pty tests fail on Fedora29

[DannyThomas]

possibly related to #56 probably not specific to Fedora29 probably not specific to perl-5.29.4

NB http://www.cpantesters.org/cpan/report/4549bd4e-c71d-11e8-bca8-6f605ebbffd9 Fedora28 is green but "t/pty.t .............................. skipped: IO::Pty not installed"

I don't know much about cpantesters, but shouldn't required modules be installed before distribution built/tested?

IPC-Run-20180523.0 Readonly-2.05 Test-Simple-1.302140 (for Test::More)

For now, I'm disabling t/pty.t because IPR-Run is a dependency in several subsequent distributions I build

t/pty.t .............................. # IO::Tty 1.12, IO::Pty 1.12
t/pty.t .............................. 1/32 
#   Failed test at t/pty.t line 136.
#          got: '012-456789--'
#     expected: '012--5678---'
# Looks like you failed 1 test of 32.
t/pty.t .............................. Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/32 subtests 

as a separate issue, should these tests be run on non-Win host?

t/win32_compile.t .................... ok   
t/windows_search_path.t .............. ok     

[toddr]

The issue is with IO::Pty. On Red Hat 7, I get:

$>perl -MIO::Pty -MIPC::Run::Debug -wE'print __FILE__ . " " . __LINE__ . "  " . IPC::Run::Debug::_map_fds() . "\n";my $h = IO::Pty->new; print __FILE__ . " " . __LINE__ . "  " . IPC::Run::Debug::_map_fds() . "\n";undef $h; print __FILE__ . " " . __LINE__ . "  " . IPC::Run::Debug::_map_fds() . "\n";'
-e 1  012---------
-e 1  01234-------
-e 1  012---------

But on RedHat 8, I get:

$>perl -MIO::Pty -MIPC::Run::Debug -wE'print __FILE__ . " " . __LINE__ . "  " . IPC::Run::Debug::_map_fds() . "\n";my $h = IO::Pty->new; print __FILE__ . " " . __LINE__ . "  " . IPC::Run::Debug::_map_fds() . "\n";undef $h; print __FILE__ . " " . __LINE__ . "  " . IPC::Run::Debug::_map_fds() . "\n";'
-e 1  012---------
-e 1  012345------
-e 1  012-4-------

When the IO::Pty object goes out of scope, one of the file handles is not closing for some reason.

[toddr]

https://rt.cpan.org/Ticket/Display.html?id=130491

[eserte]

Running the test with strace shows the following difference (both runs with perl 5.30.1):

• Fedora 31 (passes)

  42484 12:12:06.747088 dup(0)            = 3 <0.000005>
  42484 12:12:06.747113 close(3)          = 0 <0.000013>
  42484 12:12:06.747146 dup(1)            = 3 <0.000005>
  42484 12:12:06.747165 close(3)          = 0 <0.000005>
  42484 12:12:06.747185 dup(2)            = 3 <0.000005>
  42484 12:12:06.747203 close(3)          = 0 <0.000005>
  42484 12:12:06.747222 dup(3)            = -1 EBADF (Bad file descriptor) <0.000005>
  42484 12:12:06.747245 dup(4)            = -1 EBADF (Bad file descriptor) <0.000005>
  42484 12:12:06.747265 dup(5)            = 3 <0.000005>
  42484 12:12:06.747284 close(3)          = 0 <0.000005>
  42484 12:12:06.747303 dup(6)            = 3 <0.000005>
  42484 12:12:06.747321 close(3)          = 0 <0.000005>
  42484 12:12:06.747340 dup(7)            = 3 <0.000005>
  42484 12:12:06.747360 close(3)          = 0 <0.000005>
  42484 12:12:06.747379 dup(8)            = 3 <0.000005>
  42484 12:12:06.747397 close(3)          = 0 <0.000005>
  42484 12:12:06.747416 dup(9)            = -1 EBADF (Bad file descriptor) <0.000005>

• CentOS8 (fails)

  13882 12:10:15.705938 dup(0)            = 3 <0.000006>
  13882 12:10:15.705969 close(3)          = 0 <0.000005>
  13882 12:10:15.705997 dup(1)            = 3 <0.000005>
  13882 12:10:15.706020 close(3)          = 0 <0.000005>
  13882 12:10:15.706045 dup(2)            = 3 <0.000007>
  13882 12:10:15.706078 close(3)          = 0 <0.000007>
  13882 12:10:15.706111 dup(3)            = -1 EBADF (Bad file descriptor) <0.000007>
  13882 12:10:15.706149 dup(4)            = 3 <0.000005>
  13882 12:10:15.706176 close(3)          = 0 <0.000008>
  13882 12:10:15.706211 dup(5)            = 3 <0.000007>
  13882 12:10:15.706242 close(3)          = 0 <0.000005>
  13882 12:10:15.706266 dup(6)            = 3 <0.000005>
  13882 12:10:15.706305 close(3)          = 0 <0.000005>
  13882 12:10:15.706331 dup(7)            = 3 <0.000005>
  13882 12:10:15.706354 close(3)          = 0 <0.000005>
  13882 12:10:15.706400 dup(8)            = 3 <0.000005>
  13882 12:10:15.706430 close(3)          = 0 <0.000005>
  13882 12:10:15.706454 dup(9)            = -1 EBADF (Bad file descriptor) <0.000005>

[eserte]

... and lsof right before the failing test shows this additional open file descriptor on the failing system:

perl5.30. 14546 cpansand    4r   REG    8,1  6406312   144033 /var/lib/sss/mc/group

On the passing Fedora system this file does not exist (so it cannot be opened).

[eserte]

A combination of perl -d:Trace and strace shows that /var/lib/sss/mc/group is opened during this call:

/opt/perl-5.30.1/lib/site_perl/5.30.1/x86_64-linux/IO/Pty.pm:24:   my ($ptyfd, $ttyfd, $ttyname) = pty_allocate();

So, possible workarounds:

• create a dummy IO::Pty object before calling _map_fds e.g. like this:

  {my $pty = IO::Pty->new}

  This would force a pty_allocate call which would open the group sss file early.

• alternatively: do something which opens the sss group file as a side effect, e.g.

  getgrnam($();

[eserte]

First alternative could look like this: http://cpan.cpantesters.org/authors/id/S/SR/SREZIC/patches/IPC-Run-20180523.0-issue130.patch But now I think it would be a little bit cleaner if the dummy pty creation was done earlier, before the first SKIP. Theoretically sssd exists also for some *BSD systems listed for $platform_skip

[toddr]

I’m happy to fix the test, but do you think this might indicate a problem that we should be fixing in the code?

[kloczek]

@eserte

First alternative could look like this: http://cpan.cpantesters.org/authors/id/S/SR/SREZIC/patches/IPC-Run-20180523.0-issue130.patch

I've tested that patch on rawhide and looks like test suite passes correctly.

+ cd IPC-Run-20180523.0
+ /usr/bin/make -O -j48 V=1 VERBOSE=1 test
PERL_DL_NONLAZY=1 "/usr/bin/perl" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(0, 'blib/lib', 'blib/arch')" t/*.t
t/97_meta.t .......................... skipped: Author tests not required for installation
t/98_pod.t ........................... skipped: Author tests not required for installation
t/98_pod_coverage.t .................. skipped: Author tests not required for installation
t/99_perl_minimum_version.t .......... skipped: Author tests not required for installation
t/adopt.t ............................ skipped: adopt not implemented yet
t/autoflush.t ........................ ok
t/binmode.t .......................... ok
t/bogus.t ............................ ok
t/eintr.t ............................ ok
t/filter.t ........................... ok
t/harness.t .......................... ok
t/io.t ............................... ok
t/kill_kill.t ........................ ok
t/parallel.t ......................... ok
# IO::Tty 1.14, IO::Pty 1.14
t/pty.t .............................. ok
t/pump.t ............................. ok
t/readonly.t ......................... ok
t/run.t .............................. ok
t/run_stdin-callback-return-array.t .. ok
t/signal.t ........................... ok
t/timeout.t .......................... ok
t/timer.t ............................ ok
t/utf8.t ............................. ok
t/windows_search_path.t .............. ok
All tests successful.
Files=24, Tests=721, 16 wallclock secs ( 0.21 usr  0.15 sys +  1.98 cusr  1.99 csys =  4.33 CPU)

[kloczek]

BTW: do you have any plans to make new release? :)

[toddr]

https://metacpan.org/release/TODDR/IPC-Run-20200505.0