[Bug libfortran/25039] New: broken formatted I/O

[Bug libfortran/25039] New: broken formatted I/O

[iwan at irs dot phy dot nrc dot ca]

I get a runtime error on the following simple program:
cat simple.f
      program junk
      read(5,'(2i9)') i1,i2
      write(6,*) i1,i2
      end
gfortran simple.f
./a.out
1,1
At line 2 of file simple.f
Fortran runtime error: Bad value during integer read

Contrary to that:
g77 simple.f
./a.out
1,1
 1 1

A similar error occurs when reading floating point numbers 
using e.g. F10.0:
cat simple1.f
      program junk
      read(5,'(2f9.0)') x1,x2
      write(6,*) x1,x2
      end
gfortran simple1.f
./a.out
1,1
At line 2 of file simple1.f
Fortran runtime error: Bad value during floating point read

g77 simple1.f
./a.out
1,1
  1.  1.


-- 
           Summary: broken formatted I/O
           Product: gcc
           Version: 4.0.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libfortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: iwan at irs dot phy dot nrc dot ca
 GCC build triplet: x86_64-unknown-linux-gnu
  GCC host triplet: x86_64-unknown-linux-gnu
GCC target triplet: x86_64-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[jvdelisle at gcc dot gnu dot org]

------- Comment #1 from jvdelisle at gcc dot gnu dot org  2005-11-25 22:00 -------
Confirmed on 4.2


-- 

jvdelisle at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |jvdelisle at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-11-25 22:00:42
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[jvdelisle at gcc dot gnu dot org]

------- Comment #2 from jvdelisle at gcc dot gnu dot org  2005-11-26 01:35 -------
-fdump-tree-original:  See below.  What is this unprintable character?  Is that
right?

MAIN__ ()
{
  int4 i2;
  int4 i1;

  {
    struct __st_parameter_dtquireY\b dt_parm.0;

    dt_parm.0.common.filename = "junk.f";
    dt_parm.0.common.line = 4;
    dt_parm.0.common.unit = 5;
    dt_parm.0.format = "(2i9)";
    dt_parm.0.format_len = 5;
    dt_parm.0.common.flags = 4096;
    _gfortran_st_read (&dt_parm.0);
    _gfortran_transfer_integer (&dt_parm.0, &i1, 4);
    _gfortran_transfer_integer (&dt_parm.0, &i2, 4);
    _gfortran_st_read_done (&dt_parm.0);
  }
  {
    struct __st_parameter_dtquireY\b dt_parm.1;

    dt_parm.1.common.filename = "junk.f";
    dt_parm.1.common.line = 5;
    dt_parm.1.common.unit = 6;
    dt_parm.1.common.flags = 128;
    _gfortran_st_write (&dt_parm.1);
    _gfortran_transfer_integer (&dt_parm.1, &i1, 4);
    _gfortran_transfer_integer (&dt_parm.1, &i2, 4);
    _gfortran_st_write_done (&dt_parm.1);
  }
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[fxcoudert at gcc dot gnu dot org]

------- Comment #3 from fxcoudert at gcc dot gnu dot org  2005-11-26 09:58 -------
(In reply to comment #2)
> What is this unprintable character?  Is that right?

I don't see an unprintable character in my i686-linux dump:

MAIN__ ()
{
  int4 i1;
  int4 i2;

  {
    struct __st_parameter_dtquire dt_parm.0;

    dt_parm.0.common.filename = "u.f";
    dt_parm.0.common.line = 2;
    dt_parm.0.common.unit = 5;
    dt_parm.0.format = "(2i9)";
    dt_parm.0.format_len = 5;
    dt_parm.0.common.flags = 4096;
    _gfortran_st_read (&dt_parm.0);
    _gfortran_transfer_integer (&dt_parm.0, &i1, 4);
    _gfortran_transfer_integer (&dt_parm.0, &i2, 4);
    _gfortran_st_read_done (&dt_parm.0);
  }
  {
    struct __st_parameter_dtquire dt_parm.1;

    dt_parm.1.common.filename = "u.f";
    dt_parm.1.common.line = 3;
    dt_parm.1.common.unit = 6;
    dt_parm.1.common.flags = 128;
    _gfortran_st_write (&dt_parm.1);
    _gfortran_transfer_integer (&dt_parm.1, &i1, 4);
    _gfortran_transfer_integer (&dt_parm.1, &i2, 4);
    _gfortran_st_write_done (&dt_parm.1);
  }
}

The dump looks just fine, this must be a library-side problem (and an old one,
probably, considering that g95 is throwing the same error).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[jvdelisle at gcc dot gnu dot org]

------- Comment #4 from jvdelisle at gcc dot gnu dot org  2005-11-26 19:10 -------
The problem with the test case is not related to the weird dump I am getting. 
Notice I also have a 'Y' in there too as well as a '\b' character.  This is
really weird.

I also noticed some namelist failures in the testsuite out of the blue.  I am
going to bootstrap from clean directory and see if it remains.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[iwan at irs dot phy dot nrc dot ca]

------- Comment #5 from iwan at irs dot phy dot nrc dot ca  2005-11-28 20:22 -------
Created an attachment (id=10358)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10358&action=view)
Fixes bug 25039

The patch is against revision 107612 on gcc trunc, 
to be applied in gcc/libgfortran/io.
It is just a demonstration how the bug can be fixed 
( changes in the prototypes of read_block() and 
read_sf() )

Reading is terminated in 
read_sf(st_parameter_dt *dtp, int *length, int comma_separator) 
when a comma is encountered 
in the input and comma_separator=1.
comma_separator=1 when reading is initiated from 
read_l(),read_decimal(),read_radix() or read_f(). 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[kargl at gcc dot gnu dot org]

------- Comment #6 from kargl at gcc dot gnu dot org  2005-11-29 00:08 -------
The error is actually correct behavior.  The comma is
not a valid character in an integer and the '(2I9)'
format is telling the compiler that you are inputting
2 integers with 9 digits each.  The comma does not
short-circuit the read statement.

I have changed the severity to "enhancement"


-- 

kargl at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[iwan at irs dot phy dot nrc dot ca]

------- Comment #7 from iwan at irs dot phy dot nrc dot ca  2005-11-29 00:35 -------
(In reply to comment #6)
I'm not sure what the "correct" behavior is. 
But it is a fact that gfortran is the first 
compiler I have seen to respond with a runtime error, 
given this format and input. This is extract of a legacy 
code dating back to the sixtees (and Fortran IV) 
that has worked on any platform/compiler combination except 
gfortran. 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[kargl at gcc dot gnu dot org]

------- Comment #8 from kargl at gcc dot gnu dot org  2005-11-29 01:55 -------
(In reply to comment #7)
> (In reply to comment #6)
> I'm not sure what the "correct" behavior is. 
> But it is a fact that gfortran is the first 
> compiler I have seen to respond with a runtime error, 
> given this format and input. This is extract of a legacy 
> code dating back to the sixtees (and Fortran IV) 
> that has worked on any platform/compiler combination except 
> gfortran. 

So, every compiler you've used is broken :-)

I'm too busy at the moment to find the actual text in the standard,
but I'll believe the technical editor of the standard when he says
commas aren't allowed in this context.
http://gcc.gnu.org/ml/fortran/2005-11/msg00776.html

Jerry has been looking at "fixing" this problem.  Your patch does
not issue an error or warning, which I believe Jerry is considering.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[jvdelisle at gcc dot gnu dot org]

------- Comment #9 from jvdelisle at gcc dot gnu dot org  2005-11-29 02:28 -------
Yes, I have a patch under review right now that fixes this.  However, I am
going to add a requirement that -std=legacy or similar be invoked, otherwise we
would be non-compliant with F95 Standard which disallows commas in numeric
input strings.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] broken formatted I/O

[jvdelisle at gcc dot gnu dot org]

------- Comment #10 from jvdelisle at gcc dot gnu dot org  2005-11-29 03:50 -------
Dang, I just noticed that Iwan had proposed a patch in comment #5 so I did my
own.

My apologies please.

Does Iwan have copyright assignment in place?

I took a different approach with my patch because I did not want to change the
read_block interface, though there is nothing wrong with the approach Iwan
took.

I am also incorporatiung some runtime warnings after Richards comments.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] comma short-circuit field width

[jvdelisle at gcc dot gnu dot org]

------- Comment #11 from jvdelisle at gcc dot gnu dot org  2005-12-09 02:53 -------
Subject: Bug 25039

Author: jvdelisle
Date: Fri Dec  9 02:53:41 2005
New Revision: 108271

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=108271
Log:
2005-12-08  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

        PR libgfortran/25039
        * io/io.h: Create a new flag sf_read_comma to control comma
        separators in numeric reads.
        * io/transfer.c (formatted_transfer_scalar): Initialize the flag.
        (read_sf): Check for commas coming in and if the flag is set,
        shortcut the read.
        * io/read.c (read_a) (read_x): Clear the flag for character reads and
        reset it after the reads.

Modified:
    trunk/libgfortran/ChangeLog
    trunk/libgfortran/io/io.h
    trunk/libgfortran/io/read.c
    trunk/libgfortran/io/transfer.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] comma short-circuit field width

[jvdelisle at gcc dot gnu dot org]

------- Comment #12 from jvdelisle at gcc dot gnu dot org  2005-12-09 02:57 -------
Subject: Bug 25039

Author: jvdelisle
Date: Fri Dec  9 02:57:13 2005
New Revision: 108272

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=108272
Log:
2005-12-08  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

        PR libgfortran/25039
        * gfortran.dg/read_comma.f: New test.

Added:
    trunk/gcc/testsuite/gfortran.dg/read_comma.f
Modified:
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] [4.1 only] comma short-circuit field width

[jvdelisle at gcc dot gnu dot org]

------- Comment #13 from jvdelisle at gcc dot gnu dot org  2005-12-16 20:37 -------
Subject: Bug 25039

Author: jvdelisle
Date: Fri Dec 16 20:37:28 2005
New Revision: 108676

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=108676
Log:
2005-12-16  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

        PR libgfortran/25039
        * io/io.h: Create a new flag sf_read_comma to control comma
        separators in numeric reads.
        * io/transfer.c (formatted_transfer_scalar): Initialize the flag.
        (read_sf): Check for commas coming in and if the flag is set,
        shortcut the read.
        * io/read.c (read_a) (read_x): Clear the flag for character reads and
        reset it after the reads.

Modified:
    branches/gcc-4_1-branch/libgfortran/ChangeLog
    branches/gcc-4_1-branch/libgfortran/io/io.h
    branches/gcc-4_1-branch/libgfortran/io/read.c
    branches/gcc-4_1-branch/libgfortran/io/transfer.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] [4.1 only] comma short-circuit field width

[jvdelisle at gcc dot gnu dot org]

------- Comment #14 from jvdelisle at gcc dot gnu dot org  2005-12-16 20:40 -------
Subject: Bug 25039

Author: jvdelisle
Date: Fri Dec 16 20:40:13 2005
New Revision: 108677

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=108677
Log:
2005-12-16  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

        PR libgfortran/25039
        * gfortran.dg/read_comma.f: New test.

Added:
    branches/gcc-4_1-branch/gcc/testsuite/gfortran.dg/read_comma.f
Modified:
    branches/gcc-4_1-branch/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] [4.1 only] comma short-circuit field width

[jvdelisle at gcc dot gnu dot org]

------- Comment #15 from jvdelisle at gcc dot gnu dot org  2005-12-19 07:20 -------
Fixed in 4.1 and 4.2


-- 

jvdelisle at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039

[Bug libfortran/25039] [4.1 only] comma short-circuit field width

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25039