[Bug libfortran/59836] New: Wrong outputs with rounding formats for some values.

[Bug libfortran/59836] New: Wrong outputs with rounding formats for some values.

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

            Bug ID: 59836
           Summary: Wrong outputs with rounding formats for some values.
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libfortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dominiq at lps dot ens.fr
                CC: jvdelisle at gcc dot gnu.org

This is a split off pr59774 comment 9. The code

 print "(ru,g45.3)", 891.1
 print "(rd,g45.3)", -891.1
end

compiled with trunk (r206559) gives the output

                                       9.    
                                      -9.    

I have also found that the code

print '(RU,F2.0)', 0.6
print '(RD,F3.0)', -0.6
end

gives

7.
-7.

Both problems occur because the Fw.d format is not properly handled when d==0.
The following patch fixes both problems

--- ../_clean/libgfortran/io/write_float.def    2014-01-04 15:51:53.000000000
+0100
+++ libgfortran/io/write_float.def    2014-01-15 22:22:17.000000000 +0100
@@ -373,7 +373,7 @@ output_float (st_parameter_dt *dtp, cons
   updown:

   rchar = '0';
-  if (w > 0 && d == 0 && p == 0)
+  if (ft != FMT_F && nbefore == 0 && w > 0 && d == 0 && p == 0)
     nbefore = 1;
   /* Scan for trailing zeros to see if we really need to round it.  */
   for(i = nbefore + nafter; i < ndigits; i++)
@@ -386,13 +386,14 @@ output_float (st_parameter_dt *dtp, cons
   do_rnd:

   if (nbefore + nafter == 0)
+    /* Handle the case Fw.0 and value < 1.0 */
     {
       ndigits = 0;
       if (nzero_real == d && digits[0] >= rchar)
     {
       /* We rounded to zero but shouldn't have */
-      nzero--;
-      nafter = 1;
+      nbefore = 1;
+          digits--;
       digits[0] = '1';
       ndigits = 1;
     }

The first patch handles the first problem by restricting the test to the E*
formats and the second one fixes the way Fw.0 handles values < 1.0 (yes, I have
seen the formatting issue for the line digits--. I'll fix it when I'll submit
the patch).
AFAICT the line nzero--; does not seem necessary, but I don't know why.

With the patch the first test gives

                                     892.    
                                    -892.    

and the second

1.
-1.

and I did not see any regression when retesting with it.

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-01-16
      Known to work|                            |4.6.4
            Summary|Wrong outputs with rounding |[4.7/4.8/4.9 Regression]
                   |formats for some values.    |Wrong outputs with rounding
                   |                            |formats for some values.
     Ever confirmed|0                           |1
      Known to fail|                            |4.7.3, 4.8.2, 4.9.0

--- Comment #2 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Gfortran 4.6.4 gives the right answer for both codes. So it is a regression.

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #3 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
> I will also test your patches and review and commit for you if you need. 
> Please let me know. 

Could you check in particular that I did not miss something when removing the
line

-      nzero--;

I have not followed in detail the use of nzero.

I don't have write access so I cannot do the commit, but I can do the packaging
and the submit if it helps (I have the FSF form signed). Feel free to use the
posted patches as you like.

> I knew you were close to the solution on this one. Great job!

Thanks! Actually I have tried

-  if (w > 0 && d == 0 && p == 0)
+  if (ft != FMT_F && nbefore == 0 && w > 0 && d == 0 && p == 0)

but gfortran.dg/round_3.f08 regressed and I did not catch the need for the
second part of the patch, hence the kludge in pr59774 comment 9.

I think the formats and values in comment 0, can be added to
gfortran.dg/round_3.f08
(as well as some tests for pr59774). I'ld like also to add some tests for RC
around 0.5.

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #4 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Any comment about the following patch for testing the fix of this PR?

--- ../_clean/gcc/testsuite/gfortran.dg/round_3.f08    2011-05-05
03:54:21.000000000 +0200
+++ gcc/testsuite/gfortran.dg/round_3.f08    2014-01-18 11:17:21.000000000
+0100
@@ -16,8 +16,19 @@ program pr48615
     call checkfmt("(RU,1P,G6.0E2)", 2.0,  "2.E+00")
     call checkfmt("(RU,1P,G10.4E2)", 2.3456e5,  "2.3456E+05")

+    call checkfmt("(RU,G9.3)",  891.1,  " 892.")    ! pr59836
+    call checkfmt("(RD,G9.3)", -891.1,  "-892.")    ! pr59836
+
     call checkfmt("(RU,F2.0)",     0.09,  "1.")     ! 0.
     call checkfmt("(RD,F3.0)",     -0.09,  "-1.")     ! -0.
+    call checkfmt("(RU,F2.0)",     0.9,  "1.")       ! pr59836
+    call checkfmt("(RC,F2.0)",     0.4,  "0.")       ! pr59836
+    call checkfmt("(RC,F2.0)",     0.5,  "1.")       ! pr59836
+    call checkfmt("(RC,F2.0)",     0.6,  "1.")       ! pr59836
+    call checkfmt("(RD,F3.0)",     -0.9,  "-1.")     ! pr59836
+    call checkfmt("(RC,F3.0)",     -0.4,  "-0.")     ! pr59836
+    call checkfmt("(RC,F3.0)",     -0.5,  "-1.")     ! pr59836
+    call checkfmt("(RC,F3.0)",     -0.6,  "-1.")     ! pr59836
     call checkfmt("(RU,F2.0)",      2.0,  "2.")     ! 3.
     call checkfmt("(RD,F3.0)",     -2.0,  "-2.")     ! -3.
     call checkfmt("(RU,F6.4)",      2.0,  "2.0000")     ! 2.0001

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

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

--- Comment #5 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
The line

  if (w > 0 && d == 0 && p == 0)

has been introduced at r173408 and obviously the first code in comment 0 has
started to give the wrong output after it.

The second code in comment 0 has started to give the wrong output after r185433
(so far I did not try to understand why).

Note that the block

   if (nbefore + nafter == 0)
     {
       ndigits = 0;
       if (nzero_real == d && digits[0] >= rchar)
     {
       /* We rounded to zero but shouldn't have */
      nzero--;
      nafter = 1;
       digits[0] = '1';
       ndigits = 1;
     }

is present in the first version of libgfortran/io/write_float.def (r128114)
with char replaced with '5' (the change was introduced at r152263).

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[jvdelisle at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #6 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
Dominiq,

Well, obviously we have a lot of history on this code and many changes to get
where we are now. Very much appreciate your analysis.  A fresh look is always
helpful.  If you could prepare a changelog.  I will get ready to commit these
fixes. We probably should post to the main list and the patches list per normal
protocol.

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #7 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
I'll do the packaging over the week-end.

BTW this PR is a duplicate of what is left in pr48787 (part of it was already
analyzed in pr48787 comment 24).

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |thenlich at users dot sourceforge.
                   |                            |net

--- Comment #8 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
*** Bug 48787 has been marked as a duplicate of this bug. ***

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[jvdelisle at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #9 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
Author: jvdelisle
Date: Sun Jan 19 23:17:43 2014
New Revision: 206785

URL: http://gcc.gnu.org/viewcvs?rev=206785&root=gcc&view=rev
Log:
2014-01-19  Jerry DeLisle  <jvdelisle@gcc.gnu>
        Dominique d'Humieres  <dominiq@lps.ens.fr>

    PR libfortran/59771
    PR libfortran/59774
    PR libfortran/59836
    * io/write_float.def (output_float): Fix wrong handling of the
    Fw.0 format.
    (output_float_FMT_G_): Fixes rounding issues with -m32.

Modified:
    trunk/libgfortran/ChangeLog
    trunk/libgfortran/io/write_float.def

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[jvdelisle at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #10 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
Author: jvdelisle
Date: Sun Jan 19 23:21:10 2014
New Revision: 206786

URL: http://gcc.gnu.org/viewcvs?rev=206786&root=gcc&view=rev
Log:
2014-01-19  Steven G. Kargl  <kargl@gcc.gnu.org>

    PR libfortran/59771
    PR libfortran/59774
    PR libfortran/59836
    * gfortran.dg/round_3.f08: New cases added.
    * gfortran.dg/fmt_g_1.f90: New test.

Added:
    trunk/gcc/testsuite/gfortran.dg/fmt_g_1.f90
Modified:
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/gfortran.dg/round_3.f08

[Bug libfortran/59836] [4.7/4.8/4.9 Regression] Wrong outputs with rounding formats for some values.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P4
   Target Milestone|---                         |4.7.4

[Bug libfortran/59836] [4.7/4.8 Regression] Wrong outputs with rounding formats for some values.

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.7/4.8/4.9 Regression]    |[4.7/4.8 Regression] Wrong
                   |Wrong outputs with rounding |outputs with rounding
                   |formats for some values.    |formats for some values.

--- Comment #11 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Fixed on trunk (4.9).

[Bug libfortran/59836] [4.7/4.8 Regression] Wrong outputs with rounding formats for some values.

[jvdelisle at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #12 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
Author: jvdelisle
Date: Sat Feb 15 15:48:48 2014
New Revision: 207801

URL: http://gcc.gnu.org/viewcvs?rev=207801&root=gcc&view=rev
Log:
2014-02-15  Jerry DeLisle  <jvdelisle@gcc.gnu>
        Dominique d'Humieres  <dominiq@lps.ens.fr>

    Backport from mainline
    PR libfortran/59771
    PR libfortran/59774
    PR libfortran/59836
    * io/write_float.def (output_float): Fix wrong handling of the
    Fw.0 format.
    (output_float_FMT_G_): Fixes rounding issues with -m32.

Modified:
    branches/gcc-4_8-branch/libgfortran/ChangeLog
    branches/gcc-4_8-branch/libgfortran/io/write_float.def

[Bug libfortran/59836] [4.7/4.8 Regression] Wrong outputs with rounding formats for some values.

[jvdelisle at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #13 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
Author: jvdelisle
Date: Sat Feb 15 15:57:35 2014
New Revision: 207802

URL: http://gcc.gnu.org/viewcvs?rev=207802&root=gcc&view=rev
Log:
2014-02-15  Jerry DeLisle  <jvdelisle@gcc.gnu>
        Dominique d'Humieres  <dominiq@lps.ens.fr>

    Backport from mainline
    PR libfortran/59771
    PR libfortran/59774
    PR libfortran/59836
    * gfortran.dg/fmt_g_1.f90: New test.
    * gfortran.dg/round_3.f08: New cases added.

Added:
    branches/gcc-4_8-branch/gcc/testsuite/gfortran.dg/fmt_g_1.f90
Modified:
    branches/gcc-4_8-branch/gcc/testsuite/ChangeLog
    branches/gcc-4_8-branch/gcc/testsuite/gfortran.dg/round_3.f08

[Bug libfortran/59836] [4.7/4.8 Regression] Wrong outputs with rounding formats for some values.

[jvdelisle at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #14 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
Author: jvdelisle
Date: Sat Feb 15 16:53:07 2014
New Revision: 207803

URL: http://gcc.gnu.org/viewcvs?rev=207803&root=gcc&view=rev
Log:
2014-02-15  Jerry DeLisle  <jvdelisle@gcc.gnu>
        Dominique d'Humieres  <dominiq@lps.ens.fr>

    Backport from mainline
    PR libfortran/59771
    PR libfortran/59774
    PR libfortran/59836
    * io/write_float.def (output_float): Fix wrong handling of the
    Fw.0 format.
    (output_float_FMT_G_): Fixes rounding issues with -m32.

Modified:
    branches/gcc-4_7-branch/libgfortran/ChangeLog
    branches/gcc-4_7-branch/libgfortran/io/write_float.def

[Bug libfortran/59836] [4.7/4.8 Regression] Wrong outputs with rounding formats for some values.

[jvdelisle at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

--- Comment #15 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
Author: jvdelisle
Date: Sat Feb 15 16:55:19 2014
New Revision: 207804

URL: http://gcc.gnu.org/viewcvs?rev=207804&root=gcc&view=rev
Log:
2014-02-15  Jerry DeLisle  <jvdelisle@gcc.gnu>
        Dominique d'Humieres  <dominiq@lps.ens.fr>

    Backport from mainline
    PR libfortran/59771
    PR libfortran/59774
    PR libfortran/59836
    * gfortran.dg/fmt_g_1.f90: New test.
    * gfortran.dg/round_3.f08: New cases added.

Added:
    branches/gcc-4_7-branch/gcc/testsuite/gfortran.dg/fmt_g_1.f90
Modified:
    branches/gcc-4_7-branch/gcc/testsuite/ChangeLog
    branches/gcc-4_7-branch/gcc/testsuite/gfortran.dg/round_3.f08

[Bug libfortran/59836] [4.7/4.8 Regression] Wrong outputs with rounding formats for some values.

[jvdelisle at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59836

Jerry DeLisle <jvdelisle at gcc dot gnu.org> changed:

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

--- Comment #16 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
Fixed on 4.8 and 4.7, closing