From @stifynsemons

Using the more elaborate version of the testcase behind #​58294, which has a
two and a three argument version of max, what use to compile and execute
incorrectly before the bug fix now compiles and segfaults. I suspect this
has nothing to do with the changes that fixed #​58294, and everything to do
with recent MMD changes. But that's speculation
I've stripped it down to the max(a,b) and max(a,b,c) definitions and one
three-arg call, though a two-arg call produces a similar result. The result
is​:

sully​:perl6 stephensimmons$ perl6 experiment/max3.p6
Null PMC access in invoke()
current instr.​: '_block11' pc 44 (EVAL_14​:19)
called from Sub 'parrot;PCT​::HLLCompiler;eval' pc 806
(src/PCT/HLLCompiler.pir​:481)
called from Sub 'parrot;PCT​::HLLCompiler;evalfiles' pc 1078
(src/PCT/HLLCompiler.pir​:610)
called from Sub 'parrot;PCT​::HLLCompiler;command_line' pc 1257
(src/PCT/HLLCompiler.pir​:699)
called from Sub 'parrot;Perl6​::Compiler;main' pc 16414 (perl6.pir​:172)
perl6(22765) malloc​: *** error for object 0x343ea10​: double free
*** set a breakpoint in malloc_error_break to debug
Segmentation fault

sysinfo​:
sully​:perl6 stephensimmons$ perl6 -v
This is Rakudo Perl 6, revision 31194 built on parrot 0.7.0-devel
for darwin-thread-multi-2level.

Copyright 2006-2008, The Perl Foundation.

sully​:perl6 stephensimmons$ uname -a
Darwin sully.local 9.4.0 Darwin Kernel Version 9.4.0​: Mon Jun 9 19​:30​:53
PDT 2008; root​:xnu-1228.5.20~1/RELEASE_I386 i386

Note that this does not segfault in r30669, though it does not handle
defined correctly.

From @stifynsemons

max3.p6

From @pmichaud

On Tue, Sep 16, 2008 at 09​:01​:06PM -0700, Stephen Simmons wrote​:

Using the more elaborate version of the testcase behind #​58294, which has a
two and a three argument version of max, what use to compile and execute
incorrectly before the bug fix now compiles and segfaults. I suspect this
has nothing to do with the changes that fixed #​58294, and everything to do
with recent MMD changes. But that's speculation
I've stripped it down to the max(a,b) and max(a,b,c) definitions and one
three-arg call, though a two-arg call produces a similar result. The result
is​:

sully​:perl6 stephensimmons$ perl6 experiment/max3.p6
Null PMC access in invoke()
current instr.​: '_block11' pc 44 (EVAL_14​:19)

I think there's something wrong with Rakudo's code generation
here; see the --target=pir output for max3.p6. The code generated
for the 2-arg and 3-arg forms of max are coming out as​:

  ## 2 argument form
  .sub "max" :multi() :lexid("25") :outer("23")
  .param pmc param_16
  .param pmc param_19

  ## 3 argument form
  .sub "max" :multi() :lexid("33") :outer("23")
  .param pmc param_101
  .param pmc param_104
  .param pmc param_107

Those two empty :multi() declarations look very suspicious to me,
but I'm not familiar enough with the new MMD implementation to know
how it's supposed to work.

Pm

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

From @pmichaud

On Wed, Sep 17, 2008 at 10​:01​:12AM -0500, Patrick R. Michaud wrote​:

On Tue, Sep 16, 2008 at 09​:01​:06PM -0700, Stephen Simmons wrote​:

Using the more elaborate version of the testcase behind #​58294, which has a
two and a three argument version of max, what use to compile and execute
incorrectly before the bug fix now compiles and segfaults. I suspect this
has nothing to do with the changes that fixed #​58294, and everything to do
with recent MMD changes. But that's speculation
I've stripped it down to the max(a,b) and max(a,b,c) definitions and one
three-arg call, though a two-arg call produces a similar result. The result
is​:

sully​:perl6 stephensimmons$ perl6 experiment/max3.p6
Null PMC access in invoke()
current instr.​: '_block11' pc 44 (EVAL_14​:19)

I think there's something wrong with Rakudo's code generation
here; see the --target=pir output for max3.p6. [...]
## 2 argument form
.sub "max" :multi() :lexid("25") :outer("23")
## 3 argument form
.sub "max" :multi() :lexid("33") :outer("23")

Those two empty :multi() declarations look very suspicious to me,
but I'm not familiar enough with the new MMD implementation to know
how it's supposed to work.

I just checked with Jonathan on IRC, and he says the empty :multi()
declarations are correct here. So that's not the problem.

Actually, I suspect the problem may be that 'max' is already a
builtin function in Perl, and that the conflict comes because
the builtins (written in PIR) are still using Parrot's MultiSub
while the ones written in Perl 6 are using the new Perl6MultiSub
implementation. If I change the function name in max3.p6 to
be something other than 'max', it appears to work​:

  $ cat max3.p6
  multi sub max3($a, $b) {
  if defined $a && defined $b {
  if $a >= $b { return $a } else { return $b }
  }
  elsif defined $a { return $a }
  elsif defined $b { return $b }
  else { return undef }
  }
 
  multi sub max3($a, $b, $c) {
  return max3(max3($a, $b), $c);
  }
 
  say "6 <" ~ max3(9,2,1) ~ ">";
  $ ./parrot perl6.pbc max3.p6
  6 <9>
  $

So, this problem should be "fixed" when we get the builtins to be
written in Perl 6 (or at least using Perl6MultiSubs).

Pm

From @jnthn

On Wed Sep 17 08​:35​:08 2008, pmichaud wrote​:

Actually, I suspect the problem may be that 'max' is already a
builtin function in Perl, and that the conflict comes because
the builtins (written in PIR) are still using Parrot's MultiSub
while the ones written in Perl 6 are using the new Perl6MultiSub
implementation. If I change the function name in max3.p6 to
be something other than 'max', it appears to work​:

$ cat max3\.p6
multi sub max3\($a, $b\) \{
        if defined $a && defined $b \{
                if $a >= $b \{ return $a \} else \{ return $b \}
        \}
        elsif defined $a \{ return $a \}
        elsif defined $b \{ return $b \}
        else \{ return undef \}
\}

multi sub max3\($a, $b, $c\) \{
        return max3\(max3\($a, $b\), $c\);
\}

say "6 \<" ~ max3\(9,2,1\) ~ ">";
$ \./parrot perl6\.pbc max3\.p6
6 \<9>
$

So, this problem should be "fixed" when we get the builtins to be
written in Perl 6 (or at least using Perl6MultiSubs).

max is now moved to the setting, and the problem is indeed cleared up.
Resolving ticket.

Thanks,

Jonathan

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

From @moritz

We have regressed on these tests, so I'm opening the ticket again.

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

From @moritz

Tests pass again, closing ticket again

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