From @masak

<alksentrs> rakudo​: say 1 % 0
<p6eval> rakudo f8e959​: OUTPUT«1␤»
<masak> alksentrs​: o.O
<masak> alksentrs​: that should be a 'div by zorro' error, if you ask me.
<alksentrs> rakudo​: say 0 % 0
<p6eval> rakudo f8e959​: OUTPUT«0␤»
<colomon> $x % $y == $x - floor($x / $y) * $y
<alksentrs> rakudo​: say 0 / 0
<p6eval> rakudo f8e959​: OUTPUT«Divide by zero␤ in 'infix​:<div>' [...]
<colomon> rakudo​: say floor(1/0).perl
<p6eval> rakudo f8e959​: OUTPUT«Inf␤»
<TimToady> rakudo​: say (0/0).WHAT
<p6eval> rakudo f8e959​: OUTPUT«Divide by zero␤ in 'infix​:<div>' [...]
<colomon> rakudo​: say floor(1/0) * 0
<p6eval> rakudo f8e959​: OUTPUT«NaN␤»
<colomon> okay, so that's not how it's being calculated...
<colomon> ah, just a call to pir​::mod
* masak submits $x % 0 rakudobug
* colomon hopes masak doesn't expect us to trap LHS of 0 on infix​:<%>
before calling into Parrot...
<masak> colomon​: I expect $x % 0 to die just as much as $x / 0 does.

From @ShimmerFairy

Properly fails now​:

  <lue> r​: say 1 % 0
  <camelia> rakudo-parrot a24091​: OUTPUT«Type check failed for return value; expected 'Int' but got 'Failure'␤ in sub infix​:<%> at gen/parrot/CORE.setting​:4490␤ in sub infix​:<%> at gen/parrot/CORE.setting​:4094␤ in block at /tmp/tmpfile​:1␤␤»
  <camelia> ..rakudo-moar a24091​: OUTPUT«Divide by zero using infix​:<%>␤ in block at src/gen/m-CORE.setting​:12450␤ in sub infix​:<%> at src/gen/m-CORE.setting​:4486␤ in sub infix​:<%> at src/gen/m-CORE.setting​:4090␤ in block at /tmp/tmpfile​:1␤␤»
  <camelia> ..rakudo-jvm a24091​: OUTPUT«Type check failed for return value; expected 'Int' but got 'Failure'␤ in sub infix​:<%> at gen/jvm/CORE.setting​:4486␤ in sub infix​:<%> at gen/jvm/CORE.setting​:4090␤ in block at /tmp/tmpfile​:1␤␤»

A test for this is in ../roast/S03-operators/arith.t, as well as similar tests for %% in ../roast/S03-operators/is-divisible-by.t. Can be closed unless more tests are desired.

The RT System itself - Status changed from 'new' to 'open'

From @usev6

As it was said before, infix​:<%> now gives the same divide-by-zero errors as infix​:</>.

In roast we currently only test with 'dies_ok' and 'eval_dies_ok'. On #perl6 it was pointed out by psch++ two days ago that we don't fail directly with X​::Numeric​::DivideByZero, but with X​::TypeCheck​::Return (or X​::Typecheck​::Assignment)​:

<psch> right
<psch> lazy rats or something i think
<psch> m​: my $y = 3; my $x = do { CATCH { default { say .perl } }; ($y / 0).Int }; say $x.WHAT
<camelia> rakudo-moar a75337​: OUTPUT«X​::TypeCheck​::Return.new(operation => "returning", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => Any)), expected => Int)␤(Any)␤»
<psch> that means the $y / 0 returns something non-Int-y, which throws differently, with DivideByZero as additional payload
<psch> that seems a bit convoluted
<psch> as in, i can't imagine a case where we get *just* X​::Numeric​::DivideByZero not wrapped in some X​::TypeCheck

On a related note​: 'say 0 / 0' fails differently than the other three​:

$ perl6 -e 'my $x = do { CATCH { default { say .perl } }; ( 0 / 0 ).Int };'
X​::TypeCheck​::Assignment.new(symbol => "\$numerator", operation => "assignment", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => Any)), expected => Int)

$ perl6 -e 'my $x = do { CATCH { default { say .perl } }; ( 1 / 0 ).Int };'
X​::TypeCheck​::Return.new(operation => "returning", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => Any)), expected => Int)

$ perl6 -e 'my $x = do { CATCH { default { say .perl } }; ( 0 % 0 ).Int };'
X​::TypeCheck​::Return.new(operation => "returning", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => "infix​:<\%>")), expected => Int)

$ perl6 -e 'my $x = do { CATCH { default { say .perl } }; ( 1 % 0 ).Int };'
X​::TypeCheck​::Return.new(operation => "returning", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => "infix​:<\%>")), expected => Int)

IMHO that meets the requirements of this ticket. So I'd vote for "closable (maybe after tweaking the tests to reflect the typed exceptions)".

From @lizmat

FWIW, I think almost *all* dies_ok and eval_dies_ok should be eradicated from roast. The generate way too many false positives. Should probably be all converted to throws_like.

Liz

On 23 Nov 2014, at 23​:10, Christian Bartolomaeus via RT <bugs-comment@​bugs6.perl.org> wrote​:

As it was said before, infix​:<%> now gives the same divide-by-zero errors as infix​:</>.

In roast we currently only test with 'dies_ok' and 'eval_dies_ok'. On #perl6 it was pointed out by psch++ two days ago that we don't fail directly with X​::Numeric​::DivideByZero, but with X​::TypeCheck​::Return (or X​::Typecheck​::Assignment)​:

<psch> right
<psch> lazy rats or something i think
<psch> m​: my $y = 3; my $x = do { CATCH { default { say .perl } }; ($y / 0).Int }; say $x.WHAT
<camelia> rakudo-moar a75337​: OUTPUT«X​::TypeCheck​::Return.new(operation => "returning", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => Any)), expected => Int)␤(Any)␤»
<psch> that means the $y / 0 returns something non-Int-y, which throws differently, with DivideByZero as additional payload
<psch> that seems a bit convoluted
<psch> as in, i can't imagine a case where we get *just* X​::Numeric​::DivideByZero not wrapped in some X​::TypeCheck

On a related note​: 'say 0 / 0' fails differently than the other three​:

$ perl6 -e 'my $x = do { CATCH { default { say .perl } }; ( 0 / 0 ).Int };'
X​::TypeCheck​::Assignment.new(symbol => "\$numerator", operation => "assignment", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => Any)), expected => Int)

$ perl6 -e 'my $x = do { CATCH { default { say .perl } }; ( 1 / 0 ).Int };'
X​::TypeCheck​::Return.new(operation => "returning", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => Any)), expected => Int)

$ perl6 -e 'my $x = do { CATCH { default { say .perl } }; ( 0 % 0 ).Int };'
X​::TypeCheck​::Return.new(operation => "returning", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => "infix​:<\%>")), expected => Int)

$ perl6 -e 'my $x = do { CATCH { default { say .perl } }; ( 1 % 0 ).Int };'
X​::TypeCheck​::Return.new(operation => "returning", got => Failure.new(exception => X​::Numeric​::DivideByZero.new(using => "infix​:<\%>")), expected => Int)

IMHO that meets the requirements of this ticket. So I'd vote for "closable (maybe after tweaking the tests to reflect the typed exceptions)".

From @usev6

On Sun Nov 23 14​:32​:43 2014, elizabeth wrote​:

FWIW, I think almost *all* dies_ok and eval_dies_ok should be
eradicated from roast. The generate way too many false positives.
Should probably be all converted to throws_like.

Liz

Hah! On that note I went ahead and changed those 'dies_ok' and 'eval_dies_ok' to more specific tests -- see commit Raku/roast@d025d6c4a7

If the new tests are too specific (or if there are too many of them), please feel free to modify the tests or reopen the ticket. (Which I'm closing now.)

Christian

@usev6 - Status changed from 'open' to 'resolved'