Re: strange behavior of modulus operator

[p5pRT]

Migrated from rt.perl.org#924 (status was 'resolved')

Searchable as RT924$

[p5pRT]

From @gsar

This is important.

modulus operator as implemented by your C compiler\.  This
operator is not as well defined for negative operands\, but
it will execute faster\.

I don't know if the technical description correctly describes what
perl is doing, but it says that it should "give you direct access to
the modulus operator as implemented by your C compiler." On both my
Sun WorkShop Compilers 4.2 (SunOS5.6) and my gcc version egcs-2.91.66
(SunOS5.7) give

-1 % 2 = -1

while perl gives me

-1 % 2 = 1

Doesn't too. Perl knows both ways to skin a modulus, depending on
whether C<use integer> is in scope. Try it.

Sarathy
gsar@​activestate.com

[p5pRT]

From [Unknown Contact. See original ticket]

On Tue, 14 Mar 2000 19​:27​:51 MST, Tom Christiansen wrote​:

DB<1> use POSIX

DB<2> O UsageOnly
UsageOnly = '1'
DB<3> V POSIX
Not a GLOB reference at lib/dumpvar.pl line 408, <IN> line 3.

Looks like dumpvar doesn't know about these funny bits​:

% perl -lwe 'sub foo; print $main​::{foo}'
-1
% perl -lwe 'sub foo ($@​); print $main​::{foo}'
$@​

Is this particularly useful? It seems fragile. One mention of
the typeglob without sneaking around Perl's back, and those
properties are gone.

% perl -lwe 'sub foo ($@​); print $​::{foo}; print *foo; print $​::{foo}'
*main​::foo
*main​::foo
*main​::foo

If you disguise your intent from the compiler, you get by... once.

% perl -lwe 'sub foo; print $main​::{foo}; eval q{print *foo}; print $main​::{foo}'
-1
*main​::foo
*main​::foo

% perl -lwe 'sub foo ($@​); print $main​::{foo}; eval q{print *foo}; print $main​::{foo}'
$@​
*main​::foo
*main​::foo

And of course, no difference here​:

% perl -lwe 'sub foo ($@​); print $​::{foo}; eval q{print *foo}; print $​::{foo}'
$@​
*main​::foo
*main​::foo

Hm... I wonder whether the compiler could ever look into single-quoted,
compiled-time resolved strings on its eval q{...} stuff to learn more
about what's afoot?

--tom