Problem with modulo division with "use integer"

[p5pRT]

Migrated from rt.perl.org#1008 (status was 'resolved')

Searchable as RT1008$

[p5pRT]

From sbeck@cise.ufl.edu

  #!/usr/local/bin/perl -w

  $d = -1 % 7;
  print "$d\n";

  use integer;

  $d = -1 % 7;
  print "$d\n";

In all cases, it gives the output

  6
  -1

The first result is correct.

The man page says that the operator is not as well defined for negative
operands, so perhaps this can't be treated as a bug per-se, but it
would be nice if the "use integer" didn't change the behavior of the
operator.

I've included the "perl -V" info for Solaris 2.6 below, but I doubt
that it's relevant to the problem.

Summary of my perl5 (5.0 patchlevel 5 subversion 3) configuration​:
  Platform​:
  osname=solaris, osvers=2.6, archname=sun4-solaris-thread
  uname='sunos cave 5.6 generic_105181-12 sun4u sparc sunw,ultra-2 '
  hint=recommended, useposix=true, d_sigaction=define
  usethreads=define useperlio=undef d_sfio=undef
  Compiler​:
  cc='cc', optimize='-xO4', gccversion=
  cppflags='-D_REENTRANT -I/usr/local/include'
  ccflags ='-D_REENTRANT -I/usr/local/include'
  stdchar='unsigned char', d_stdstdio=define, usevfork=true
  intsize=4, longsize=4, ptrsize=4, doublesize=8
  d_longlong=define, longlongsize=8, d_longdbl=define, longdblsize=16
  alignbytes=8, usemymalloc=y, prototype=define
  Linker and Libraries​:
  ld='cc', ldflags =' -L/usr/local/lib'
  libpth=/usr/local/lib /lib /usr/lib /usr/ccs/lib
  libs=-lsocket -lnsl -ldl -lm -lpthread -lc -lcrypt
  libc=/lib/libc.so, so=so, useshrplib=true, libperl=libperl.so
  Dynamic Linking​:
  dlsrc=dl_dlopen.xs, dlext=so, d_dlsymun=undef, ccdlflags='
  -R /usr/local/lib/perl5/5.00503/sun4-solaris-thread/CORE'
  cccdlflags='-KPIC', lddlflags='-G -L/usr/local/lib'

---------------------------- Sullivan Beck -----------------------------
sbeck@​cise.ufl.edu | This space reserved for some saying demonstrating
CSE 314E | great wisdom, wit, or insight. I'll fill it in
PH : (352) 392-1057 | just as soon as I have any of the above.
Fax​: (352) 392-1220 |
------------------- http​://www.cise.ufl.edu/~sbeck/ --------------------

[p5pRT]

From @rspier

From the 'old bug patrol'....

It doesn't seem that there is a good, fast, and platform independent
solution to the fact that C leaves certain aspects of modulo with
negative numbers undefined. (Or more specifically - does allow it to
be machine dependent.)

There are several bugs in the DB related to this​:
  19990716.004
  19991022.018
  20000821.002

Here's a patch (against perl-current) which adds a warning. [below]

The warning probably isn't the best solution - but there probably
isn't a 'best' one - but this at least explains what's going on to the
user who didn't read the documentation.

There is also a fallback in pp_modulo, for UV's which ends up doing
  ans = left % right;
Should a similar warning go there too?

-R

[rspier@​speed perl-current]$ diff -pc pp.c.orig pp.c
*** pp.c.orig Sun Oct 8 19​:10​:43 2000
--- pp.c Sun Oct 8 20​:05​:55 2000
*************** PP(pp_i_modulo)
*** 1525,1530 ****
--- 1525,1533 ----
  dPOPTOPiirl;
  if (!right)
  DIE(aTHX_ "Illegal modulus zero");
+ if (ckWARN(WARN_PORTABLE) && (left < 0 || right < 0))
+ Perl_warner(aTHX_ WARN_PORTABLE,
+ "Results of %% with negative numbers and use integer is platform dependent");
  SETi( left % right );
  RETURN;
  }

[p5pRT]

From @gsar

On Sun, 08 Oct 2000 20​:14​:40 EDT, Robert Spier wrote​:

[rspier@​speed perl-current]$ diff -pc pp.c.orig pp.c
*** pp.c.orig Sun Oct 8 19​:10​:43 2000
--- pp.c Sun Oct 8 20​:05​:55 2000
*************** PP(pp_i_modulo)
*** 1525,1530 ****
--- 1525,1533 ----
dPOPTOPiirl;
if (!right)
DIE(aTHX_ "Illegal modulus zero");
+ if (ckWARN(WARN_PORTABLE) && (left < 0 || right < 0))

Putting the cheaper test will be more efficient in such cases​:

  if ((left < 0 || right < 0) && ckWARN(WARN_PORTABLE))

+ Perl_warner(aTHX_ WARN_PORTABLE,
+ "Results of %% with negative numbers and use integer is pl
atform dependent");
SETi( left % right );
RETURN;
}

Given that what it is warning about is what it has been documented to
do for a long time, I don't think I agree with the need for this patch.

Sarathy
gsar@​ActiveState.com

[p5pRT]

From @rspier

GS> Putting the cheaper test will be more efficient in such cases​:
GS> if ((left < 0 || right < 0) && ckWARN(WARN_PORTABLE))

Oops. Changed.

GS> Given that what it is warning about is what it has been documented
GS> to do for a long time, I don't think I agree with the need for
GS> this patch.

I agree with that.

But also - there have been a few bug reports on this subject, and
related to it. Yes - the problem is with the C standard (or lack
thereof, in this case), and it seems that people expect modulo to be
the same on all systems.

I think the warning is harmless - and might encourage people to "do
the proper thing" and be aware of cross platform issues. perl 5.6
added many warnings, as compared to 5.6.0.

If the warning annoys people - they should be using 'no warnings
"portable";'

If your opinion is still No, I'll close the bug and note that "It's
working as documented."

-R

[p5pRT]

From The RT System itself

Functionality is documented to work this way. The "bug" is in the definition of the C standard for %.
Closing because there seems to be little interest in fixing or adding warnings.