[patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[Jerry DeLisle]

[-- Attachment #1: Type: text/plain, Size: 1302 bytes --]

:ADDPATCH fortran:

Hi all,

After a lot of fussing, I think I have sorted this out.

The attached patch includes my original fix to this pr which checked for 
delimiters to decide whether to terminate reading.

In addition, this patch adjusts the conditions that allow extended reads of 
namelists.  This now requires -std=legacy.  The reason is that all these cases 
with non standard namelists such as extra data or no delimiters end up 
conflicting.  The mechanism for handling these is present, its a matter of 
choice when to allow what.

With this patch, non-delimited (no ' or ") strings work by default.  The test 
case namelist_24.f90 is modified to -std=legacy.  Two new test cases are 
provided. The test case namelist_39.f90 is derived from Toon Moene's case in 
pr33421.

I want to mention that according to the F95 standard, character strings in 
namelists are supposed to be delimited.  People should do that. (strong hint to 
anyone reading this :) ).

Regression tested on x86-64.

OK for trunk?

Regards,

Jerry

2007-09-30  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

	PR libfortran/33252
	* io/list_read.c (read_character): Use DELIM_APOSTROPHE and DELIM_QUOTE
         in check of first character in string.
	(nml_parse_qualifier): Only allow extended reads if -std=legacy
	is given.

[-- Attachment #2: pr33253.diff --]
[-- Type: text/x-patch, Size: 1066 bytes --]

Index: list_read.c
===================================================================
--- list_read.c	(revision 128892)
+++ list_read.c	(working copy)
@@ -893,7 +893,9 @@ read_character (st_parameter_dt *dtp, in
       goto get_string;
 
     default:
-      if (dtp->u.p.namelist_mode)
+      if (dtp->u.p.namelist_mode
+	  && (dtp->u.p.current_unit->flags.delim == DELIM_APOSTROPHE
+	      || dtp->u.p.current_unit->flags.delim == DELIM_QUOTE))
 	{
 	  unget_char (dtp,c);
 	  return;
@@ -1797,10 +1799,10 @@ nml_parse_qualifier (st_parameter_dt *dt
 		{
 		  memcpy (&ls[dim].start, dtp->u.p.value, sizeof(ssize_t));
 
-		  /*  If -std=f95/2003 or an array section is specified,
-		      do not allow excess data to be processed.  */
+		  /*  Do not allow excess data to be processed if this is
+		      an array section or not -std=legacy.  */
                   if (is_array_section == 1
-		      || compile_options.allow_std < GFC_STD_GNU)
+		      || compile_options.warn_std)
 		    ls[dim].end = ls[dim].start;
 		  else
 		    dtp->u.p.expanded_read = 1;

[-- Attachment #3: namelist_38.f90 --]
[-- Type: text/x-fortran, Size: 778 bytes --]

! { dg-do run }
! PR33253 namelist: reading back a string, also fixed writing with delimiters.
! Test case modified from that of the PR by
! Jerry DeLisle  <jvdelisle@gcc.gnu.org>
program main
  implicit none
  character(len=3) :: a
  namelist /foo/ a

  open(10, status="scratch", delim="quote")
  a = 'a"a'
  write(10,foo) 
  rewind 10
  a = ""
  read (10,foo) ! This gave a runtime error before the patch.
  if (a.ne.'a"a') call abort
  close (10)

  open(10, status="scratch", delim="apostrophe")
  a = "a'a"
  write(10,foo) 
  rewind 10
  a = ""
  read (10,foo)
  if (a.ne."a'a") call abort
  close (10)

  open(10, status="scratch", delim="none")
  a = "a'a"
  write(10,foo) 
  rewind 10
  a = ""
  read (10,foo)
  if (a.ne."a'a") call abort
  close (10)
end program main

[-- Attachment #4: namelist_39.f90 --]
[-- Type: text/x-fortran, Size: 705 bytes --]

! { dg-do run }
! PR33421 and PR33252 Weird quotation of namelist output of character arrays
! Test case from Toon Moene, adapted by Jerry DeLisle  <jvdelisle@gcc.gnu.org>

program test
implicit none
character(len=45) :: b(3)
namelist /nam/ b
b = 'x'
open(99, status="scratch")
write(99,'(4(a,/),a)') "&NAM", &
      " b(1)=' AAP NOOT MIES WIM ZUS JET',", &
      " b(2)='SURF.PRESSURE',", &
      " b(3)='APEKOOL',", &
      " /"
rewind(99)
read(99,nml=nam)
close(99)

if (b(1).ne." AAP NOOT MIES WIM ZUS JET                   ") call abort
if (b(2).ne."SURF.PRESSURE                                ") call abort
if (b(3).ne."APEKOOL                                      ") call abort

end program test


[-- Attachment #5: namelist_24.f90 --]
[-- Type: text/x-fortran, Size: 1748 bytes --]

!{ dg-do run }
!{ dg-options -std=legacy }
! Tests namelist read when more data is provided then specified by 
! array qualifier in list.
! Contributed by Jerry DeLisle  <jvdelisle@gcc.gnu.org>.
      program pr24459
      implicit none
      integer nd, ier, i, j
      parameter ( nd = 5 )
      character*(8) names(nd,nd)
      character*(8) names2(nd,nd)
      character*(8) names3(nd,nd)
      namelist / mynml /  names, names2, names3
      open(unit=20,status='scratch', delim='apostrophe')
      write (20, '(a)') "&MYNML"
      write (20, '(a)') "NAMES = 25*'0'"
      write (20, '(a)') "NAMES2 = 25*'0'"
      write (20, '(a)') "NAMES3 = 25*'0'"
      write (20, '(a)') "NAMES(2,2) = 'frogger'"
      write (20, '(a)') "NAMES(1,1) = 'E123' 'E456' 'D789' 'P135' 'P246'"
      write (20, '(a)') "NAMES2(1:5:2,2) = 'abcde' 'fghij' 'klmno'"
      write (20, '(a)') "NAMES3 = 'E123' 'E456' 'D789' 'P135' 'P246' '0' 'frogger'"
      write (20, '(a)') "/"
      rewind(20)
      read(20,nml=mynml, iostat=ier)
      if (ier.ne.0) call abort()
      if (any(names(:,3:5).ne."0")) call abort()
      if (names(2,2).ne."frogger") call abort()
      if (names(1,1).ne."E123") call abort()
      if (names(2,1).ne."E456") call abort()
      if (names(3,1).ne."D789") call abort()
      if (names(4,1).ne."P135") call abort()
      if (names(5,1).ne."P246") call abort()
      if (any(names2(:,1).ne."0")) call abort()
      if (any(names2(:,3:5).ne."0")) call abort()
      if (names2(1,2).ne."abcde") call abort()
      if (names2(2,2).ne."0") call abort()
      if (names2(3,2).ne."fghij") call abort()
      if (names2(4,2).ne."0") call abort()
      if (names2(5,2).ne."klmno") call abort()
      if (any(names3.ne.names)) call abort()
      end

Re: [patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[FX Coudert]

> In addition, this patch adjusts the conditions that allow extended  
> reads of namelists.  This now requires -std=legacy.  The reason is  
> that all these cases with non standard namelists such as extra data  
> or no delimiters end up conflicting.  The mechanism for handling  
> these is present, its a matter of choice when to allow what.

Short of a proper review, which I might do later this week, I want to  
note that as this requires users to turn change compiler options for  
their (broken) code, this should probably be noted in the release notes.

FX

Re: [patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[Jerry DeLisle]

Jerry DeLisle wrote:
> FX Coudert wrote:
>>> In addition, this patch adjusts the conditions that allow extended 
>>> reads of namelists.  This now requires -std=legacy.  The reason is 
>>> that all these cases with non standard namelists such as extra data 
>>> or no delimiters end up conflicting.  The mechanism for handling 
>>> these is present, its a matter of choice when to allow what.
>>
>> Short of a proper review, which I might do later this week, I want to 
>> note that as this requires users to turn change compiler options for 
>> their (broken) code, this should probably be noted in the release notes.
>>
>> FX
>>
> I agree.  Another thing that would be helpful is to see how other 
> compilers handle the test cases given.  I only have ifort.
> 
> ifort fails on namelist_15.f90 but passes on namelist_24.f90 with no 
> compiler options.
> 
> Any comment on the behavior we want would be appreciated.
> 
> Jerry
> 

Fx,

After some additional study on this one, we are able to fix everything without 
changing the require compiler options.  The following patch does it. Addition of 
one more test for the bail out logic.

Index: list_read.c
===================================================================
--- list_read.c (revision 128954)
+++ list_read.c (working copy)
@@ -893,7 +893,10 @@ read_character (st_parameter_dt *dtp, in
        goto get_string;

      default:
-      if (dtp->u.p.namelist_mode)
+      if (dtp->u.p.namelist_mode
+         && (dtp->u.p.current_unit->flags.delim == DELIM_APOSTROPHE
+             || dtp->u.p.current_unit->flags.delim == DELIM_QUOTE
+             || quote == ' '))
         {
           unget_char (dtp,c);
           return;

OK for trunk?  Test cases remain the same except no -std=legacy for namelist_24.f90

Jerry

Re: [patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[Jerry DeLisle]

Jerry DeLisle wrote:
> Jerry DeLisle wrote:
>> FX Coudert wrote:
>>>> In addition, this patch adjusts the conditions that allow extended 
>>>> reads of namelists.  This now requires -std=legacy.  The reason is 
>>>> that all these cases with non standard namelists such as extra data 
>>>> or no delimiters end up conflicting.  The mechanism for handling 
>>>> these is present, its a matter of choice when to allow what.
>>>
>>> Short of a proper review, which I might do later this week, I want to 
>>> note that as this requires users to turn change compiler options for 
>>> their (broken) code, this should probably be noted in the release notes.
>>>
>>> FX
>>>
>> I agree.  Another thing that would be helpful is to see how other 
>> compilers handle the test cases given.  I only have ifort.
>>
>> ifort fails on namelist_15.f90 but passes on namelist_24.f90 with no 
>> compiler options.
>>
>> Any comment on the behavior we want would be appreciated.
>>
>> Jerry
>>
> 
> Fx,
> 
> After some additional study on this one, we are able to fix everything 
> without changing the require compiler options.  The following patch does 
> it. Addition of one more test for the bail out logic.
> 
> Index: list_read.c
> ===================================================================
> --- list_read.c (revision 128954)
> +++ list_read.c (working copy)
> @@ -893,7 +893,10 @@ read_character (st_parameter_dt *dtp, in
>        goto get_string;
> 
>      default:
> -      if (dtp->u.p.namelist_mode)
> +      if (dtp->u.p.namelist_mode
> +         && (dtp->u.p.current_unit->flags.delim == DELIM_APOSTROPHE
> +             || dtp->u.p.current_unit->flags.delim == DELIM_QUOTE
> +             || quote == ' '))
>         {
>           unget_char (dtp,c);
>           return;
> 
> OK for trunk?  Test cases remain the same except no -std=legacy for 
> namelist_24.f90
> 
> Jerry
> 
Approved on IRC by Fx. NIST tested and regression tested on x86-64.

Will commit shortly.

Regards,

Jerry

Re: [patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[Tobias Burnus]

:REVIEWMAIL:

Jerry DeLisle wrote:
> The attached patch fixes this buglet.  We were not handling the
> interface to write_character quite right to convey the correct
> delimiting.
> OK for trunk?
Looks OK as far as I can see.

Thanks for fixing this.

Tobias

Re: [patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[Jerry DeLisle]

[-- Attachment #1: Type: text/plain, Size: 1413 bytes --]

:ADDPATCH fortran]

Tobias Burnus wrote:
> Hi Jerry,
> 
> Jerry DeLisle wrote:
>> The title of this PR is a bit misleading.  The failure occurs with a
>> string without the apostrophe as well.  The problem here is that when
>> reading a string, we look at the first character and see if it is a
>> delimiter ' or " .  If not we would skip out and not read the rest of
>> the string.
> 
> For me, it does not fully work yet. Using the following program:
> --------------------
> program main
>   implicit none
>   character(len=3) :: a
>   namelist /foo/ a
>   open(10,delim="APOSTROPHE")
>   a = "a'a"
>   write(10,foo)
>   write (*,foo)
>   rewind 10
>   read (10,foo)
> end program main
> --------------------

The attached patch fixes this buglet.  We were not handling the interface to 
write_character quite right to convey the correct delimiting.

Updated the test case to  test the three possible delimiters that can be 
specified to assure the double quotes happen when they should.

Interestingly, the write_character function had all the necessary workings to do 
this. :)

Regression tested on x86-64-gnu-linux.

OK for trunk?

Regards,

Jerry

2007-09-04  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

	PR libfortran/33253
	* io/write.c (nml_write_obj): Set the delimiter correctly before calling
	write_character. (namelist_write): Clean up the code a little and add
	comments to clarify what its doing.



[-- Attachment #2: namelist.diff --]
[-- Type: text/x-patch, Size: 2182 bytes --]

Index: write.c
===================================================================
--- write.c	(revision 128114)
+++ write.c	(working copy)
@@ -868,6 +868,7 @@ nml_write_obj (st_parameter_dt *dtp, nam
   size_t base_name_len;
   size_t base_var_name_len;
   size_t tot_len;
+  unit_delim tmp_delim;
 
   /* Write namelist variable names in upper case. If a derived type,
      nothing is output.  If a component, base and base_name are set.  */
@@ -984,11 +985,13 @@ nml_write_obj (st_parameter_dt *dtp, nam
               break;
 
 	    case GFC_DTYPE_CHARACTER:
-	      if (dtp->u.p.nml_delim)
-		write_character (dtp, &dtp->u.p.nml_delim, 1);
+	      tmp_delim = dtp->u.p.current_unit->flags.delim;
+	      if (dtp->u.p.nml_delim == '"')
+		dtp->u.p.current_unit->flags.delim = DELIM_QUOTE;
+	      if (dtp->u.p.nml_delim == '\'')
+		dtp->u.p.current_unit->flags.delim = DELIM_APOSTROPHE;
 	      write_character (dtp, p, obj->string_length);
-	      if (dtp->u.p.nml_delim)
-		write_character (dtp, &dtp->u.p.nml_delim, 1);
+	      dtp->u.p.current_unit->flags.delim = tmp_delim;
               break;
 
 	    case GFC_DTYPE_REAL:
@@ -1130,7 +1133,6 @@ namelist_write (st_parameter_dt *dtp)
   /* Set the delimiter for namelist output.  */
 
   tmp_delim = dtp->u.p.current_unit->flags.delim;
-  dtp->u.p.current_unit->flags.delim = DELIM_NONE;
   switch (tmp_delim)
     {
     case (DELIM_QUOTE):
@@ -1146,10 +1148,12 @@ namelist_write (st_parameter_dt *dtp)
       break;
     }
 
+  /* Temporarily disable namelist delimters.  */
+  dtp->u.p.current_unit->flags.delim = DELIM_NONE;
+
   write_character (dtp, "&", 1);
 
   /* Write namelist name in upper case - f95 std.  */
-
   for (i = 0 ;i < dtp->namelist_name_len ;i++ )
     {
       c = toupper (dtp->namelist_name[i]);
@@ -1165,14 +1169,14 @@ namelist_write (st_parameter_dt *dtp)
 	  t1 = nml_write_obj (dtp, t2, dummy_offset, dummy, dummy_name);
 	}
     }
+
 #ifdef HAVE_CRLF
   write_character (dtp, "  /\r\n", 5);
 #else
   write_character (dtp, "  /\n", 4);
 #endif
 
-  /* Recover the original delimiter.  */
-
+  /* Restore the original delimiter.  */
   dtp->u.p.current_unit->flags.delim = tmp_delim;
 }
 

[-- Attachment #3: namelist_38.f90 --]
[-- Type: text/x-fortran, Size: 778 bytes --]

! { dg-do run }
! PR33253 namelist: reading back a string, also fixed writing with delimiters.
! Test case modified from that of the PR by
! Jerry DeLisle  <jvdelisle@gcc.gnu.org>
program main
  implicit none
  character(len=3) :: a
  namelist /foo/ a

  open(10, status="scratch", delim="quote")
  a = 'a"a'
  write(10,foo) 
  rewind 10
  a = ""
  read (10,foo) ! This gave a runtime error before the patch.
  if (a.ne.'a"a') call abort
  close (10)

  open(10, status="scratch", delim="apostrophe")
  a = "a'a"
  write(10,foo) 
  rewind 10
  a = ""
  read (10,foo)
  if (a.ne."a'a") call abort
  close (10)

  open(10, status="scratch", delim="none")
  a = "a'a"
  write(10,foo) 
  rewind 10
  a = ""
  read (10,foo)
  if (a.ne."a'a") call abort
  close (10)
end program main

Re: [patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[Jerry DeLisle]

Tobias Burnus wrote:
> Hi Jerry,
> 
> Jerry DeLisle wrote:
>> The title of this PR is a bit misleading.  The failure occurs with a
>> string without the apostrophe as well.  The problem here is that when
>> reading a string, we look at the first character and see if it is a
>> delimiter ' or " .  If not we would skip out and not read the rest of
>> the string.
> 
> For me, it does not fully work yet. Using the following program:
> --------------------
> program main
>   implicit none
>   character(len=3) :: a
>   namelist /foo/ a
>   open(10,delim="APOSTROPHE")
>   a = "a'a"
>   write(10,foo)
>   write (*,foo)
>   rewind 10
>   read (10,foo)
> end program main
> --------------------
> 
> ifort, NAG f95, g95, openf95 write all (modulo spacing):
> 
>  &FOO
>  A       = 'a''a'
>  /
> 
> gfortran writes:
> 
> &FOO
>  A='a'a',  /
> 
> and has then problems reading this back:
> Fortran runtime error: Invalid string input in item 0
> 
> (As gfortran properly reads  A='a''a' this seems to be only a writing
> and not a reading problem.)
> 
> Tobias
> 
Ok, I will will see what I can do here.

Jerry

Re: [patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[Tobias Burnus]

Hi Jerry,

Jerry DeLisle wrote:
> The title of this PR is a bit misleading.  The failure occurs with a
> string without the apostrophe as well.  The problem here is that when
> reading a string, we look at the first character and see if it is a
> delimiter ' or " .  If not we would skip out and not read the rest of
> the string.

For me, it does not fully work yet. Using the following program:
--------------------
program main
  implicit none
  character(len=3) :: a
  namelist /foo/ a
  open(10,delim="APOSTROPHE")
  a = "a'a"
  write(10,foo)
  write (*,foo)
  rewind 10
  read (10,foo)
end program main
--------------------

ifort, NAG f95, g95, openf95 write all (modulo spacing):

 &FOO
 A       = 'a''a'
 /

gfortran writes:

&FOO
 A='a'a',  /

and has then problems reading this back:
Fortran runtime error: Invalid string input in item 0

(As gfortran properly reads  A='a''a' this seems to be only a writing
and not a reading problem.)

Tobias

Re: [patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[Steve Kargl]

On Mon, Sep 03, 2007 at 11:59:27AM -0700, Jerry DeLisle wrote:
> 
> 2007-09-03  Jerry DeLisle  <jvdelisle@gcc.gnu.org>
> 
> 	PR libfortran/33253
> 	* io/list_read.c (read_character): Use DELIM_APOSTROPHE and 
> 	DELIM_QUOTE
> 	in check of first character in string.
> 

OK.

-- 
Steve

[patch, libfortran] PR33253 namelist: reading back a string with apostrophe

[Jerry DeLisle]

[-- Attachment #1: Type: text/plain, Size: 879 bytes --]

:ADDPATCH fortran:

Hi all,

The title of this PR is a bit misleading.  The failure occurs with a string 
without the apostrophe as well.  The problem here is that when reading a string, 
we look at the first character and see if it is a delimiter ' or " .  If not we 
would skip out and not read the rest of the string.

This patch fixes this by not doing this bail out unless a specific delimiter was 
specified in the OPEN statement, otherwise, read_character proceeds to read the 
string until it finds a valid separator.

Minor tweaks to namelist_15.f90 and namelist_24.f90 are required. New test case 
attached.

Regression tested on x86-64-gnu-linux.

OK for trunk?

Regards,

Jerry

2007-09-03  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

	PR libfortran/33253
	* io/list_read.c (read_character): Use DELIM_APOSTROPHE and DELIM_QUOTE
	in check of first character in string.



[-- Attachment #2: pr33253.diff --]
[-- Type: text/x-patch, Size: 1665 bytes --]

Index: libgfortran/io/list_read.c
===================================================================
--- libgfortran/io/list_read.c	(revision 128052)
+++ libgfortran/io/list_read.c	(working copy)
@@ -887,7 +887,9 @@ read_character (st_parameter_dt *dtp, in
       goto get_string;
 
     default:
-      if (dtp->u.p.namelist_mode)
+      if (dtp->u.p.namelist_mode
+	  && (dtp->u.p.current_unit->flags.delim == DELIM_APOSTROPHE
+	      || dtp->u.p.current_unit->flags.delim == DELIM_QUOTE))
 	{
 	  unget_char (dtp,c);
 	  return;
Index: gcc/testsuite/gfortran.dg/namelist_15.f90
===================================================================
--- gcc/testsuite/gfortran.dg/namelist_15.f90	(revision 128052)
+++ gcc/testsuite/gfortran.dg/namelist_15.f90	(working copy)
@@ -20,7 +20,7 @@ program namelist_15
 
   namelist /mynml/ x
 
-  open (10, status = "scratch")
+  open (10, status = "scratch", delim='apostrophe')
   write (10, '(A)') "&MYNML"
   write (10, '(A)') " x = 3, 4, 'dd', 'ee', 'ff', 'gg',"
   write (10, '(A)') "     4, 5, 'hh', 'ii', 'jj', 'kk',"
Index: gcc/testsuite/gfortran.dg/namelist_24.f90
===================================================================
--- gcc/testsuite/gfortran.dg/namelist_24.f90	(revision 128052)
+++ gcc/testsuite/gfortran.dg/namelist_24.f90	(working copy)
@@ -11,7 +11,7 @@
       character*(8) names2(nd,nd)
       character*(8) names3(nd,nd)
       namelist / mynml /  names, names2, names3
-      open(unit=20,status='scratch')
+      open(unit=20,status='scratch', delim='apostrophe')
       write (20, '(a)') "&MYNML"
       write (20, '(a)') "NAMES = 25*'0'"
       write (20, '(a)') "NAMES2 = 25*'0'"

[-- Attachment #3: namelist_38.f90 --]
[-- Type: text/x-fortran, Size: 415 bytes --]

! { dg-do run }
! PR33253 namelist: reading back a string
! Test case modified from that of the PR by
! Jerry DeLisle  <jvdelisle@gcc.gnu.org>
program main
  implicit none
  character(len=8) :: a
  namelist /foo/ a
  open(10, status="scratch")
  a = "a'a"
  write(10,foo) 
  rewind 10
  a = ""
  read (10,foo) ! This gave a runtime error before the patch.
  if (a.ne."a'a") call abort
  close (10)
end program main