#!./perl -w BEGIN { chdir 't' if -d 't'; @INC = '../lib'; require './test.pl'; } use strict; use Config; # Tests of post/pre - increment/decrement operators. # Verify that addition/subtraction properly upgrade to doubles. # These tests are only significant on machines with 32 bit longs, # and two's complement negation, but shouldn't fail anywhere. my $a = 2147483647; my $c=$a++; cmp_ok($a, '==', 2147483648, "postincrement properly upgrades to double"); $a = 2147483647; $c=++$a; cmp_ok($a, '==', 2147483648, "preincrement properly upgrades to double"); $a = 2147483647; $a=$a+1; cmp_ok($a, '==', 2147483648, "addition properly upgrades to double"); $a = -2147483648; $c=$a--; cmp_ok($a, '==', -2147483649, "postdecrement properly upgrades to double"); $a = -2147483648; $c=--$a; cmp_ok($a, '==', -2147483649, "predecrement properly upgrades to double"); $a = -2147483648; $a=$a-1; cmp_ok($a, '==', -2147483649, "subtraction properly upgrades to double"); $a = 2147483648; $a = -$a; $c=$a--; cmp_ok($a, '==', -2147483649, "negation and postdecrement properly upgrade to double"); $a = 2147483648; $a = -$a; $c=--$a; cmp_ok($a, '==', -2147483649, "negation and predecrement properly upgrade to double"); $a = 2147483648; $a = -$a; $a=$a-1; cmp_ok($a, '==', -2147483649, "negation and subtraction properly upgrade to double"); $a = 2147483648; $b = -$a; $c=$b--; cmp_ok($b, '==', -$a-1, "negation, postdecrement and additional negation"); $a = 2147483648; $b = -$a; $c=--$b; cmp_ok($b, '==', -$a-1, "negation, predecrement and additional negation"); $a = 2147483648; $b = -$a; $b=$b-1; cmp_ok($b, '==', -(++$a), "negation, subtraction, preincrement and additional negation"); $a = undef; is($a++, '0', "postinc undef returns '0'"); $a = undef; is($a--, undef, "postdec undef returns undef"); # Verify that shared hash keys become unshared. sub check_same { my ($orig, $suspect) = @_; my $fail; while (my ($key, $value) = each %$suspect) { if (exists $orig->{$key}) { if ($orig->{$key} ne $value) { print "# key '$key' was '$orig->{$key}' now '$value'\n"; $fail = 1; } } else { print "# key '$key' is '$orig->{$key}', unexpect.\n"; $fail = 1; } } foreach (keys %$orig) { next if (exists $suspect->{$_}); print "# key '$_' was '$orig->{$_}' now missing\n"; $fail = 1; } ok (!$fail, "original hashes unchanged"); } my (%orig) = my (%inc) = my (%dec) = my (%postinc) = my (%postdec) = (1 => 1, ab => "ab"); my %up = (1=>2, ab => 'ac'); my %down = (1=>0, ab => -1); foreach (keys %inc) { my $ans = $up{$_}; my $up; eval {$up = ++$_}; is($up, $ans, "key '$_' incremented correctly"); is($@, '', "no error condition"); } check_same (\%orig, \%inc); foreach (keys %dec) { my $ans = $down{$_}; my $down; eval {$down = --$_}; is($down, $ans, "key '$_' decremented correctly"); is($@, '', "no error condition"); } check_same (\%orig, \%dec); foreach (keys %postinc) { my $ans = $postinc{$_}; my $up; eval {$up = $_++}; is($up, $ans, "assignment preceded postincrement"); is($@, '', "no error condition"); } check_same (\%orig, \%postinc); foreach (keys %postdec) { my $ans = $postdec{$_}; my $down; eval {$down = $_--}; is($down, $ans, "assignment preceded postdecrement"); is($@, '', "no error condition"); } check_same (\%orig, \%postdec); { no warnings 'uninitialized'; my ($x, $y); eval { $y ="$x\n"; ++$x; }; cmp_ok($x, '==', 1, "preincrement of previously uninitialized variable"); is($@, '', "no error condition"); my ($p, $q); eval { $q ="$p\n"; --$p; }; cmp_ok($p, '==', -1, "predecrement of previously uninitialized variable"); is($@, '', "no error condition"); } $a = 2147483648; $c=--$a; cmp_ok($a, '==', 2147483647, "predecrement properly downgrades from double"); $a = 2147483648; $c=$a--; cmp_ok($a, '==', 2147483647, "postdecrement properly downgrades from double"); { use integer; my $x = 0; $x++; cmp_ok($x, '==', 1, "(void) i_postinc"); $x--; cmp_ok($x, '==', 0, "(void) i_postdec"); } SKIP: { if ($Config{uselongdouble} && ($Config{longdblkind} == 6 || $Config{longdblkind} == 5)) { skip "the double-double format is weird", 1; } # I'm sure that there's an IBM format with a 48 bit mantissa # IEEE doubles have a 53 bit mantissa # 80 bit long doubles have a 64 bit mantissa # sparcs have a 112 bit mantissa for their long doubles. Just to be awkward :-) my $h_uv_max = 1 + (~0 >> 1); my $found; for my $n (47..113) { my $power_of_2 = 2**$n; my $plus_1 = $power_of_2 + 1; next if $plus_1 != $power_of_2; my ($start_p, $start_n); if ($h_uv_max > $power_of_2 / 2) { my $uv_max = 1 + 2 * (~0 >> 1); # UV_MAX is 2**$something - 1, so subtract 1 to get the start value $start_p = $uv_max - 1; # whereas IV_MIN is -(2**$something), so subtract 2 $start_n = -$h_uv_max + 2; print "# Mantissa overflows at 2**$n ($power_of_2)\n"; print "# But max UV ($uv_max) is greater so testing that\n"; } else { print "# Testing 2**$n ($power_of_2) which overflows the mantissa\n"; $start_p = int($power_of_2 - 2); $start_n = -$start_p; my $check = $power_of_2 - 2; die "Something wrong with our rounding assumptions: $check vs $start_p" unless $start_p == $check; } foreach ([$start_p, '++$i', 'pre-inc', 'inc'], [$start_p, '$i++', 'post-inc', 'inc'], [$start_n, '--$i', 'pre-dec', 'dec'], [$start_n, '$i--', 'post-dec', 'dec']) { my ($start, $action, $description, $act) = @$_; my $code = eval << "EOC" or die $@; sub { no warnings 'imprecision'; my \$i = \$start; for(0 .. 3) { my \$a = $action; } } EOC warning_is($code, undef, "$description under no warnings 'imprecision'"); $code = eval << "EOC" or die $@; sub { use warnings 'imprecision'; my \$i = \$start; for(0 .. 3) { my \$a = $action; } } EOC warnings_like($code, [(qr/Lost precision when ${act}rementing -?\d+/) x 2], "$description under use warnings 'imprecision'"); } $found = 1; last; } ok($found, "found a NV value which overflows the mantissa"); } # SKIP # these will segfault if they fail sub PVBM () { 'foo' } { my $dummy = index 'foo', PVBM } isnt(scalar eval { my $pvbm = PVBM; $pvbm++ }, undef, "postincrement defined"); isnt(scalar eval { my $pvbm = PVBM; $pvbm-- }, undef, "postdecrement defined"); isnt(scalar eval { my $pvbm = PVBM; ++$pvbm }, undef, "preincrement defined"); isnt(scalar eval { my $pvbm = PVBM; --$pvbm }, undef, "predecrement defined"); # #9466 # don't use pad TARG when the thing you're copying is a ref, or the referent # won't get freed. { package P9466; my $x; sub DESTROY { $x = 1 } for (0..1) { $x = 0; my $a = bless {}; my $b = $_ ? $a++ : $a--; undef $a; undef $b; ::is($x, 1, "9466 case $_"); } } # *Do* use pad TARG if it is actually a named variable, even when the thing # you’re copying is a ref. The fix for #9466 broke this. { package P9466_2; my $x; sub DESTROY { $x = 1 } for (2..3) { $x = 0; my $a = bless {}; my $b; use integer; if ($_ == 2) { $b = $a--; # sassign optimised away } else { $b = $a++; } ::is(ref $b, __PACKAGE__, 'i_post(in|de)c/TARGMY on ref'); undef $a; undef $b; ::is($x, 1, "9466 case $_"); } } $_ = ${qr //}; $_--; is($_, -1, 'regexp--'); { no warnings 'numeric'; $_ = ${qr //}; $_++; is($_, 1, 'regexp++'); } if ($::IS_EBCDIC) { $_ = v129; $_++; isnt(ref\$_, 'VSTRING', '++ flattens vstrings'); } else { $_ = v97; $_++; isnt(ref\$_, 'VSTRING', '++ flattens vstrings'); } sub TIESCALAR {bless\my $x} sub STORE { ++$store::called } tie my $t, ""; { $t = $_++; $t = $_--; use integer; $t = $_++; $t = $_--; } is $store::called, 4, 'STORE called on "my" target'; { # Temporarily broken between before 5.6.0 (b162f9ea/21f5b33c) and # between 5.21.5 and 5.21.6 (9e319cc4fd) my $x = 7; $x = $x++; is $x, 7, '$lex = $lex++'; $x = 7; # broken in b162f9ea (5.6.0); fixed in 5.21.6 use integer; $x = $x++; is $x, 7, '$lex = $lex++ under use integer'; } done_testing();