view release on metacpan or  search on metacpan

Stack.pod  view on Meta::CPAN

=pod

=head1 NAME

  Acme::Stack - examine a curious stack handling event.

=head1 DISCUSSION

  This module demonstrates (more simply) the behaviour that resulted in
  https://github.com/sisyphus/math-gmpz/issues/5.
  The t/test1.t script is expected to fail tests 6 and 7. This doesn't really
  matter as this module doesn't do anything useful - apart from demonstrating
  the puzzling behaviour.

  The t/test2.t script is a rewrite of t/test1.t, but is expected to pass all
  tests. The first 4 tests are exactly as written in t/test2.t and therefore
  it's no surprise that they pass.
  The remaining 3 tests in t/test2.t differ from their respective counterparts
  in t/test1.t only in that a perl scalar is used to save an intermediate
  result, before that intermediate result is passed on.
  One question is therefore "How does that avoid the test failures that occur
  with, t/test1.t ?"

  There are 3 (unexported) XSubs in this module, namely abc(), def() and ghi().
  They all do the same thing - ie they do nothing other than check that they have
  been passed 4 arguments.
  If they have been passed other than 4 arguments, then they report the
  number of arguments they've received, print out the stringification of the
  first argument, and croak.
  In the test scripts, they're always called explicitly with 4 arguments, so
  it's a bit of a surprise when, under specific conditions, 2 of them will see 5
  5 arguments.

  As best I can tell:
  When perl calls abc(), PL_markstack_ptr is explicitly incremented BEFORE the
  abc() function is actually executed;
  When perl calls def(), PL_markstack_ptr is explicitly incremented just prior
  to the termination of def();
  When perl calls ghi(); PL_markstack_ptr is not explicitly incremented at all.

  Those are the only differences between the 3 functions.
  I conclude that abc() is the one that has been correctly formulated because
  never croaks.
  But I don't really understand how both def() and ghi() manage to see 5 args.
  And I don't understand how it is that, in test1.t, def() and ghi() pass tests
  3 and 4 (respectively), yet fail tests 6 and 7 respectively.

  I guess it has something to do with being inside another separate package
  (named Number) via method calls.
  BTW, when 5 arguments are detected, the first one is a Number object - with the
  four IV args that are given explicitly being tacked on behind it.

  I'm pretty weak on understanding the finer points of stack manipulation.
  I can usually get by via trial-and-error - which is what I've done here.
  But if someone can pass on some words of explanation then maybe enough
  pennies will drop as to allow me to better comprehend the docs on this.

  Is it possible to create a simpler demo of this behaviour - eg by not having
  to involve the extra package ("package Number;") in the test scripts ?

  Is there any illegitimacy in the way that the problematic functions are being
  accessed by the perl code in the test scripts ?

 =head1 LICENSE

  This program is free software; you may redistribute it and/or modify it
  under the same terms as Perl itself.
  Copyright 20223, Sisyphus

=head1 AUTHOR

  Sisyphus <sisyphus at(@) cpan dot (.) org>

=cut