[Bug fortran/54234] New: -Wconversion or -Wconversion-extra should warn for CMPLX(dp,dp)

[Bug fortran/54234] New: -Wconversion or -Wconversion-extra should warn for CMPLX(dp,dp)

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54234

             Bug #: 54234
           Summary: -Wconversion or -Wconversion-extra should warn for
                    CMPLX(dp,dp)
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Keywords: diagnostic
          Severity: normal
          Priority: P3
         Component: fortran
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: burnus@gcc.gnu.org


When calling CMPLX with non-default-kind arguments, the compiler should warn
about the conversion, i.e. for
  CMPLX (0.1_dp, 0.1_dp)
as the expression is complex(4) while the arguments are REALs with kind=8.

There should be no warning for  cmplx (complex number).

Longer example – compile with -Wconversion -Wconversion-extra. Again, I'd
expect a warning for the CMPLX:

module fft_mod
  implicit none
  integer,       parameter :: dp=selected_real_kind(15,300)
  real(kind=dp), parameter :: pi=3.141592653589793238460
contains
  subroutine test
    integer :: x
    x = int (abs (cmplx(2.3_dp,0.1_dp)))
  end subroutine test
end module fft_mod

[Bug fortran/54234] -Wconversion or -Wconversion-extra should warn for CMPLX(dp,dp)

[kargl at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54234

kargl at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |kargl at gcc dot gnu.org

--- Comment #1 from kargl at gcc dot gnu.org 2012-08-12 16:12:09 UTC ---
I believe that a warning is not necessary and
in fact should not be issued.  The Fortran
standard explicitly states that CMPLX returns
a complex number with default real kind if the
KIND= optional argument is not used.

[Bug fortran/54234] -Wconversion or -Wconversion-extra should warn for CMPLX(dp,dp)

[tkoenig at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54234

Thomas Koenig <tkoenig at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tkoenig at gcc dot gnu.org

--- Comment #2 from Thomas Koenig <tkoenig at gcc dot gnu.org> 2012-08-12 16:23:19 UTC ---
This is certainly a mistake made by lots of people,
so I think a warning would be appropriate.

[Bug fortran/54234] -Wconversion or -Wconversion-extra should warn for CMPLX(dp,dp)

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54234

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu.org

--- Comment #3 from Tobias Burnus <burnus at gcc dot gnu.org> 2012-08-12 19:21:41 UTC ---
(In reply to comment #1)
> The Fortran standard explicitly states that CMPLX returns
> a complex number with default real kind if the KIND= optional
> argument is not used.

Well, if it weren't valid, gfortran shouldn't warn but print an error. I
recently only found a bug in our program because of a warning about assigning
an REAL to an INTEGER, which should have been a REAL – thus, such warnings can
be helpful.


(In reply to comment #2)
> This is certainly a mistake made by lots of people,
> so I think a warning would be appropriate.

That's also the reason for suggesting this. The question is whether
-Wconversion (which is enabled by -Wall) or -Wconversion-extra (which is hidden
in the manual) is the better choice. It seems to be a somewhat common problem.
The case that one wants to have the default kind but uses a higher-precision
argument, seems to be rather rare; hence, I am inclined to warn already with
-Wconversion/-Wall.

(Side note: REAL always confuses me – for a complex number it returns the kind
of the argument - but for integer/real the default-kind.)


Untested patch – for -Wconversion-extra
--- a/gcc/fortran/check.c
+++ b/gcc/fortran/check.c
@@ -1278,6 +1278,17 @@ gfc_check_cmplx (gfc_expr *x, gfc_expr *y, gfc_expr
*kind)
   if (kind_check (kind, 2, BT_COMPLEX) == FAILURE)
     return FAILURE;

+  if (y && !kind && gfc_option.warn_conversion_extra
+      && x->ts.type == BT_REAL && x->ts.kind > gfc_default_real_kind)
+    gfc_warning_now ("Conversion from %s to default-kind COMPLEX(%d) at %L "
+                    "might loose precision, consider using the KIND argument",
+                    gfc_typename (&x->ts), gfc_default_real_kind, &x->where);
+  else if (y && !kind && gfc_option.warn_conversion_extra
+          && y->ts.type == BT_REAL && y->ts.kind > gfc_default_real_kind)
+    gfc_warning_now ("Conversion from %s to default-kind COMPLEX(%d) at %L "
+                    "might loose precision, consider using the KIND argument",
+                    gfc_typename (&y->ts), gfc_default_real_kind, &y->where);
+
   return SUCCESS;
 }

[Bug fortran/54234] -Wconversion or -Wconversion-extra should warn for CMPLX(dp,dp)

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54234

--- Comment #4 from Tobias Burnus <burnus at gcc dot gnu.org> 2012-08-14 09:35:02 UTC ---
Forgot to mention in the initial report that this issue came up at c.l.f, cf.
http://www.rhinocerus.net/forum/lang-fortran/711699-quality-fortran-code-posted-rosettacode.html
But the issue has been risen before.

Submitted patch http://gcc.gnu.org/ml/fortran/2012-08/msg00064.html

[Bug fortran/54234] -Wconversion or -Wconversion-extra should warn for CMPLX(dp,dp)

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54234

--- Comment #5 from Tobias Burnus <burnus at gcc dot gnu.org> 2012-08-14 10:22:11 UTC ---
Author: burnus
Date: Tue Aug 14 10:22:06 2012
New Revision: 190378

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=190378
Log:
2012-08-14  Tobias Burnus  <burnus@net-b.de>

        PR fortran/54234
        * check.c (gfc_check_cmplx): Add -Wconversion warning
        when converting higher-precision REAL to default-precision
        CMPLX without kind= parameter.

2012-08-14  Tobias Burnus  <burnus@net-b.de>

        PR fortran/54234
        * gfortran.dg/warn_conversion_4.f90: New.


Added:
    trunk/gcc/testsuite/gfortran.dg/warn_conversion_4.f90
Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/check.c
    trunk/gcc/testsuite/ChangeLog

[Bug fortran/54234] -Wconversion or -Wconversion-extra should warn for CMPLX(dp,dp)

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54234

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED

--- Comment #6 from Tobias Burnus <burnus at gcc dot gnu.org> 2012-08-14 10:30:42 UTC ---
FIXED on the 4.8 trunk.