Bug #111910 for Math-Units-PhysicalValue: Tests fail with German locale and perl >= 5.22.0

Wed Feb 10 03:03:47 2016 SREZIC [...] cpan.org - Ticket created

Subject:

Tests fail with German locale and perl >= 5.22.0

With a German locale (LC_ALL=de_DE.UTF-8; but other locales with a non-dot radix are probably also affected), on FreeBSD and Debian systems, with perl 5.22.0, 5.22.1, and 5.23.7: Output from '/usr/bin/make test': PERL_DL_NONLAZY=1 "/usr/perl5.22.0p/bin/perl5.22.0" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(0, 'blib/lib', 'blib/arch')" t/*.t t/01_load.t .................. ok # Test 1 got: "6500 ft" (t/05_skydiving.t at line 24) # Expected: "6,500 ft" # t/05_skydiving.t line 24 is: ok( "$dist", "6,500 ft" ); # You shouldn't really have to convert these to strings, # Test 2 got: "151,16 ft/s" (t/05_skydiving.t at line 25) # Expected: "151.16 ft/s" # t/05_skydiving.t line 25 is: ok( "$rate", "151.16 ft/s" ); # perl realizes then $rdate is an obj and "151..." is a string... # Test 3 got: "3761,82 kg*m/s" (t/05_skydiving.t at line 26) # Expected: "3,761.82 kg*m/s" # t/05_skydiving.t line 26 is: ok( "$momentum", "3,761.82 kg*m/s" ); t/05_skydiving.t ............. Failed 3/9 subtests t/10_metric.t ................ ok t/15_exponents.t ............. ok # Test 1 got: "27182818284,59 g" (t/20_longBIGfloats.t at line 16) # Expected: "27,182,818,284.59 g" # t/20_longBIGfloats.t line 16 is: ok( $s1, "27,182,818,284.59 g" ); t/20_longBIGfloats.t ......... Failed 1/1 subtests t/25_scalar_reductions.t ..... ok # Test 1 got: "8,65e5 miles" (t/30_sigfigs.t at line 17) # Expected: "8.65e5 miles" # t/30_sigfigs.t line 17 is: ok( $sun->sci( 3 ), "8.65e5 miles" ); t/30_sigfigs.t ............... Failed 1/1 subtests # Test 1 got: "1024 s^2" (t/35_powers.t at line 17) # Expected: "1,024 s^2" # t/35_powers.t line 17 is: ok( "$time", "1,024 s^2" ); t/35_powers.t ................ Failed 1/2 subtests t/40_some_standard_values.t .. ok t/45_boolean.t ............... ok t/47_compare.t ............... ok t/49_assbackwards.t .......... ok t/50_zero.t .................. ok t/55_exponents.t ............. skipped: (no reason given) t/60_plurals.t ............... ok LC_ALL=C on these systems, or other perl versions even with German locale are fine.

Wed Feb 10 11:25:07 2016 jettero [...] cpan.org - Correspondence added

Thanks for taking the time to contact me about this. I've been trying to track this down for years. When I set my locale to any of de, de_DE, de_DE.utf8, de_DE.UTF-8; and other de_* locales that exist on my machine; the output of /bin/ls is affected like I'd expect, but I still can't reproduce the error in perl. Is this something that's only an issue in very new Perls? My latest attempt to address it for the tests is to BEGIN { $ENV{LC_ALL}="C" }; but I also considered POSIX::setlocale("en_US") or something like that. How would you recommend I set up for my tests so the output is predictably LC_NUMERIC=en_US? (There's probably a CPAN Testers page for this. I should have looked there yesterday.) -Paul On Wed Feb 10 03:03:47 2016, SREZIC wrote: Show quoted text

> With a German locale (LC_ALL=de_DE.UTF-8; but other locales with a > non-dot radix are probably also affected), on FreeBSD and Debian > systems, with perl 5.22.0, 5.22.1, and 5.23.7: > > Output from '/usr/bin/make test': > > PERL_DL_NONLAZY=1 "/usr/perl5.22.0p/bin/perl5.22.0" "- > MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef > *Test::Harness::Switches; test_harness(0, 'blib/lib', 'blib/arch')" > t/*.t > t/01_load.t .................. ok > # Test 1 got: "6500 ft" (t/05_skydiving.t at line 24) > # Expected: "6,500 ft" > # t/05_skydiving.t line 24 is: ok( "$dist", "6,500 ft" ); > # You shouldn't really have to convert these to strings, > # Test 2 got: "151,16 ft/s" (t/05_skydiving.t at line 25) > # Expected: "151.16 ft/s" > # t/05_skydiving.t line 25 is: ok( "$rate", "151.16 ft/s" ); > # perl realizes then $rdate is an obj and "151..." is a string... > # Test 3 got: "3761,82 kg*m/s" (t/05_skydiving.t at line 26) > # Expected: "3,761.82 kg*m/s" > # t/05_skydiving.t line 26 is: ok( "$momentum", "3,761.82 kg*m/s" ); > t/05_skydiving.t ............. > Failed 3/9 subtests > t/10_metric.t ................ ok > t/15_exponents.t ............. ok > # Test 1 got: "27182818284,59 g" (t/20_longBIGfloats.t at line 16) > # Expected: "27,182,818,284.59 g" > # t/20_longBIGfloats.t line 16 is: ok( $s1, "27,182,818,284.59 g" > ); > t/20_longBIGfloats.t ......... > Failed 1/1 subtests > t/25_scalar_reductions.t ..... ok > # Test 1 got: "8,65e5 miles" (t/30_sigfigs.t at line 17) > # Expected: "8.65e5 miles" > # t/30_sigfigs.t line 17 is: ok( $sun->sci( 3 ), "8.65e5 miles" > ); > t/30_sigfigs.t ............... > Failed 1/1 subtests > # Test 1 got: "1024 s^2" (t/35_powers.t at line 17) > # Expected: "1,024 s^2" > # t/35_powers.t line 17 is: ok( "$time", "1,024 s^2" ); > t/35_powers.t ................ > Failed 1/2 subtests > t/40_some_standard_values.t .. ok > t/45_boolean.t ............... ok > t/47_compare.t ............... ok > t/49_assbackwards.t .......... ok > t/50_zero.t .................. ok > t/55_exponents.t ............. skipped: (no reason given) > t/60_plurals.t ............... ok > > > LC_ALL=C on these systems, or other perl versions even with German > locale are fine.

-- If riding in an airplane is flying, then riding in a boat is swimming. 116 jumps, 48.6 minutes of freefall, 92.9 freefall miles.

Wed Feb 10 11:25:07 2016 The RT System itself - Status changed from 'new' to 'open'

Wed Feb 10 11:25:12 2016 jettero [...] cpan.org - Taken

Wed Feb 10 15:53:05 2016 SREZIC [...] cpan.org - Correspondence added

On 2016-02-10 11:25:07, JETTERO wrote: Show quoted text

> Thanks for taking the time to contact me about this. > > I've been trying to track this down for years. When I set my locale > to any of de, de_DE, de_DE.utf8, de_DE.UTF-8; and other de_* locales > that exist on my machine; the output of /bin/ls is affected like I'd > expect, but I still can't reproduce the error in perl. > > Is this something that's only an issue in very new Perls?

Yes, only with perl >= 5.22.0 Show quoted text

> > My latest attempt to address it for the tests is to BEGIN { > $ENV{LC_ALL}="C" }; but I also considered POSIX::setlocale("en_US") or > something like that. > > How would you recommend I set up for my tests so the output is > predictably LC_NUMERIC=en_US? > > (There's probably a CPAN Testers page for this. I should have looked > there yesterday.)

For former perls, setting $ENV{LC_ALL} was sufficient. In newer perls, it seems that a change like this is necessary: diff --git a/t/05_skydiving.t b/t/05_skydiving.t index 4ed2068..709b60d 100644 --- a/t/05_skydiving.t +++ b/t/05_skydiving.t @@ -1,6 +1,6 @@ # vi:fdm=marker fdl=0 syntax=perl: -BEGIN { $ENV{LC_ALL} = "C" } +BEGIN { use POSIX qw(setlocale LC_ALL); setlocale(LC_ALL, "C") } use strict; use Test; (Probably similar changes are necessary for the other three failing test scripts) Sorry, I cannot say what exactly changed in perl 5.22. perl5220delta.pod lists quite a number of changes regarding to locale handling. Maybe you can ask Karl Williamson (khw at cpan dot org) or the p5p mailing list for details. Regards, Slaven Show quoted text

> > -Paul > > On Wed Feb 10 03:03:47 2016, SREZIC wrote:

> > With a German locale (LC_ALL=de_DE.UTF-8; but other locales with a > > non-dot radix are probably also affected), on FreeBSD and Debian > > systems, with perl 5.22.0, 5.22.1, and 5.23.7: > > > > Output from '/usr/bin/make test': > > > > PERL_DL_NONLAZY=1 "/usr/perl5.22.0p/bin/perl5.22.0" "- > > MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef > > *Test::Harness::Switches; test_harness(0, 'blib/lib', 'blib/arch')" > > t/*.t > > t/01_load.t .................. ok > > # Test 1 got: "6500 ft" (t/05_skydiving.t at line 24) > > # Expected: "6,500 ft" > > # t/05_skydiving.t line 24 is: ok( "$dist", "6,500 ft" > > ); > > # You shouldn't really have to convert these to strings, > > # Test 2 got: "151,16 ft/s" (t/05_skydiving.t at line 25) > > # Expected: "151.16 ft/s" > > # t/05_skydiving.t line 25 is: ok( "$rate", "151.16 ft/s" ); > > # perl realizes then $rdate is an obj and "151..." is a string... > > # Test 3 got: "3761,82 kg*m/s" (t/05_skydiving.t at line 26) > > # Expected: "3,761.82 kg*m/s" > > # t/05_skydiving.t line 26 is: ok( "$momentum", "3,761.82 kg*m/s" > > ); > > t/05_skydiving.t ............. > > Failed 3/9 subtests > > t/10_metric.t ................ ok > > t/15_exponents.t ............. ok > > # Test 1 got: "27182818284,59 g" (t/20_longBIGfloats.t at line 16) > > # Expected: "27,182,818,284.59 g" > > # t/20_longBIGfloats.t line 16 is: ok( $s1, "27,182,818,284.59 > > g" > > ); > > t/20_longBIGfloats.t ......... > > Failed 1/1 subtests > > t/25_scalar_reductions.t ..... ok > > # Test 1 got: "8,65e5 miles" (t/30_sigfigs.t at line 17) > > # Expected: "8.65e5 miles" > > # t/30_sigfigs.t line 17 is: ok( $sun->sci( 3 ), "8.65e5 miles" > > ); > > t/30_sigfigs.t ............... > > Failed 1/1 subtests > > # Test 1 got: "1024 s^2" (t/35_powers.t at line 17) > > # Expected: "1,024 s^2" > > # t/35_powers.t line 17 is: ok( "$time", "1,024 s^2" ); > > t/35_powers.t ................ > > Failed 1/2 subtests > > t/40_some_standard_values.t .. ok > > t/45_boolean.t ............... ok > > t/47_compare.t ............... ok > > t/49_assbackwards.t .......... ok > > t/50_zero.t .................. ok > > t/55_exponents.t ............. skipped: (no reason given) > > t/60_plurals.t ............... ok > > > > > > LC_ALL=C on these systems, or other perl versions even with German > > locale are fine.

Thu Feb 11 09:18:30 2016 jettero [...] cpan.org - Correspondence added

Ahh, thanks! That explains some things and I can use this commit/patch directly. Awesome.

Thu Feb 11 09:27:36 2016 jettero [...] cpan.org - Correspondence added

I can't wait to see if the tests pass on Perls 5.22+

Thu Feb 11 09:27:42 2016 jettero [...] cpan.org - Status changed from 'open' to 'resolved'