skip tainting tests if perl was built without taint support

[neilb]

From Perl 5.35.11 onwards, Configure lets you build a perl without taint support.

These changes skip tests that rely on taint support, so that this distribution can be installed under such a perl.

[jkeenan]

@neilb, problem!

When I finally manage to correctly configure and build a no-taint perl on my principal FreeBSD-12 machine, I get test failures ... and then one test file stops abruptly, necessitating a Ctrl-C.

$ pwd
/usr/home/jkeenan/gitwork/ipc-system-simple

$ bleadperl -v | head -2 | tail -1
This is perl 5, version 35, subversion 12 (v5.35.12 (v5.35.11-9-g99db5f9692)) built for amd64-freebsd-thread-multi

$ bleadperl -V:config_args
config_args='-des -Dusedevel -Uversiononly -Dprefix=/home/jkeenan/testing/blead -Dman1dir=none -Dman3dir=none -Duseithreads -Doptimize=-O2 -pipe -fstack-protector -fno-strict-aliasing -Utaint_support';

$ bleadperl -MConfig -E \
> 'say "taint: ", !exists($Config{taint_support}) || $Config{taint_support} ? "yes" : "no";'
taint: no

$ bleadperl Makefile.PL
Checking if your kit is complete...
Looks good
Generating a Unix-style Makefile
Writing Makefile for IPC::System::Simple
Writing MYMETA.yml and MYMETA.json

$ make
cp lib/IPC/System/Simple.pm blib/lib/IPC/System/Simple.pm

$ make test
PERL_DL_NONLAZY=1 "/usr/home/jkeenan/testing/blead/bin/perl" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(0, 'blib/lib', 'blib/arch')" t/*.t
t/01_load.t ........... ok   
t/02_exit.t ........... ok     
t/03_signal.t ......... ok   
t/04_capture.t ........ ok     
t/05_multi_capture.t .. ok     
t/06_fail.t ........... ok   
t/07_taint.t .......... 1/13 
#   Failed test 'Sanity - executable name is tainted'
#   at t/07_taint.t line 21.

#   Failed test 'Sanity - Evil zero is tainted'
#   at t/07_taint.t line 25.

#   Failed test 'Single-arg, tainted data'
#   at t/07_taint.t line 41.
#                   ''
#     doesn't match '(?^:called with tainted argument)'

#   Failed test 'multi-arg, tainted data'
#   at t/07_taint.t line 44.
#                   ''
#     doesn't match '(?^:called with tainted argument)'

#   Failed test 'Returns of multi-arg capture should be tainted'
#   at t/07_taint.t line 53.

#   Failed test 'Returns of single-arg capture should be tainted'
#   at t/07_taint.t line 56.
# Looks like you failed 6 tests of 13.
t/07_taint.t .......... Dubious, test returned 6 (wstat 1536, 0x600)
Failed 6/13 subtests 
	(less 2 skipped subtests: 5 okay)
t/08_core.t ........... ok   
t/09_system.t ......... ok     
t/10_formatting.t ..... 1/5 ^C*** Signal 2

I believe there has been a segfault.

$ gitst
On branch master
Your branch is up to date with 'origin/master'.

Untracked files:
  (use "git add <file>..." to include in what will be committed)
	t/perl.core

nothing added to commit but untracked files present (use "git add" to track)

Attempting the remaining tests with a prove built on this perl:

$ bleadprove -b t/11_newlines.t t/12_systemx.t t/13_exports.t t/14_uninitialised.t t/args.t t/internal.t t/win32.t
t/11_newlines.t ....... ok   
t/12_systemx.t ........ ok   
t/13_exports.t ........ ok   
t/14_uninitialised.t .. ok   
t/args.t .............. ok     
t/internal.t .......... ok   
t/win32.t ............. skipped: Win32 only tests
All tests successful.
Files=7, Tests=75,  1 wallclock secs ( 0.06 usr  0.02 sys +  0.58 cusr  0.15 csys =  0.80 CPU)
Result: PASS

[jkeenan]

Double-checking now; I think I was working in master.

[jkeenan]

When I switch to your branch, all tests PASS -- but there's a segfault!

[perlmonger: ipc-system-simple] $ bleadperl Makefile.PL && make
Checking if your kit is complete...
Looks good
Generating a Unix-style Makefile
Writing Makefile for IPC::System::Simple
Writing MYMETA.yml and MYMETA.json
cp lib/IPC/System/Simple.pm blib/lib/IPC/System/Simple.pm
[perlmonger: ipc-system-simple] $ make test
PERL_DL_NONLAZY=1 "/usr/home/jkeenan/testing/blead/bin/perl" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(0, 'blib/lib', 'blib/arch')" t/*.t
t/01_load.t ........... ok   
t/02_exit.t ........... ok     
t/03_signal.t ......... ok   
t/04_capture.t ........ ok     
t/05_multi_capture.t .. ok     
t/06_fail.t ........... ok   
t/07_taint.t .......... skipped: your perl was built without taint support
t/08_core.t ........... ok   
t/09_system.t ......... ok     
t/10_formatting.t ..... ok   
t/11_newlines.t ....... ok   
t/12_systemx.t ........ ok   
t/13_exports.t ........ ok   
t/14_uninitialised.t .. ok   
t/args.t .............. ok     
t/internal.t .......... ok   
t/win32.t ............. skipped: Win32 only tests
All tests successful.
Files=17, Tests=173,  2 wallclock secs ( 0.05 usr  0.04 sys +  1.48 cusr  0.34 csys =  1.91 CPU)
Result: PASS
[perlmonger: ipc-system-simple] $ gitst
On branch no-taint
Your branch is up to date with 'neilb/no-taint'.

Untracked files:
  (use "git add <file>..." to include in what will be committed)
	t/perl.core

nothing added to commit but untracked files present (use "git add" to track)

The segfault appears to be occuring when running t/08_core.t:

[perlmonger: ipc-system-simple] $ bleadperl -MBSD::Resource -E 'say qq|hello world|;'
hello world
[perlmonger: ipc-system-simple] $ bleadprove -vb t/08_core.t 
t/08_core.t .. 
1..3
ok 1 - use IPC::System::Simple;
ok 2 - Signal caught,   $? = 134
ok 3 - Coredump caught, $? = 134
ok
All tests successful.
Files=1, Tests=3,  0 wallclock secs ( 0.02 usr  0.02 sys +  0.09 cusr  0.03 csys =  0.16 CPU)
Result: PASS
[perlmonger: ipc-system-simple] $ gitst
On branch no-taint
Your branch is up to date with 'neilb/no-taint'.

Untracked files:
  (use "git add <file>..." to include in what will be committed)
	t/perl.core

nothing added to commit but untracked files present (use "git add" to track)

[jkeenan]

When I switch to your branch, all tests PASS -- but there's a segfault!

Upon further investigation, this appears to be IPC-System-Simple's problem, not the problem of -Utaint_support in Perl 5 blead.

On my FreeBSD machine, the BSD::Resource module had never been installed against the "system" perl, which is perl-5.32.1. So whenever I previously tested IPC::System::Simple, t/08_core.t was (correctly) being skipped. So the core dump (apparently generated for testing purposes) was never generated. Once I have installed BSD::Resource against any perl, the test is executed and a core dump is generated -- but the test program doesn't clean up after itself.

[jkeenan]

And just to be self-aggravating ... it appears the core dump is not generated on Linux, with or without taint support. Inside t/08_core.t I see the following:

# Core dumps on OS X are complicated. See GH #5.

BEGIN {
    if ($^O eq 'darwin') {
        plan skip_all => "Coredump tests skipped under OS X/Darwin";
    }
}

So maybe @pjf encountered this problem on OS X, which is a descendant of FreeBSD.

[jkeenan]

@neilb, I'm inclined to accept your pull request, but am going to hold off a bit, given that some problems with support for no-taint perls have been raised in Perl/perl5#19657.

[neilb]

@neilb, I'm inclined to accept your pull request, but am going to hold off a bit, given that some problems with support for no-taint perls have been raised in Perl/perl5#19657.

I agree that it's best to hold off until we resolve that issue. I'll close this if it comes to that.

[neilb]

Closing this, as the current implementation of the Configure question is going to be reverted.
I'll be back ;-)

[jkeenan]

Closing this, as the current implementation of the Configure question is going to be reverted. I'll be back ;-)

@neilb, please see #43, wherein @DrHyde encountered what looks to be the same problem you did last year at this time. Thoughts?