^__END__/) { $found = $found + 1; # don't use ++ last; } } # This is an internal error, and should never happen. All bar one of # the files had an __END__ marker to signal the end of their preamble, # although for some it wasn't technically necessary as they have no # tests. It might be possible to process files without an __END__ by # seeking back to the start and treating the whole file as tests, but # it's simpler and more reliable just to make the rule that all files # must have __END__ in. This should never fail - a file without an # __END__ should not have been checked in, because the regression tests # would not have passed. die "Could not find '__END__' in $file" unless $found; my ($t, @p) = _setup_one_file($fh, $file); $tests += $t; push @prgs, @p; close $fh or die "Cannot close $file: $!\n"; } return ($tests, @prgs); } sub run_multiple_progs { my $up = shift; my @prgs; if ($up) { # The tests in lib run in a temporary subdirectory of t, and always # pass in a list of "programs" to run @prgs = @_; } else { # The tests below t run in t and pass in a file handle. In theory we # can pass (caller)[1] as the second argument to report errors with # the filename of our caller, as the handle is always DATA. However, # line numbers in DATA count from the __END__ token, so will be wrong. # Which is more confusing than not providing line numbers. So, for now, # don't provide line numbers. No obvious clean solution - one hack # would be to seek DATA back to the start and read to the __END__ token, # but that feels almost like we should just open $0 instead. # Not going to rely on undef in list assignment. my $dummy; ($dummy, @prgs) = _setup_one_file(shift); } my $tmpfile = tempfile(); my $count_failures = 0; my ($file, $line); PROGRAM: while (defined ($line = shift @prgs)) { $_ = shift @prgs; unless ($line) { $file = $_; if (defined $file) { print "# From $file\n"; } next; } my $switch = ""; my @temps ; my @temp_path; if (s/^(\s*-\w+)//) { $switch = $1; } my ($prog, $expected) = split(/\nEXPECT(?:\n|$)/, $_, 2); my %reason; foreach my $what (qw(skip todo)) { $prog =~ s/^#\s*\U$what\E\s*(.*)\n//m and $reason{$what} = $1; # If the SKIP reason starts ? then it's taken as a code snippet to # evaluate. This provides the flexibility to have conditional SKIPs if ($reason{$what} && $reason{$what} =~ s/^\?//) { my $temp = eval $reason{$what}; if ($@) { die "# In \U$what\E code reason:\n# $reason{$what}\n$@"; } $reason{$what} = $temp; } } my $name = ''; if ($prog =~ s/^#\s*NAME\s+(.+)\n//m) { $name = $1; } elsif (defined $file) { $name = "test from $file at line $line"; } if ($reason{skip}) { SKIP: { skip($name ? "$name - $reason{skip}" : $reason{skip}, 1); } next PROGRAM; } if ($prog =~ /--FILE--/) { my @files = split(/\n?--FILE--\s*([^\s\n]*)\s*\n/, $prog) ; shift @files ; die "Internal error: test $_ didn't split into pairs, got " . scalar(@files) . "[" . join("%%%%", @files) ."]\n" if @files % 2; while (@files > 2) { my $filename = shift @files; my $code = shift @files; push @temps, $filename; if ($filename =~ m#(.*)/# && $filename !~ m#^\.\./#) { require File::Path; File::Path::mkpath($1); push(@temp_path, $1); } open my $fh, '>', $filename or die "Cannot open $filename: $!\n"; print $fh $code; close $fh or die "Cannot close $filename: $!\n"; } shift @files; $prog = shift @files; } open my $fh, '>', $tmpfile or die "Cannot open >$tmpfile: $!"; print $fh q{ BEGIN { push @INC, '.'; open STDERR, '>&', STDOUT or die "Can't dup STDOUT->STDERR: $!;"; } }; print $fh "\n#line 1\n"; # So the line numbers don't get messed up. print $fh $prog,"\n"; close $fh or die "Cannot close $tmpfile: $!"; my $results = runperl( stderr => 1, progfile => $tmpfile, stdin => undef, $up ? (switches => ["-I$up/lib", $switch], nolib => 1) : (switches => [$switch]) ); my $status = $?; $results =~ s/\n+$//; # allow expected output to be written as if $prog is on STDIN $results =~ s/$::tempfile_regexp/-/g; if ($^O eq 'VMS') { # some tests will trigger VMS messages that won't be expected $results =~ s/\n?%[A-Z]+-[SIWEF]-[A-Z]+,.*//; # pipes double these sometimes $results =~ s/\n\n/\n/g; } # bison says 'parse error' instead of 'syntax error', # various yaccs may or may not capitalize 'syntax'. $results =~ s/^(syntax|parse) error/syntax error/mig; # allow all tests to run when there are leaks $results =~ s/Scalars leaked: \d+\n//g; $expected =~ s/\n+$//; my $prefix = ($results =~ s#^PREFIX(\n|$)##) ; # any special options? (OPTIONS foo bar zap) my $option_regex = 0; my $option_random = 0; my $fatal = $FATAL; if ($expected =~ s/^OPTIONS? (.+)(?:\n|\Z)//) { foreach my $option (split(' ', $1)) { if ($option eq 'regex') { # allow regular expressions $option_regex = 1; } elsif ($option eq 'random') { # all lines match, but in any order $option_random = 1; } elsif ($option eq 'fatal') { # perl should fail $fatal = 1; } elsif ($option eq 'nonfatal') { # used to turn off default fatal $fatal = 0; } else { die "$0: Unknown OPTION '$option'\n"; } } } die "$0: can't have OPTION regex and random\n" if $option_regex + $option_random > 1; my $ok = 0; if ($results =~ s/^SKIPPED\n//) { print "$results\n" ; $ok = 1; } else { if ($option_random) { my @got = sort split "\n", $results; my @expected = sort split "\n", $expected; $ok = "@got" eq "@expected"; } elsif ($option_regex) { $ok = $results =~ /^$expected/; } elsif ($prefix) { $ok = $results =~ /^\Q$expected/; } else { $ok = $results eq $expected; } if ($ok && $fatal && !($status >> 8)) { $ok = 0; } } local $::TODO = $reason{todo}; unless ($ok) { my $err_line = ''; $err_line .= "FILE: $file ; line $line\n" if defined $file; $err_line .= "PROG: $switch\n$prog\n" . "EXPECTED:\n$expected\n"; $err_line .= "EXIT STATUS: != 0\n" if $fatal; $err_line .= "GOT:\n$results\n"; $err_line .= "EXIT STATUS: " . ($status >> 8) . "\n" if $fatal; if ($::TODO) { $err_line =~ s/^/# /mg; print $err_line; # Harness can't filter it out from STDERR. } else { print STDERR $err_line; ++$count_failures; die "PERL_TEST_ABORT_FIRST_FAILURE set Test Failure" if $ENV{PERL_TEST_ABORT_FIRST_FAILURE}; } } if (defined $file) { _ok($ok, "at $file line $line", $name); } else { # We don't have file and line number data for the test, so report # errors as coming from our caller. local $Level = $Level + 1; ok($ok, $name); } foreach (@temps) { unlink $_ if $_; } foreach (@temp_path) { File::Path::rmtree $_ if -d $_; } } if ( $count_failures ) { print STDERR <<'EOS'; # # Note: 'run_multiple_progs' run has one or more failures # you can consider setting the environment variable # PERL_TEST_ABORT_FIRST_FAILURE=1 before running the test # to stop on the first error. # EOS } return; } sub can_ok ($@) { my($proto, @methods) = @_; my $class = ref $proto || $proto; unless( @methods ) { return _ok( 0, _where(), "$class->can(...)" ); } my @nok = (); foreach my $method (@methods) { local($!, $@); # don't interfere with caller's $@ # eval sometimes resets $! eval { $proto->can($method) } || push @nok, $method; } my $name; $name = @methods == 1 ? "$class->can('$methods[0]')" : "$class->can(...)"; _ok( !@nok, _where(), $name ); } # Call $class->new( @$args ); and run the result through object_ok. # See Test::More::new_ok sub new_ok { my($class, $args, $obj_name) = @_; $args ||= []; $obj_name = "The object" unless defined $obj_name; local $Level = $Level + 1; my $obj; my $ok = eval { $obj = $class->new(@$args); 1 }; my $error = $@; if($ok) { object_ok($obj, $class, $obj_name); } else { ok( 0, "new() died" ); diag("Error was: $@"); } return $obj; } sub isa_ok ($$;$) { my($object, $class, $obj_name) = @_; my $diag; $obj_name = 'The object' unless defined $obj_name; my $name = "$obj_name isa $class"; if( !defined $object ) { $diag = "$obj_name isn't defined"; } else { my $whatami = ref $object ? 'object' : 'class'; # We can't use UNIVERSAL::isa because we want to honor isa() overrides local($@, $!); # eval sometimes resets $! my $rslt = eval { $object->isa($class) }; my $error = $@; # in case something else blows away $@ if( $error ) { if( $error =~ /^Can't call method "isa" on unblessed reference/ ) { # It's an unblessed reference $obj_name = 'The reference' unless defined $obj_name; if( !UNIVERSAL::isa($object, $class) ) { my $ref = ref $object; $diag = "$obj_name isn't a '$class' it's a '$ref'"; } } elsif( $error =~ /Can't call method "isa" without a package/ ) { # It's something that can't even be a class $obj_name = 'The thing' unless defined $obj_name; $diag = "$obj_name isn't a class or reference"; } else { die <isa on your object and got some weird error. This should never happen. Please contact the author immediately. Here's the error. $@ WHOA } } elsif( !$rslt ) { $obj_name = "The $whatami" unless defined $obj_name; my $ref = ref $object; $diag = "$obj_name isn't a '$class' it's a '$ref'"; } } _ok( !$diag, _where(), $name ); } sub class_ok { my($class, $isa, $class_name) = @_; # Written so as to count as one test local $Level = $Level + 1; if( ref $class ) { ok( 0, "$class is a reference, not a class name" ); } else { isa_ok($class, $isa, $class_name); } } sub object_ok { my($obj, $isa, $obj_name) = @_; local $Level = $Level + 1; if( !ref $obj ) { ok( 0, "$obj is not a reference" ); } else { isa_ok($obj, $isa, $obj_name); } } # Purposefully avoiding a closure. sub __capture { push @::__capture, join "", @_; } sub capture_warnings { my $code = shift; local @::__capture; local $SIG {__WARN__} = \&__capture; local $Level = 1; &$code; return @::__capture; } # This will generate a variable number of tests. # Use done_testing() instead of a fixed plan. sub warnings_like { my ($code, $expect, $name) = @_; local $Level = $Level + 1; my @w = capture_warnings($code); cmp_ok(scalar @w, '==', scalar @$expect, $name); foreach my $e (@$expect) { if (ref $e) { like(shift @w, $e, $name); } else { is(shift @w, $e, $name); } } if (@w) { diag("Saw these additional warnings:"); diag($_) foreach @w; } } sub _fail_excess_warnings { my($expect, $got, $name) = @_; local $Level = $Level + 1; # This will fail, and produce diagnostics is($expect, scalar @$got, $name); diag("Saw these warnings:"); diag($_) foreach @$got; } sub warning_is { my ($code, $expect, $name) = @_; die sprintf "Expect must be a string or undef, not a %s reference", ref $expect if ref $expect; local $Level = $Level + 1; my @w = capture_warnings($code); if (@w > 1) { _fail_excess_warnings(0 + defined $expect, \@w, $name); } else { is($w[0], $expect, $name); } } sub warning_like { my ($code, $expect, $name) = @_; die sprintf "Expect must be a regexp object" unless ref $expect eq 'Regexp'; local $Level = $Level + 1; my @w = capture_warnings($code); if (@w > 1) { _fail_excess_warnings(0 + defined $expect, \@w, $name); } else { like($w[0], $expect, $name); } } # Set a watchdog to timeout the entire test file. The input seconds is # multiplied by $ENV{PERL_TEST_TIME_OUT_FACTOR} (default 1; minimum 1). # Set this in your profile for slow boxes, or use it to override the timeout # temporarily for debugging. # # NOTE: If the test file uses 'threads', then call the watchdog() function # _AFTER_ the 'threads' module is loaded. { # Closure my $watchdog; my $watchdog_thread; sub watchdog ($;$) { my $timeout = shift; # If cancelling, use the state variables to know which method was used to # create the watchdog. if ($timeout == 0) { if ($watchdog_thread) { $watchdog_thread->kill('KILL'); undef $watch_dog_thread; } elsif ($watchdog) { kill('KILL', $watchdog); undef $watch_dog; } else { alarm(0); } return; } # Make sure these aren't defined. undef $watchdog; undef $watchdog_thread; my $method = shift || ""; my $timeout_msg = 'Test process timed out - terminating'; # Accept either spelling my $timeout_factor = $ENV{PERL_TEST_TIME_OUT_FACTOR} || $ENV{PERL_TEST_TIMEOUT_FACTOR} || 1; $timeout_factor = 1 if $timeout_factor < 1; $timeout_factor = $1 if $timeout_factor =~ /^(\d+)$/; # Valgrind slows perl way down so give it more time before dying. $timeout_factor = 10 if $timeout_factor < 10 && $ENV{PERL_VALGRIND}; $timeout *= $timeout_factor; my $pid_to_kill = $$; # PID for this process if ($method eq "alarm") { goto WATCHDOG_VIA_ALARM; } # shut up use only once warning my $threads_on = $threads::threads && $threads::threads; # Don't use a watchdog process if 'threads' is loaded - # use a watchdog thread instead if (!$threads_on || $method eq "process") { # On Windows and VMS, try launching a watchdog process # using system(1, ...) (see perlport.pod). system() returns # immediately on these platforms with effectively a pid of the new # process if ($is_mswin || $is_vms) { # On Windows, try to get the 'real' PID if ($is_mswin) { eval { require Win32; }; if (defined(&Win32::GetCurrentProcessId)) { $pid_to_kill = Win32::GetCurrentProcessId(); } } # If we still have a fake PID, we can't use this method at all return if ($pid_to_kill <= 0); # Launch watchdog process undef $watchdog; eval { local $SIG{'__WARN__'} = sub { _diag("Watchdog warning: $_[0]"); }; my $sig = $is_vms ? 'TERM' : 'KILL'; my $prog = "sleep($timeout);" . "warn qq/# $timeout_msg" . '\n/;' . "kill(q/$sig/, $pid_to_kill);"; # If we're in taint mode PATH will be tainted $ENV{PATH} =~ /(.*)/s; local $ENV{PATH} = untaint_path($1); # On Windows use the indirect object plus LIST form to guarantee # that perl is launched directly rather than via the shell (see # perlfunc.pod), and ensure that the LIST has multiple elements # since the indirect object plus COMMANDSTRING form seems to # hang (see perl #121283). Don't do this on VMS, which doesn't # support the LIST form at all. if ($is_mswin) { my $runperl = which_perl(); $runperl =~ /(.*)/; $runperl = $1; if ($runperl =~ m/\s/) { $runperl = qq{"$runperl"}; } $watchdog = system({ $runperl } 1, $runperl, '-e', $prog); } else { my $cmd = _create_runperl(prog => $prog); $watchdog = system(1, $cmd); } }; if ($@ || ($watchdog <= 0)) { _diag('Failed to start watchdog'); _diag($@) if $@; undef($watchdog); return; } # Add END block to parent to terminate and # clean up watchdog process eval("END { local \$! = 0; local \$? = 0; wait() if kill('KILL', $watchdog); };"); return; } # Try using fork() to generate a watchdog process undef $watchdog; eval { $watchdog = fork() }; if (defined($watchdog)) { if ($watchdog) { # Parent process # Add END block to parent to terminate and # clean up watchdog process eval "END { local \$! = 0; local \$? = 0; wait() if kill('KILL', $watchdog); };"; return; } ### Watchdog process code # Load POSIX if available eval { require POSIX; }; # Execute the timeout sleep($timeout - 2) if ($timeout > 2); # Workaround for perlbug #49073 sleep(2); # Kill test process if still running if (kill(0, $pid_to_kill)) { _diag($timeout_msg); kill('KILL', $pid_to_kill); if ($is_cygwin) { # sometimes the above isn't enough on cygwin sleep 1; # wait a little, it might have worked after all system("/bin/kill -f $pid_to_kill") if kill(0, $pid_to_kill); } } # Don't execute END block (added at beginning of this file) $NO_ENDING = 1; # Terminate ourself (i.e., the watchdog) POSIX::_exit(1) if (defined(&POSIX::_exit)); exit(1); } # fork() failed - fall through and try using a thread } # Use a watchdog thread because either 'threads' is loaded, # or fork() failed if (eval {require threads; 1}) { $watchdog_thread = 'threads'->create(sub { # Load POSIX if available eval { require POSIX; }; $SIG{'KILL'} = sub { threads->exit(); }; # Detach after the signal handler is set up; the parent knows # not to signal until detached. 'threads'->detach(); # Execute the timeout my $time_left = $timeout; do { $time_left = $time_left - sleep($time_left); } while ($time_left > 0); # Kill the parent (and ourself) select(STDERR); $| = 1; _diag($timeout_msg); POSIX::_exit(1) if (defined(&POSIX::_exit)); my $sig = $is_vms ? 'TERM' : 'KILL'; kill($sig, $pid_to_kill); }); # Don't proceed until the watchdog has set up its signal handler. # (Otherwise there is a possibility that we will exit with threads # running.) The watchdog tells us the handler is set by detaching # itself. (The 'is_running()' is a fail-safe.) while ( $watchdog_thread->is_running() && ! $watchdog_thread->is_detached()) { 'threads'->yield(); } return; } # If everything above fails, then just use an alarm timeout WATCHDOG_VIA_ALARM: if (eval { alarm($timeout); 1; }) { # Load POSIX if available eval { require POSIX; }; # Alarm handler will do the actual 'killing' $SIG{'ALRM'} = sub { select(STDERR); $| = 1; _diag($timeout_msg); POSIX::_exit(1) if (defined(&POSIX::_exit)); my $sig = $is_vms ? 'TERM' : 'KILL'; kill($sig, $pid_to_kill); }; } } } # End closure # Orphaned Docker or Linux containers do not necessarily attach to PID 1. They might attach to 0 instead. sub is_linux_container { if ($^O eq 'linux' && open my $fh, '<', '/proc/1/cgroup') { while(<$fh>) { if (m{^\d+:pids:(.*)} && $1 ne '/init.scope') { return 1; } } } return 0; } 1;