qpdf/qtest/module/TestDriver.pm

1863 lines
49 KiB
Perl

# -*- perl -*-
#
# This file is part of qtest.
#
# Copyright 1993-2022, Jay Berkenbilt
#
# QTest is distributed under the terms of version 2.0 of the Artistic
# license which may be found in the source distribution.
#
# Search for "PUBLIC METHODS" to find the public methods and
# documentation on how to use them.
require 5.008;
use strict;
package TestDriver::PidKiller;
use vars qw($f_pid);
$f_pid = 'pid';
sub new
{
my $class = shift;
my $rep = +{+__PACKAGE__ => {} };
$rep->{+__PACKAGE__}{$f_pid} = shift;
bless $rep, $class;
}
sub DESTROY
{
my $rep = shift;
my $pid = $rep->{+__PACKAGE__}{$f_pid};
defined($$pid) && $$pid && kill 15, $$pid;
}
package TestDriver;
use IO::Handle;
use IO::File;
use IO::Socket;
use IO::Select;
use POSIX ':sys_wait_h';
use File::Copy;
use File::Find;
use Carp;
use Cwd;
require QTC;
# Constants
# Possible test case outcomes
use constant PASS => 'PASS';
use constant FAIL => 'FAIL';
# Input/Output keys
use constant STRING => 'STRING';
use constant FILE => 'FILE';
use constant COMMAND => 'COMMAND';
use constant FILTER => 'FILTER';
use constant REGEXP => 'REGEXP';
use constant EXIT_STATUS => 'EXIT_STATUS';
use constant THREAD_DATA => 'THREAD_DATA';
use constant TD_THREADS => 'TD_THREADS';
use constant TD_SEQGROUPS => 'TD_SEQGROUPS';
# Flags
use constant NORMALIZE_NEWLINES => 1 << 0;
use constant NORMALIZE_WHITESPACE => 1 << 1;
use constant EXPECT_FAILURE => 1 << 2;
use constant RM_WS_ONLY_LINES => 1 << 3;
# Field names
use vars (qw($f_socket $f_origdir $f_tempdir $f_testlog),
qw($f_testxml $f_testjunit $f_suitename));
$f_socket = 'socket';
$f_origdir = 'origdir';
$f_tempdir = 'tempdir';
$f_testlog = 'testlog';
$f_testxml = 'testxml';
$f_testjunit = 'testjunit';
$f_suitename = 'suitename';
use vars qw($f_passes $f_fails $f_xpasses $f_xfails $f_testnum);
$f_passes = 'passes'; # expected passes
$f_fails = 'fails'; # unexpected failures
$f_xpasses = 'xpasses'; # unexpected passes
$f_xfails = 'xfails'; # expected failures
$f_testnum = 'testnum';
# Static Variables
# QTEST_MARGIN sets the number of spaces to after PASSED or FAILED and
# before the rightmost column of the screen.
my $margin = $ENV{'QTEST_MARGIN'} || 8;
$margin += $ENV{'QTEST_EXTRA_MARGIN'} || 0;
my $ncols = 80;
my $color_reset = "";
my $color_green = "";
my $color_yellow = "";
my $color_red = "";
my $color_magenta = "";
my $color_emph = "";
# MSWin32 support
my $in_windows = 0;
my $winbin = undef;
if (($^O eq 'MSWin32') || ($^O eq 'msys'))
{
$in_windows = 1;
}
sub get_tty_features
{
my $got_size = 0;
eval
{
require Term::ReadKey;
if (-t STDOUT)
{
# This prints error messages if STDOUT is not a tty, so
# check even if -stdout-tty=1 was given.
($ncols, undef, undef, undef) =
Term::ReadKey::GetTerminalSize(\*STDOUT);
}
$got_size = 1;
};
if (! $got_size)
{
eval
{
# Get screen columns if possible
no strict;
local $^W = 0;
local *X;
{
local $SIG{'__WARN__'} = sub {};
require 'sys/ioctl.ph';
}
if ((defined &TIOCGWINSZ) && open(X, "+</dev/tty"))
{
my $winsize = "";
if (ioctl(X, &TIOCGWINSZ, $winsize))
{
(undef, $ncols) = unpack('S4', $winsize);
$got_size = 1;
}
close(X);
}
};
}
if (! defined $ncols)
{
$ncols = 80;
}
eval
{
if ($in_windows)
{
eval
{
# If you don't have this module, you may want to set
# the environment variable ANSI_COLORS_DISABLED to 1
# to avoid "garbage" output around PASSED, FAILED,
# etc.
require Win32::Console::ANSI;
}
}
require Term::ANSIColor;
$color_reset = Term::ANSIColor::RESET();
$color_green = Term::ANSIColor::GREEN();
$color_yellow = Term::ANSIColor::YELLOW();
$color_red = Term::ANSIColor::RED();
$color_magenta = Term::ANSIColor::MAGENTA();
$color_emph = Term::ANSIColor::color('bold blue on_black');
};
}
# Static Methods
sub print_and_pad
{
my $str = shift;
my $spaces = $ncols - 10 - length($str) - $margin;
$spaces = 0 if $spaces < 0;
print $str . (' ' x $spaces) . ' ... ';
}
sub print_results
{
my ($outcome, $exp_outcome) = @_;
my $color = "";
my $outcome_text;
if ($outcome eq $exp_outcome)
{
if ($outcome eq PASS)
{
&QTC::TC("testdriver", "TestDriver expected pass");
$color = $color_green;
$outcome_text = "PASSED";
}
else
{
&QTC::TC("testdriver", "TestDriver expected fail");
$color = $color_yellow;
# " (exp)" is fewer characters than the default margin
# which keeps this from wrapping lines with default
# settings.
$outcome_text = "FAILED (exp)";
}
}
else
{
if ($outcome eq PASS)
{
&QTC::TC("testdriver", "TestDriver unexpected pass");
$color = $color_magenta;
$outcome_text = "PASSED-UNEXP";
}
else
{
&QTC::TC("testdriver", "TestDriver unexpected fail");
$color = $color_red;
$outcome_text = "FAILED";
}
}
print $color, $outcome_text, $color_reset, "\n";
$outcome_text;
}
# Normal Methods
sub new
{
my $class = shift;
my $rep = +{+__PACKAGE__ => {} };
if (@_ != 1)
{
croak "Usage: ", __PACKAGE__, "->new(\"test-suite name\")\n";
}
my $suitename = shift;
if (! ((@ARGV == 13) &&
(($ARGV[0] eq '-fd') || ($ARGV[0] eq '-port')) &&
($ARGV[2] eq '-origdir') &&
($ARGV[4] eq '-tempdir') &&
($ARGV[6] eq '-testlog') &&
($ARGV[8] eq '-testxml') &&
($ARGV[10] eq '-testjunit') &&
($ARGV[12] =~ m/^-stdout-tty=([01])$/) &&
(-d $ARGV[5])))
{
die +__PACKAGE__, ": improper invocation of test driver $0 (" .
join(' ', @ARGV) . ")\n";
}
my $fd = ($ARGV[0] eq '-fd') ? $ARGV[1] : undef;
my $port = ($ARGV[0] eq '-port') ? $ARGV[1] : undef;
my $origdir = $ARGV[3];
my $tempdir = $ARGV[5];
my $testlogfile = $ARGV[7];
my $testxmlfile = $ARGV[9];
my $testjunitfile = $ARGV[11];
my $testlog = new IO::File(">>$testlogfile");
binmode $testlog;
my $testxml = new IO::File(">>$testxmlfile");
binmode $testxml;
my $testjunit = new IO::File(">>$testjunitfile");
binmode $testjunit;
$ARGV[12] =~ m/=([01])/ or die +__PACKAGE__, ": INTERNAL ERROR in ARGV[10]";
my $stdout_is_tty = $1;
if ($stdout_is_tty)
{
get_tty_features();
}
my $socket;
if (defined $fd)
{
$socket = new IO::Handle;
if (! $socket->fdopen($fd, "w+"))
{
warn +__PACKAGE__, ": unable to open file descriptor $fd.\n";
warn +__PACKAGE__, " must be created from a program invoked by" .
" the test driver system\n";
die +__PACKAGE__, ": initialization failed";
}
}
else
{
$socket = IO::Socket::INET->new(
PeerAddr => '127.0.0.1', PeerPort => $port) or
die "unable to connect to port $port: $!\n";
}
$socket->autoflush();
binmode $socket;
# Do some setup that would ordinarily be reserved for a main
# program. We want test suites to behave in a certain way so tha
# the overall system works as desired.
# Killing the driver should cause to to exit. Without this, it
# may cause whatever subsidiary program is being run to exit and
# the driver to continue to the next test case.
$SIG{'INT'} = $SIG{'HUP'} = $SIG{'TERM'} = $SIG{'QUIT'} = sub { exit 2 };
# Unbuffer our output.
$| = 1;
$rep->{+__PACKAGE__}{$f_socket} = $socket;
$rep->{+__PACKAGE__}{$f_origdir} = $origdir;
$rep->{+__PACKAGE__}{$f_tempdir} = $tempdir;
$rep->{+__PACKAGE__}{$f_testlog} = $testlog;
$rep->{+__PACKAGE__}{$f_testxml} = $testxml;
$rep->{+__PACKAGE__}{$f_testjunit} = $testjunit;
$rep->{+__PACKAGE__}{$f_suitename} = $suitename;
$rep->{+__PACKAGE__}{$f_passes} = 0;
$rep->{+__PACKAGE__}{$f_fails} = 0;
$rep->{+__PACKAGE__}{$f_xpasses} = 0;
$rep->{+__PACKAGE__}{$f_xfails} = 0;
$rep->{+__PACKAGE__}{$f_testnum} = 1;
# Do protocol handshaking with the test driver system
my $init = scalar(<$socket>);
if ($init !~ m/^TEST_DRIVER 1$/)
{
die +__PACKAGE__, ": incorrect protocol with test driver system\n";
}
$socket->print("TEST_DRIVER_CLIENT 1\n");
bless $rep, $class;
}
sub _socket
{
my $rep = shift;
$rep->{+__PACKAGE__}{$f_socket};
}
sub _tempdir
{
my $rep = shift;
$rep->{+__PACKAGE__}{$f_tempdir};
}
sub _testlog
{
my $rep = shift;
$rep->{+__PACKAGE__}{$f_testlog};
}
sub _testxml
{
my $rep = shift;
$rep->{+__PACKAGE__}{$f_testxml};
}
sub _testjunit
{
my $rep = shift;
$rep->{+__PACKAGE__}{$f_testjunit};
}
sub _suitename
{
my $rep = shift;
$rep->{+__PACKAGE__}{$f_suitename};
}
sub _testnum
{
my $rep = shift;
$rep->{+__PACKAGE__}{$f_testnum} = $_[0] if @_;
$rep->{+__PACKAGE__}{$f_testnum};
}
# PUBLIC METHODS
# Usage: report(n)
# Specify the number of tests that are expected to have been run.
# Please note: the purpose of reporting the number of test cases with
# "report" is as an extra check to make sure that the test suite
# itself didn't have a logic error that caused some test cases to be
# skipped. The argument to "report" should therefore be a hard-coded
# number or a number computed only from static features in the test
# suite. It should not be a number that is counted up during the
# process of running the test suite. Computing this number as a side
# effect of running test cases would defeat the purpose of the number.
# For example, if the test suite consists of an array of test cases,
# and the test suite code iterates through that loop and calls
# "runtest" twice for each element, it would be reasonable to pass an
# expression that includes the size of the array as an argument to
# "report", but it would not be appropriate to have a variable called
# "$ntests" that is incremented each time "runtest" is called and then
# passed to "report".
sub report
{
my $rep = shift;
croak "Usage: ", __PACKAGE__, "->report(num-tests-expected)\n"
unless @_ && $_[0] =~ m/^\d+$/;
# Message to test driver system:
# n-expected-tests passes fails unexpected-passes expected-fails
my @vals = (shift);
push(@vals, map { $rep->{+__PACKAGE__}{$_} } ($f_passes, $f_fails,
$f_xpasses, $f_xfails));
my $socket = $rep->_socket();
$socket->print(join(' ', @vals));
$socket->flush();
}
# Usage: notify(string)
# Prints the string followed by a newline to standard output of the
# test suite.
sub notify
{
my $rep = shift;
my $msg = shift;
&QTC::TC("testdriver", "TestDriver notify");
print $msg, "\n";
}
# Usage: emphasize(string)
# Prints the string followed by a newline to standard output of the
# test suite. The string is printed with emphasis if the terminal
# supports color.
sub emphasize
{
my $rep = shift;
my $msg = shift;
&QTC::TC("testdriver", "TestDriver emphasize");
print $color_emph, $msg, $color_reset, "\n";
}
# Usage: prompt(msg, env, default)
# If the environment variable "env" is set, its value is returned.
# Otherwise, if STDIN is a tty, the user is prompted for an answer
# using msg as the prompt, or if STDIN is not a tty, the value
# specified in "default" is returned. Note that careless use of
# prompt in test suites may make the test suites unable to be run in
# batch mode.
sub prompt
{
my $rep = shift;
my ($msg, $env, $default) = @_;
&QTC::TC("testdriver", "TestDriver prompt");
my $answer = $ENV{$env};
if (defined $answer)
{
print "$msg\n";
print "[Question answered from environment variable \$$env: $answer]\n";
}
else
{
print "To avoid question, place answer in" .
" environment variable \$$env\n";
# Note: ActiveState perl 5.10.1 gives the wrong answer for -t
# STDIN when NUL (http://bugs.activestate.com/show_bug.cgi?id=85614).
if ((-t STDIN) && (-t STDOUT))
{
print "$msg ";
chop($answer = <STDIN>);
if ($answer eq '')
{
print "[Using default answer for question: $default]\n";
$answer = $default;
}
}
else
{
print "$msg\n";
print "[Using default answer for question: $default]\n";
$answer = $default;
}
}
$answer;
}
# Usage: get_start_dir()
# Returns the name of the directory from which the test driver was
# originally invoked. This can be useful for test suites that are
# designed to be run from read-only areas or from multiple locations
# simultaneously: they can get the original invocation directory and
# use it as a place to write temporary files.
sub get_start_dir
{
my $rep = shift;
$rep->{+__PACKAGE__}{$f_origdir};
}
# Usage: runtest description input output [ flags ]
# Returns true iff test passes; i.e., input matches output
# Parameters:
# description: a short textual description of the test case
# input: a hash reference that defines the input to the test case
# input keys and associated values:
# STRING: a string that is used verbatim as the test input
# FILE: a file whose contents are used as the test input
# COMMAND: an array reference containing a command and arguments
# or a string representing the command. This is passed to exec,
# so the rules that exec uses to determine whether to pass this
# to a shell are followed. The command is run with STDIN set to
# /dev/null, STDOUT redirected to an internal file, and STDERR
# copied to STDOUT.
# Note that exactly one of STRING, FILE, or COMMAND must appear.
# FILTER: if specified, it is a program that is run on the test
# input specified above to generate the true test input.
# output: a hash reference that defines the expected output of the
# test case
# STRING: a string that contains the expected test output
# FILE: a file that contains the expected test output
# REGEXP: a regular expression that must match the test output
# Note that exactly one of STRING, FILE, or REGEXP must appear.
# EXIT_STATUS: the exit status of the command. Required iff the
# intput is specified by COMMAND. A value of undef means that we
# don't care about the exit status of a command. The special
# value of '!0' means we allow any abnormal exit status but we
# don't care what the specific exit status is. An integer value
# is the ordinary exit status of a command. A string of the form
# SIG:n indicates that the program has exited with signal n.
# Note that SIG:n is not reliable in a Windows (non-Cygwin)
# environment.
# THREAD_DATA: If specified, the test output is expected to
# contain multithreaded output with output lines marked by thread
# and sequence group identifiers. The value must be a hash that
# contains required key TD_THREADS and optional key TD_SEQGROUPS.
# The value of each key is an array reference containing a list
# of threads or sequence groups as appropriate. When THREAD_DATA
# is specified, the single call to runtest actually generates t +
# s + 3 tests where "t" is the number of threads and "s" is the
# number of sequence groups specified. See the documentation for
# full details on how multithreaded output is handled by the test
# driver.
# flags: additional flags to control the test case; should be
# logically orred together (e.g. NORMALIZE_WHITESPACE | EXPECT_FAILURE)
# NORMALIZE_NEWLINES: If specified, all newlines or carriage
# return/newline combinations in the input are translated to
# straight UNIX-style newlines. This is done before writing
# through any filter. Newlines are also normalized in the
# expected output.
# NORMALIZE_WHITESPACE: If specified, all carriage returns are
# removed, and all strings of one or more space or tab characters
# are replaced by a single space character in the input. This is
# done before writing through any filter. The expected output
# must be normalized in this way as well in order for the test to
# pass.
# EXPECT_FAILURE: If specified, the test case is expected to
# fail. In this case, a test case failure will not generate
# verbose output or cause overall test suite failure, and a pass
# will generate test suite failure. This should be used for
# place-holder test cases that exercise a known bug that cannot
# yet be fixed.
# RM_WS_ONLY_LINES: If specified, all lines only containing any
# whitespace character like newlines, spaces or tabs are removed
# from the input. This is done before writing through any filter
# and is especially useful if some tests output more newlines on
# some platforms than on others.
sub runtest
{
my $rep = shift;
if (! ((@_ == 3) || (@_ == 4)))
{
croak +("Usage: ", +__PACKAGE__,
"->runtest(description, input, output[, flags])\n");
}
my ($description, $input, $output, $flags) = @_;
$flags = 0 unless defined $flags;
my $tempdir = $rep->_tempdir();
if (ref($description) ne '')
{
&QTC::TC("testdriver", "TestDriver description not string");
croak +__PACKAGE__, "->runtest: description must be a string\n";
}
if (ref($input) ne 'HASH')
{
&QTC::TC("testdriver", "TestDriver input not hash");
croak +__PACKAGE__, "->runtest: input must be a hash reference\n";
}
if (ref($output) ne 'HASH')
{
&QTC::TC("testdriver", "TestDriver output not hash");
croak +__PACKAGE__, "->runtest: output must be a hash reference\n";
}
if ((ref($flags) ne '') || ($flags !~ m/^\d+$/))
{
&QTC::TC("testdriver", "TestDriver flags not integer");
croak +__PACKAGE__, "->runtest: flags must be an integer\n";
}
my ($extra_in_keys, $in_string, $in_file, $in_command, $in_filter) =
check_hash_keys($input, $rep->STRING,
$rep->FILE, $rep->COMMAND, $rep->FILTER);
if ($extra_in_keys)
{
&QTC::TC("testdriver", "TestDriver extraneous input keys");
croak +(+__PACKAGE__,
"->runtest: extraneous keys in intput hash: $extra_in_keys\n");
}
my ($extra_out_keys, $out_string, $out_file, $out_regexp,
$out_exit_status, $thread_data) =
check_hash_keys($output, $rep->STRING,
$rep->FILE, $rep->REGEXP, $rep->EXIT_STATUS,
$rep->THREAD_DATA);
if ($extra_out_keys)
{
&QTC::TC("testdriver", "TestDriver extraneous output keys");
croak +(+__PACKAGE__,
"->runtest: extraneous keys in output hash: $extra_out_keys\n");
}
if ((((defined $in_string) ? 1 : 0) +
((defined $in_file) ? 1 : 0) +
((defined $in_command) ? 1 : 0)) != 1)
{
&QTC::TC("testdriver", "TestDriver invalid input");
croak +__PACKAGE__, "->runtest: exactly one of" .
" STRING, FILE, or COMMAND must be present for input\n";
}
if ((((defined $out_string) ? 1 : 0) +
((defined $out_file) ? 1 : 0) +
((defined $out_regexp) ? 1 : 0)) != 1)
{
&QTC::TC("testdriver", "TestDriver invalid output");
croak +__PACKAGE__, "->runtest: exactly one of" .
" STRING, FILE, or REGEXP must be present for output\n";
}
if ((defined $in_command) != (exists $output->{$rep->EXIT_STATUS}))
{
&QTC::TC("testdriver", "TestDriver invalid status");
croak +__PACKAGE__, "->runtest: input COMMAND and output EXIT_STATUS"
. " must either both appear both not appear\n";
}
my ($threads, $seqgroups) = (undef, undef);
if (defined $thread_data)
{
if (ref($thread_data) ne 'HASH')
{
&QTC::TC("testdriver", "TestDriver thread_data not hash");
croak +__PACKAGE__, "->runtest: THREAD_DATA" .
" must be a hash reference\n";
}
my $extra_thread_keys;
($extra_thread_keys, $threads, $seqgroups) =
check_hash_keys($thread_data, $rep->TD_THREADS, $rep->TD_SEQGROUPS);
if ($extra_thread_keys)
{
&QTC::TC("testdriver", "TestDriver extraneous thread_data keys");
croak +(+__PACKAGE__,
"->runtest: extraneous keys in THREAD_DATA hash:" .
" $extra_thread_keys\n");
}
if (! defined $threads)
{
&QTC::TC("testdriver", "TestDriver thread_data no threads");
croak +__PACKAGE__, "->runtest: THREAD_DATA" .
" must contain TD_THREADS\n";
}
elsif (ref($threads) ne 'ARRAY')
{
&QTC::TC("testdriver", "TestDriver threads not array ref");
croak +__PACKAGE__, "->runtest: TD_THREADS" .
" must be an array reference\n";
}
if ((defined $seqgroups) && (ref($seqgroups) ne 'ARRAY'))
{
&QTC::TC("testdriver", "TestDriver seqgroups not array ref");
croak +__PACKAGE__, "->runtest: TD_SEQGROUPS" .
" must be an array reference\n";
}
}
# testnum is incremented by print_testid
my $testnum = $rep->_testnum();
my $category = $rep->_suitename();
$rep->print_testid($description);
# Open a file handle to read the raw (unfiltered) test input
my $pid = undef;
my $pid_killer = new TestDriver::PidKiller(\$pid);
my $in = new IO::Handle;
my $use_tempfile = $in_windows;
my $tempout_status = undef;
if (defined $in_string)
{
&QTC::TC("testdriver", "TestDriver input string");
open($in, '<', \$in_string) or
die +(+__PACKAGE__,
"->runtest: unable to read from input string: $!\n");
}
elsif (defined $in_file)
{
&QTC::TC("testdriver", "TestDriver input file");
if (! open($in, '<', $in_file))
{
# If the input file is not found, generate a string to use
# as input, and make the string different each run to
# prevent sloppy commiting of the string as the expected
# output. This prevents complete failure in cases when the
# input file is not found, such as when one test's input
# file is generated by something that may have failed
# earlier.
my $not_found = "qtest, pid=$$: input file $in_file not found\n";
open($in, '<', \$not_found) or
croak +(+__PACKAGE__,
"->runtest: unable to read from input file $in_file: $!\n");
}
}
elsif (defined $in_command)
{
if (ref($in_command) eq 'ARRAY')
{
&QTC::TC("testdriver", "TestDriver input command array");
}
elsif (ref($in_command) eq '')
{
&QTC::TC("testdriver", "TestDriver input command string");
}
if ($use_tempfile)
{
my $tempout = "$tempdir/tempout";
$tempout_status = $rep->winrun(
$in_command, File::Spec->devnull(), $tempout);
open($in, "<$tempout") or
croak +(+__PACKAGE__,
"->runtest: unable to read from" .
" input file $tempout: $!\n");
}
else
{
$pid = open($in, "-|");
croak +__PACKAGE__, "->runtest: fork failed: $!\n"
unless defined $pid;
if ($pid == 0)
{
open(STDERR, ">&STDOUT");
open(STDIN, '<', \ "");
if (ref($in_command) eq 'ARRAY')
{
exec @$in_command or
croak+(+__PACKAGE__,
"->runtest: unable to run command ",
join(' ', @$in_command), "\n");
}
else
{
exec $in_command or
croak+(+__PACKAGE__,
"->runtest: unable to run command ",
$in_command, "\n");
}
}
}
}
else
{
die +__PACKAGE__, ": INTERNAL ERROR: invalid test input";
}
binmode $in;
# Open file handle into which to write the actual output
my $actual = new IO::File;
my $actual_file = "$tempdir/actual";
if (defined $in_filter)
{
&QTC::TC("testdriver", "TestDriver filter defined");
if ($use_tempfile)
{
my $filter_file = "$tempdir/filter";
open(F, ">$filter_file.1") or
croak+(+__PACKAGE__,
"->runtest: unable to create $filter_file.1: $!\n");
binmode F;
while (<$in>)
{
if ($flags & $rep->NORMALIZE_NEWLINES)
{
s/\r$//;
}
print F;
}
$in->close();
close(F);
$rep->winrun($in_filter, "$filter_file.1", $filter_file);
open($in, "<$filter_file") or
croak +(+__PACKAGE__,
"->runtest: unable to read from" .
" input file $filter_file: $!\n");
binmode $in;
$in_filter = undef;
}
}
if (defined $in_filter)
{
# Write through filter to actual file
open($actual, "| $in_filter > $actual_file") or
croak +(+__PACKAGE__,
": pipe to filter $in_filter failed: $!\n");
}
else
{
&QTC::TC("testdriver", "TestDriver filter not defined");
open($actual, ">$actual_file") or
die +(+__PACKAGE__, ": write to $actual_file failed: $!\n");
}
binmode $actual;
# Write from input to actual output, normalizing spaces and
# newlines if needed
my $exit_status = undef;
while (1)
{
my ($line, $status) = read_line($in, $pid);
$exit_status = $status if defined $status;
last unless defined $line;
if ($flags & $rep->NORMALIZE_WHITESPACE)
{
&QTC::TC("testdriver", "TestDriver normalize whitespace");
$line =~ s/[ \t]+/ /g;
}
else
{
&QTC::TC("testdriver", "TestDriver no normalize whitespace");
}
if ($flags & $rep->NORMALIZE_NEWLINES)
{
&QTC::TC("testdriver", "TestDriver normalize newlines");
$line =~ s/\r$//;
}
else
{
&QTC::TC("testdriver", "TestDriver no normalize newlines");
}
if ($flags & $rep->RM_WS_ONLY_LINES)
{
&QTC::TC("testdriver", "TestDriver remove empty lines");
$line =~ s/^\s+$//;
}
else
{
&QTC::TC("testdriver", "TestDriver no remove empty lines");
}
$actual->print($line);
$actual->flush();
last if defined $exit_status;
}
$in->close();
if (defined $tempout_status)
{
$exit_status = $tempout_status;
}
if (defined $in_command)
{
if (! defined $exit_status)
{
$exit_status = $?;
}
my $exit_status_number = 0;
my $exit_status_signal = 0;
if ($in_windows)
{
# WIFSIGNALED et al are not defined. This is emperically
# what happens with MSYS 1.0.11 and ActiveState Perl
# 5.10.1.
if ($exit_status & 0x8000)
{
$exit_status_signal = 1;
$exit_status = ($exit_status & 0xfff) >> 8;
$exit_status = "SIG:$exit_status";
}
elsif ($exit_status >= 256)
{
$exit_status_number = 1;
$exit_status = $exit_status >> 8;
}
}
elsif (WIFSIGNALED($exit_status))
{
$exit_status_signal = 1;
$exit_status = "SIG:" . WTERMSIG($exit_status);
}
elsif (WIFEXITED($exit_status))
{
$exit_status_number = 1;
$exit_status = WEXITSTATUS($exit_status);
}
if ($exit_status_number)
{
&QTC::TC("testdriver", "TestDriver exit status number");
}
if ($exit_status_signal)
{
&QTC::TC("testdriver", "TestDriver exit status signal");
}
}
$? = 0;
$actual->close();
$pid = undef;
if ($?)
{
die +(+__PACKAGE__,
"->runtest: failure closing actual output; status = $?\n");
}
# Compare exit statuses. This expression is always true when the
# input was not from a command.
if ((defined $out_exit_status) && ($out_exit_status eq '!0'))
{
&QTC::TC("testdriver", "TestDriver non-zero exit status");
}
my $status_match =
((! defined $out_exit_status) ||
((defined $exit_status) &&
( (($out_exit_status eq '!0') && ($exit_status ne 0)) ||
($exit_status eq $out_exit_status) )));
# Compare actual output with expected output.
my $expected_file = undef;
my $output_match = undef;
if (defined $out_string)
{
&QTC::TC("testdriver", "TestDriver output string");
# Write output string to a file so we can run diff
$expected_file = "$tempdir/expected";
my $e = new IO::File;
open($e, ">$expected_file") or
die +(__PACKAGE__,
"->runtest: unable to write to $expected_file: $!\n");
binmode $e;
$e->print($out_string);
$e->close();
}
elsif (defined $out_file)
{
&QTC::TC("testdriver", "TestDriver output file");
if ($flags & $rep->NORMALIZE_NEWLINES)
{
# Normalize newlines in expected output file
$expected_file = "$tempdir/expected";
unlink $expected_file;
my $in = new IO::File;
if (open($in, "<$out_file"))
{
binmode $in;
my $e = new IO::File;
open($e, ">$expected_file") or
die +(__PACKAGE__,
"->runtest: unable to write to $expected_file: $!\n");
binmode $e;
while (<$in>)
{
s/\r?$//;
$e->print($_);
}
$e->close();
$in->close();
}
}
else
{
$expected_file = $out_file;
}
}
elsif (defined $out_regexp)
{
&QTC::TC("testdriver", "TestDriver output regexp");
# No expected file; do regexp test to determine whether output
# matches
$actual = new IO::File;
open($actual, "<$actual_file") or
die +(__PACKAGE__,
"->runtest: unable to read $actual_file: $!\n");
binmode $actual;
local $/ = undef;
my $actual_output = <$actual>;
$actual->close();
$output_match = ($actual_output =~ m/$out_regexp/);
}
else
{
die +__PACKAGE__, ": INTERNAL ERROR: invalid test output";
}
my $output_diff = undef;
if (! defined $output_match)
{
if (! defined $expected_file)
{
die +__PACKAGE__, ": INTERNAL ERROR: expected_file not defined";
}
if (defined $threads)
{
# Real output comparisons are done later.
$output_match = 1;
}
else
{
$output_diff = "$tempdir/difference";
my $r = $rep->safe_pipe(['diff', '-a', '-u',
$expected_file, $actual_file],
$output_diff);
$output_match = ($r == 0);
}
}
my $outcome = ($output_match && $status_match) ? PASS : FAIL;
my $exp_outcome = (($flags & $rep->EXPECT_FAILURE) ? FAIL : PASS);
my $outcome_text = print_results($outcome, $exp_outcome);
my $passed = $rep->update_counters($outcome, $exp_outcome);
my $testxml = $rep->_testxml();
my $testjunit = $rep->_testjunit();
my $testlog = $rep->_testlog();
# $outcome_text is for the human-readable. We need something
# different for the xml file.
$testxml->print(" <testcase\n" .
" testid=\"" . xmlify($category, 1) . " $testnum\"\n" .
" description=\"" . xmlify($description, 1) . "\"\n" .
" outcome=\"" .
(($outcome eq PASS)
? ($passed ? "pass" : "unexpected-pass")
: ($passed ? "expected-fail" : "fail")) .
"\"\n");
$testjunit->print(" <testcase\n" .
" id=\"" . xmlify($category, 1) . " $testnum\"\n" .
" name=\"" . xmlify($description, 1) . "\"\n");
if (($outcome eq FAIL) && ($outcome ne $exp_outcome))
{
# Test failed and failure was not expected
$testxml->print(" >\n");
$testjunit->print(" >\n" .
" <failure>\n");
$testlog->printf("$category test %d (%s) FAILED\n",
$testnum, $description);
my $cwd = getcwd();
$testlog->print("cwd: $cwd\n");
$testjunit->print("cwd: " . xmlify($cwd) . "\n");
$testxml->print(" <cwd>" . xmlify($cwd) . "</cwd>\n");
my $cmd = $in_command;
if ((defined $cmd) && (ref($cmd) eq 'ARRAY'))
{
$cmd = join(' ', @$cmd);
}
if (defined $cmd)
{
$testlog->print("command: $cmd\n");
$testxml->print(" <command>" . xmlify($cmd) . "</command>\n");
$testjunit->print("command: " . xmlify($cmd) . "\n");
}
if (defined $out_file)
{
# Use $out_file, not $expected_file -- we are only
# interested in dispaying this information if the user's
# real output was original in a file.
$testlog->print("expected output in $out_file\n");
$testxml->print(
" <expected-output-file>" . xmlify($out_file) .
"</expected-output-file>\n");
$testjunit->print("expected output in " .
xmlify($out_file) . "\n");
}
# It would be nice if we could filter out internal calls for
# times when runtest is called inside of the module for
# multithreaded testing.
$testlog->print(Carp::longmess());
$testxml->print(" <stacktrace>test failure" .
xmlify(Carp::longmess()) .
"</stacktrace>\n");
$testjunit->print("stracktrace:\n" . xmlify(Carp::longmess()));
if (! $status_match)
{
&QTC::TC("testdriver", "TestDriver status mismatch");
$testlog->printf("\tExpected status: %s\n", $out_exit_status);
$testlog->printf("\tActual status: %s\n", $exit_status);
$testxml->print(
" <expected-status>$out_exit_status</expected-status>\n");
$testxml->print(
" <actual-status>$exit_status</actual-status>\n");
$testjunit->printf(" Expected status: %s\n", $out_exit_status);
$testjunit->printf(" Actual status: %s\n", $exit_status);
}
if (! $output_match)
{
&QTC::TC("testdriver", "TestDriver output mismatch");
$testlog->print("--> BEGIN EXPECTED OUTPUT <--\n");
$testxml->print(" <expected-output>");
$testjunit->print("-- BEGIN EXPECTED OUTPUT --\n");
if (defined $expected_file)
{
write_file_to_fh($expected_file, $testlog);
xml_write_file_to_fh($expected_file, $testxml);
xml_write_file_to_fh($expected_file, $testjunit);
}
elsif (defined $out_regexp)
{
$testlog->print("regexp: " . $out_regexp);
if ($out_regexp !~ m/\n$/s)
{
$testlog->print("\n");
}
$testxml->print("regexp: " . xmlify($out_regexp));
$testjunit->print("regexp: " . xmlify($out_regexp));
}
else
{
die +(+__PACKAGE__,
"->runtest: INTERNAL ERROR: no expected output\n");
}
$testlog->print("--> END EXPECTED OUTPUT <--\n" .
"--> BEGIN ACTUAL OUTPUT <--\n");
$testxml->print("</expected-output>\n" .
" <actual-output>");
$testjunit->print("-- END EXPECTED OUTPUT --\n" .
"-- ACTUAL OUTPUT --\n");
write_file_to_fh($actual_file, $testlog);
xml_write_file_to_fh($actual_file, $testxml);
xml_write_file_to_fh($actual_file, $testjunit);
$testlog->print("--> END ACTUAL OUTPUT <--\n");
$testxml->print("</actual-output>\n");
$testjunit->print("-- ACTUAL OUTPUT --\n");
if (defined $output_diff)
{
&QTC::TC("testdriver", "TestDriver display diff");
$testlog->print("--> DIFF EXPECTED ACTUAL <--\n");
$testxml->print(" <diff-output>");
$testjunit->print("-- DIFF EXPECTED ACTUAL --\n");
write_file_to_fh($output_diff, $testlog);
xml_write_file_to_fh($output_diff, $testxml);
xml_write_file_to_fh($output_diff, $testjunit);
$testlog->print("--> END DIFFERENCES <--\n");
$testxml->print("</diff-output>\n");
$testjunit->print("-- END DIFFERENCES --\n");
}
else
{
&QTC::TC("testdriver", "TestDriver display no diff");
}
}
$testxml->print(" </testcase>\n");
$testjunit->print(" </failure>\n" .
" </testcase>\n");
}
else
{
$testxml->print(" />\n");
$testjunit->print(" />\n");
}
if (defined $threads)
{
if (! defined $expected_file)
{
&QTC::TC("testdriver", "TestDriver thread data but no exp output");
croak +(+__PACKAGE__,
"->runtest: thread data invalid".
" without fixed test output\n");
}
my $thread_expected = "$tempdir/thread-expected";
my $thread_actual = "$tempdir/thread-actual";
copy($actual_file, $thread_actual);
filter_seqgroups($expected_file, $thread_expected);
$passed =
$rep->analyze_thread_data($description,
$expected_file, $actual_file,
$threads, $seqgroups)
&& $passed;
if ($passed)
{
$rep->runtest($description . ": all subcases passed",
{$rep->STRING => ""},
{$rep->STRING => ""});
}
else
{
$rep->runtest($description . ": original output",
{$rep->FILE => $thread_actual},
{$rep->FILE => $thread_expected});
}
unlink $thread_expected, $thread_actual;
}
$passed;
}
sub read_line
{
my ($fh, $pid) = @_;
my $line = undef;
my $status = undef;
if (defined $pid)
{
# It doesn't work to just call <$fh> in this case. For some
# unknown reason, some programs occasionally exit and cause an
# interrupted system call return from read which perl just
# ignores, making the call to <$fh> hang. To protect
# ourselves, we explicitly check for the program having exited
# periodically if read hasn't returned anything.
while (1)
{
my $s = new IO::Select();
$s->add($fh);
my @ready = $s->can_read(1);
if (@ready == 0)
{
if (waitpid($pid, WNOHANG) > 0)
{
$status = $?;
last;
}
next;
}
else
{
my $buf = "";
my $status = sysread($fh, $buf, 1);
if ((defined $status) && ($status == 1))
{
$line = "" unless defined $line;
$line .= $buf;
last if $buf eq "\n";
}
else
{
last;
}
}
}
}
else
{
$line = <$fh>;
}
($line, $status);
}
sub write_file_to_fh
{
my ($file, $out) = @_;
my $in = new IO::File("<$file");
if (defined $in)
{
binmode $in;
my $ended_with_newline = 1;
while (<$in>)
{
$out->print($_);
$ended_with_newline = m/\n$/s;
}
if (! $ended_with_newline)
{
$out->print("[no newline at end of data]\n");
}
$in->close();
}
else
{
$out->print("[unable to open $file: $!]\n");
}
}
sub xmlify
{
my ($str, $attr) = @_;
$attr = 0 unless defined $attr;
$str =~ s/\&/\&amp;/g;
$str =~ s/</&lt;/g;
$str =~ s/>/&gt;/g;
$str =~ s/\"/&quot;/g if $attr;
$str =~ s/([\000-\010\013-\037])/sprintf("[0x%02x]", ord($1))/ge;
$str =~ s/([\177-\377])/sprintf("&#x%02x;", ord($1))/ge;
$str;
}
sub xml_write_file_to_fh
{
my ($file, $out) = @_;
my $in = new IO::File("<$file");
if (defined $in)
{
binmode $in;
while (defined ($_ = <$in>))
{
$out->print(xmlify($_));
}
$in->close();
}
else
{
$out->print("[unable to open $file: $!]");
}
}
sub check_hash_keys
{
my ($hash, @keys) = @_;
my %actual_keys = ();
foreach my $k (keys %$hash)
{
$actual_keys{$k} = 1;
}
foreach my $k (@keys)
{
delete $actual_keys{$k};
}
my $extra_keys = join(', ', sort (keys %actual_keys));
($extra_keys, (map { $hash->{$_} } @keys));
}
sub print_testid
{
my $rep = shift;
my ($description) = @_;
my $testnum = $rep->_testnum();
my $category = $rep->_suitename();
print_and_pad(sprintf("$category %2d (%s)", $testnum, $description));
my $tc_filename = $ENV{'TC_FILENAME'} || "";
if ($tc_filename && open(F, ">>$tc_filename"))
{
binmode F;
printf F "# $category %2d (%s)\n", $testnum, $description;
close(F);
}
$rep->_testnum(++$testnum);
}
sub update_counters
{
my $rep = shift;
my ($outcome, $exp_outcome) = @_;
(($outcome eq PASS) && ($exp_outcome eq PASS)) &&
$rep->{+__PACKAGE__}{$f_passes}++;
(($outcome eq PASS) && ($exp_outcome eq FAIL)) &&
$rep->{+__PACKAGE__}{$f_xpasses}++;
(($outcome eq FAIL) && ($exp_outcome eq PASS)) &&
$rep->{+__PACKAGE__}{$f_fails}++;
(($outcome eq FAIL) && ($exp_outcome eq FAIL)) &&
$rep->{+__PACKAGE__}{$f_xfails}++;
($outcome eq PASS);
}
sub analyze_thread_data
{
my $rep = shift;
my ($description, $expected, $actual,
$expected_threads, $expected_seqgroups) = @_;
my $tempdir = $rep->_tempdir();
my %actual_threads = ();
my %actual_seqgroups = ();
my @errors = ();
$rep->thread_cleanup();
$rep->split_combined($expected);
$rep->analyze_threaded_output
($actual, \%actual_threads, \%actual_seqgroups, \@errors);
# Make sure we saw the right threads and sequences
my $desired = "threads:\n";
$desired .= join('', map { " $_\n" } (sort @$expected_threads));
$desired .= "sequence groups:\n";
if (defined $expected_seqgroups)
{
$desired .= join('', map { " $_\n" } (sort @$expected_seqgroups));
}
my $observed = "threads:\n";
$observed .= join('', map { " $_\n" } (sort keys %actual_threads));
$observed .= "sequence groups:\n";
$observed .= join('', map { " $_\n" } (sort keys %actual_seqgroups));
if (@errors)
{
$observed .= join('', @errors);
}
my $passed =
$rep->runtest("$description: multithreaded data",
{$rep->STRING => $observed},
{$rep->STRING => $desired});
foreach my $th (@{$expected_threads})
{
create_if_missing("$tempdir/$th.thread-actual",
"[no actual output]\n");
filter_seqgroups("$tempdir/$th.thread-expected",
"$tempdir/$th.thread-filtered");
$passed =
$rep->runtest($description . ": thread $th",
{$rep->FILE => "$tempdir/$th.thread-actual"},
{$rep->FILE => "$tempdir/$th.thread-filtered"})
&& $passed;
}
if (defined $expected_seqgroups)
{
foreach my $sg (@{$expected_seqgroups})
{
create_if_missing("$tempdir/$sg.seq-actual",
"[no actual output]\n");
$passed =
$rep->runtest($description . ": seqgroup $sg",
{$rep->FILE => "$tempdir/$sg.seq-actual"},
{$rep->FILE => "$tempdir/$sg.seq-expected"})
&& $passed;
}
}
$rep->thread_cleanup();
$passed;
}
sub analyze_threaded_output
{
my $rep = shift;
my ($file, $threads, $seqgroups, $errors) = @_;
my $sequence_checking = 1;
open(F, "<$file") or die +__PACKAGE__, ": can't open $file: $!\n";
binmode F;
my $cur_thread = undef;
while (<F>)
{
if (m/^(\[\[(.+?)\]\]:)/)
{
my $tag = $1;
my $thread = $2;
my $rest = $'; #' [unconfuse emacs font lock mode]
$rep->handle_line($file, $., $tag, $thread, $rest,
\$sequence_checking, $threads, $seqgroups,
$errors);
$cur_thread = $thread;
}
else
{
$rep->handle_line($file, $., "", $cur_thread, $_,
\$sequence_checking, $threads, $seqgroups,
$errors);
}
}
close(F);
}
sub handle_line
{
my $rep = shift;
my ($file, $lineno, $tag, $thread, $rest,
$sequence_checking, $threads, $seqgroups, $errors) = @_;
my $tempdir = $rep->_tempdir();
if (! exists $threads->{$thread})
{
my $fh = new IO::File("<$tempdir/$thread.thread-expected");
if (! $fh)
{
&QTC::TC("testdriver", "TestDriver no input file for thread");
$fh = undef;
$$sequence_checking = 0;
push(@$errors,
"$file:$.: no input file for thread $thread; " .
"sequence checking abandoned\n");
}
else
{
binmode $fh;
}
$threads->{$thread} = $fh;
}
my $known = defined($threads->{$thread});
my $seqs = "";
if ($$sequence_checking)
{
my $fh = $threads->{$thread};
my $next_input_line = scalar(<$fh>);
if (! defined $next_input_line)
{
$next_input_line = "[EOF]\n";
}
$seqs = $rep->strip_seqs(\$next_input_line);
if ($next_input_line eq $rest)
{
if ($seqs ne "")
{
$rep->handle_seqs($seqs, $tag . $rest, $seqgroups);
}
}
else
{
&QTC::TC("testdriver", "TestDriver thread mismatch");
$$sequence_checking = 0;
push(@$errors,
"$file:$.: thread $thread mismatch; " .
"sequencing checking abandoned\n" .
"actual $rest" .
"expected $next_input_line");
}
}
output_line("$tempdir/$thread.thread-actual", $rest);
if (! $known)
{
&QTC::TC("testdriver", "TestDriver output from unknown thread");
push(@$errors, "[[$thread]]:$rest");
}
}
sub strip_seqs
{
my $rep = shift;
my $linep = shift;
my $seqs = "";
if ($$linep =~ s/^\(\(.*?\)\)//)
{
$seqs = $&;
}
$seqs;
}
sub handle_seqs
{
my $rep = shift;
my ($seqs, $line, $seqgroups) = @_;
my $tempdir = $rep->_tempdir();
$seqs =~ s/^\(\((.*?)\)\)/$1/;
foreach my $seq (split(',', $seqs))
{
$seqgroups->{$seq} = 1;
output_line("$tempdir/$seq.seq-actual", $line);
}
}
sub filter_seqgroups
{
my ($infile, $outfile) = @_;
open(F, "<$infile") or
die +__PACKAGE__, ": can't open $infile: $!\n";
binmode F;
open(O, ">$outfile") or
die +__PACKAGE__, ": can't create $outfile: $!\n";
binmode O;
while (<F>)
{
s/^((?:\[\[.+?\]\]:)?)\(\(.+?\)\)/$1/;
print O;
}
close(O);
close(F);
}
sub output_line
{
my ($file, $line) = @_;
open(O, ">>$file") or die +__PACKAGE__, ": can't open $file: $!\n";
binmode O;
print O $line or die +__PACKAGE__, ": can't append to $file: $!\n";
close(O) or die +__PACKAGE__, ": close $file failed: $!\n";
}
sub create_if_missing
{
my ($file, $line) = @_;
if (! -e $file)
{
open(O, ">$file") or die +__PACKAGE__, ": can't create $file: $!\n";
binmode O;
print O $line;
close(O);
}
}
sub split_combined
{
my $rep = shift;
my $combined = shift;
my $tempdir = $rep->_tempdir();
open(C, "<$combined") or die +__PACKAGE__, ": can't open $combined: $!\n";
binmode C;
my %files = ();
my $last_thread_fh = undef;
while (<C>)
{
my $thread_fh = $last_thread_fh;
my $thread_out = undef;
if (m/^(\[\[(.+?)\]\]:)(\(\((.+?)\)\))?(.*\n?)$/)
{
my $thread_full = $1;
my $thread = $2;
my $seq_full = $3;
my $seq = $4;
my $rest = $5;
my $seq_out = undef;
$thread_out = $rest;
my @seq_files = ();
my $thread_file = "$tempdir/$thread.thread-expected";
if (defined $seq_full)
{
$thread_out = $seq_full . $thread_out;
$seq_out = $thread_full . $rest;
foreach my $s (split(/,/, $seq))
{
my $f = "$tempdir/$s.seq-expected";
my $fh = cache_open(\%files, $f);
$fh->print($seq_out);
}
}
$thread_fh = cache_open(\%files, $thread_file);
}
else
{
$thread_out = $_;
}
if ((defined $thread_out) && (! defined $thread_fh))
{
die +__PACKAGE__, ": no place to put output lines\n";
}
$thread_fh->print($thread_out) if defined $thread_out;
$last_thread_fh = $thread_fh;
}
close(C);
map { $_->close() } (values %files);
}
sub cache_open
{
my ($cache, $file) = @_;
if (! defined $file)
{
return undef;
}
if (! exists $cache->{$file})
{
unlink $file;
my $fh = new IO::File(">$file") or
die +__PACKAGE__, ": can't open $file: $!\n";
binmode $fh;
$cache->{$file} = $fh;
}
$cache->{$file};
}
sub thread_cleanup
{
my $rep = shift;
my $dir = $rep->_tempdir();
my @files = +(grep { m/\.(thread|seq)-(actual|expected|filtered)$/ }
(glob("$dir/*")));
if (@files)
{
unlink @files;
}
}
sub rmrf
{
my $path = shift;
return unless -e $path;
my $wanted = sub
{
if ((-d $_) && (! -l $_))
{
rmdir $_ or die "rmdir $_ failed: $!\n";
}
else
{
unlink $_ or die "unlink $_ failed: $!\n";
}
};
finddepth({wanted => $wanted, no_chdir => 1}, $path);
}
sub safe_pipe
{
my $rep = shift;
my ($cmd, $outfile) = @_;
my $result = 0;
if ($in_windows)
{
$result = $rep->winrun($cmd, File::Spec->devnull(), $outfile);
}
else
{
my $pid = open(C, "-|");
if ($pid)
{
# parent
my $out = new IO::File(">$outfile") or
die +__PACKAGE__, ": can't open $outfile: $!\n";
binmode C;
while (<C>)
{
$out->print($_);
}
close(C);
$result = $?;
$out->close();
}
else
{
# child
open(STDERR, ">&STDOUT");
exec(@$cmd) || die +__PACKAGE__, ": $cmd->[0] failed: $!\n";
}
}
$result;
}
sub winrun
{
# This function does several things to make running stuff on
# Windows look sort of like running things on UNIX. It assumes
# MinGW perl is running in an MSYS/MinGW environment.
#
# * When an MSYS/MinGW program is run with system("..."), its
# newlines generate \r\n, but when it's run from MSYS sh, its
# newlines generate \n. We want \n for UNIX-like programs.
#
# * system("...") in perl doesn't have any special magic to
# handle #! lines in scripts. A lot of test suites will count
# on that.
#
# * There's no Windows equivalent to execve with separate
# arguments, so all sorts of fancy quoting is necessary when *
# dealing with arguments with spaces, etc.
#
# * Pipes work unreliably. Fork emulation is very incomplete.
#
# To work around these issues, we ensure that everything is
# actually executed from the MSYS /bin/sh. We find the actual
# path of that and then write a shell script which we explicitly
# invoke as an argument to /bin/sh. If we have a string that we
# want executed with /bin/sh, we include the string in the shell
# script. If we have an array, we pass the array on the
# commandline to the shell script and let it preserve spacing. We
# also do our output redirection in the shell script itself since
# redirection of STDOUT and STDERR doesn't carry forward to
# programs invoked by programs we invoke. Finally, we filter out
# errors generated by the script itself, since it is supposed to
# be an invisible buffer for smoother execution of programs.
# Experience shows that its output comes from things like printing
# the names of signals generated by subsidiary programs.
my $rep = shift;
my ($in_command, $in, $out) = @_;
my $tempdir = $rep->_tempdir();
my $tempfilename = "$tempdir/winrun.tmp";
if (! defined $winbin)
{
if ((system("sh -c 'cd /bin; pwd -W' > $tempfilename") == 0) &&
open(F, "<$tempfilename"))
{
$winbin = <F>;
close(F);
$winbin =~ s,[\r\n],,g;
$winbin =~ s,\\,/,g;
}
if (! defined $winbin)
{
die +__PACKAGE__, ": unable to find windows path to /bin\n";
}
}
my $script = "$tempdir/tmpscript";
open(F, ">$script") or
croak +(+__PACKAGE__,
"->runtest: unable to open $script to write: $!\n");
binmode F;
print F "exec >$tempfilename\n";
print F "exec 2>&1\n";
print F "exec <$in\n";
my @cmd = ("$winbin/sh", $script);
if (ref($in_command) eq 'ARRAY')
{
# For debugging, write out the args
foreach my $arg (@$in_command)
{
print F "# $arg\n";
}
print F '"$@"', "\n";
push(@cmd, @$in_command);
}
else
{
print F "$in_command\n";
}
close(F);
my $status = system @cmd;
if (open(IN, "<$tempfilename") &&
open(OUT, ">$out"))
{
binmode IN;
binmode OUT;
while (<IN>)
{
next if m/^$script:/;
print OUT;
}
close(IN);
close(OUT);
}
$status;
}
1;
#
# END OF TestDriver
#