Multiple inheritance is broken in Inline::CPP. The bug wasn't detected by the module's original "11minhrt.t" test code, but as I was rewriting the code to shift from Test.pm to Test::More (taking advantage of some of Test::More's additional test features) I discovered it.
Attached is a copy of a development version of grammar/t/11minhrt.t
which illustrates the bug.
I've been walking through the code with the Perl debugger (not fun,
and so far not productive). I also ran a version of grammar.pm that
dumped a bunch of diagnostic info to the screen (that was quite
informative, found another harmless bug, but didn't help me figure out
this one). I've also looked over the output from BUILD_NOISY as well
as the .xs file that a build creates.
Unfortunately I'm not seeing the needle in the haystack... or the
forest through the trees, whichever the case may be. I even tried the
"go to bed and look at it again tomorrow" trick. :)
Here is the symptom illustrated in semi-code.
my $obj = Child->new(); # Inherits from Parent1 and Parent2, in that order.
say $obj->parent1_method() # Ok.
say $obj->parent2_method() # Bug: same return value as $obj->parent1_method().
So the method that we're trying to inherit from Parent2 is being
resolved or bound as Parent1's method. These methods do not share the
same name, so that's not an issue.
What's interesting (and probably not surprising) is that if the C++
class "Child" is created with Parent2 as the first super, and Parent1
as the second super, then the bug reverses itself:
say $obj->parent1_method() # Bug: same return value as $obj->parent2_method().
say $obj->parent2_method() # Ok.
Two files attached:
11minhrt.t (the code I'm using to tickle this bug)
another.pl (a non-Test::More version that tries to keep it as
simple as possible)
If anyone has the energy and willingness, please run 11minhrt.t and
see what happens. Some theories I might chase down would be even
better! :)
Dave
--
David Oswald
daosw...@gmail.com
another.pl
Description: Binary data
