[patch, libfortran] PR33039 Read NAMELIST: reads wrong namelist name

[patch, libfortran] PR33039 Read NAMELIST: reads wrong namelist name

[Jerry DeLisle]

[-- Attachment #1: Type: text/plain, Size: 593 bytes --]

:ADDPATCH fortran:

Hi,

This patch fixes this by checking for the required space after a namelist name. 
  I also allow end of line characters, mostly to avoid rewriting a bunch of test 
cases.  It seems reasonable.  The space is a requirement on the user, not the 
compiler.

I will commit as obvious.  Regression tested on x86-64-Gnu-Linux.

Attached is one test case that requires updating, plus one new test case.

Regards,

Jerry

2007-08-09  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

	PR libfortran/33039
	* io/list_read.c (find_nml_name): Check for a space after a namelist
	name match.

[-- Attachment #2: pr33039.diff --]
[-- Type: text/x-patch, Size: 604 bytes --]

Index: list_read.c
===================================================================
--- list_read.c	(revision 127265)
+++ list_read.c	(working copy)
@@ -2593,6 +2593,14 @@ find_nml_name:
   if (dtp->u.p.nml_read_error)
     goto find_nml_name;
 
+  /* A trailing space is required, we give a little lattitude here, 10.9.1.  */ 
+  c = next_char (dtp);
+  if (!(c == ' ' || c == '\r' || c == '\n'))
+    {
+      unget_char (dtp, c);
+      goto find_nml_name;
+    }
+
   /* Ready to read namelist objects.  If there is an error in input
      from stdin, output the error message and continue.  */
 

[-- Attachment #3: namelist_36.f90 --]
[-- Type: text/x-fortran, Size: 1102 bytes --]

! { dg-do run }
! PR33039 Read NAMELIST:  reads wrong namelist name
! Test case from PR modified by Jerry DeLisle <jvdelisle@gcc.gnu.org>
PROGRAM namelist
CHARACTER*25 CHAR
NAMELIST /CODE/ CHAR, X
NAMELIST /CODEtwo/ X

OPEN(10, status="scratch")
write(10,'(a)') "File with test NAMELIST inputs"
write(10,'(a)') " &CODVJS  char='VJS-Not a proper nml name', X=-0.5/"
write(10,'(a)') " &CODEone char='CODEone input', X=-1.0 /"
write(10,'(a)') " &CODEtwo char='CODEtwo inputs', X=-2.0/"
write(10,'(a)') " &code    char='Lower case name',X=-3.0/"
write(10,'(a)') " &CODE    char='Desired namelist sel', X=44./"
write(10,'(a)') " &CODEx   char='Should not read CODEx nml', X=-5./"
write(10,'(a)') " $CODE    char='Second desired nml', X=66.0 /"
write(10,'(a)') " $CODE    X=77.0, char='Reordered desired nml'/"
rewind(10)
CHAR = 'Initialize string ***'
X    = -777.
READ(10, nml=CODE, END=999)
if (x.ne.-3.0) call abort
READ(10, nml=CODE, END=999)
if (x.ne.44.0) call abort
READ(10, nml=CODE, END=999)
if (x.ne.66.0) call abort
READ(10, nml=CODE, END=999)
 999 if (x.ne.77.0) call abort
END PROGRAM namelist

[-- Attachment #4: pr17286.f90 --]
[-- Type: text/x-fortran, Size: 1115 bytes --]

! { dg-do run }
! PR17286
! Namelist read failed when spaces exist between the '=' and the numbers
! This is a libgfortran bug
! Derived from testcase provided by Paul Thomas <paulthomas2@wanadoo.fr>
       program bug3
       integer num1 , num2 , num3 , num4 
       data num3  / 42 /
       data num4  / 56 /
       namelist /mynml1/ num1,num2
       namelist /mynml2/ num3,num4
       logical dbg
       data dbg / .FALSE. /
       open(unit=10,status='SCRATCH')
       write(10,'(A)') "&mynml1 ,num1= 16,num2=32,&end"
!
! write mynml2
!
       write(10,mynml2)
       rewind(10)
!
! now read back
!
       num1 = -1
       num2 = -1
       read(10,mynml1)
       if (num1.eq.16.and.num2.eq.32) then
          if (dbg) write(*,mynml1)
       else
          if (dbg) print *, 'expected 16 32 got ',num1,num2
          call abort
       endif
       num3 = -1
       num4 = -1
       read(10,mynml2)
       if (num3.eq.42.and.num4.eq.56) then
          if (dbg) write(*,mynml2)
       else
          if (dbg) print *, 'expected 42 56 got ',num3,num4
          call abort
       endif

       close(10)
       end

Re: [patch, libfortran] PR33039 Read NAMELIST: reads wrong namelist name

[Paul Thomas]

:REVIEWMAIL:

Jerry,

This one is obviously OK.

Thanks

Paul
> :ADDPATCH fortran:
>
> Hi,
>
> This patch fixes this by checking for the required space after a 
> namelist name.  I also allow end of line characters, mostly to avoid 
> rewriting a bunch of test cases.  It seems reasonable.  The space is a 
> requirement on the user, not the compiler.
>
> I will commit as obvious.  Regression tested on x86-64-Gnu-Linux.
>
> Attached is one test case that requires updating, plus one new test case.
>
> Regards,
>
> Jerry
>
> 2007-08-09  Jerry DeLisle  <jvdelisle@gcc.gnu.org>
>
>     PR libfortran/33039
>     * io/list_read.c (find_nml_name): Check for a space after a namelist
>     name match.
> ------------------------------------------------------------------------
>
> Index: list_read.c
> ===================================================================
> --- list_read.c	(revision 127265)
> +++ list_read.c	(working copy)
> @@ -2593,6 +2593,14 @@ find_nml_name:
>    if (dtp->u.p.nml_read_error)
>      goto find_nml_name;
>  
> +  /* A trailing space is required, we give a little lattitude here, 10.9.1.  */ 
> +  c = next_char (dtp);
> +  if (!(c == ' ' || c == '\r' || c == '\n'))
> +    {
> +      unget_char (dtp, c);
> +      goto find_nml_name;
> +    }
> +
>    /* Ready to read namelist objects.  If there is an error in input
>       from stdin, output the error message and continue.  */
>  
>   
> ------------------------------------------------------------------------
>
> ! { dg-do run }
> ! PR33039 Read NAMELIST:  reads wrong namelist name
> ! Test case from PR modified by Jerry DeLisle <jvdelisle@gcc.gnu.org>
> PROGRAM namelist
> CHARACTER*25 CHAR
> NAMELIST /CODE/ CHAR, X
> NAMELIST /CODEtwo/ X
>
> OPEN(10, status="scratch")
> write(10,'(a)') "File with test NAMELIST inputs"
> write(10,'(a)') " &CODVJS  char='VJS-Not a proper nml name', X=-0.5/"
> write(10,'(a)') " &CODEone char='CODEone input', X=-1.0 /"
> write(10,'(a)') " &CODEtwo char='CODEtwo inputs', X=-2.0/"
> write(10,'(a)') " &code    char='Lower case name',X=-3.0/"
> write(10,'(a)') " &CODE    char='Desired namelist sel', X=44./"
> write(10,'(a)') " &CODEx   char='Should not read CODEx nml', X=-5./"
> write(10,'(a)') " $CODE    char='Second desired nml', X=66.0 /"
> write(10,'(a)') " $CODE    X=77.0, char='Reordered desired nml'/"
> rewind(10)
> CHAR = 'Initialize string ***'
> X    = -777.
> READ(10, nml=CODE, END=999)
> if (x.ne.-3.0) call abort
> READ(10, nml=CODE, END=999)
> if (x.ne.44.0) call abort
> READ(10, nml=CODE, END=999)
> if (x.ne.66.0) call abort
> READ(10, nml=CODE, END=999)
>  999 if (x.ne.77.0) call abort
> END PROGRAM namelist
>   
> ------------------------------------------------------------------------
>
> ! { dg-do run }
> ! PR17286
> ! Namelist read failed when spaces exist between the '=' and the numbers
> ! This is a libgfortran bug
> ! Derived from testcase provided by Paul Thomas <paulthomas2@wanadoo.fr>
>        program bug3
>        integer num1 , num2 , num3 , num4 
>        data num3  / 42 /
>        data num4  / 56 /
>        namelist /mynml1/ num1,num2
>        namelist /mynml2/ num3,num4
>        logical dbg
>        data dbg / .FALSE. /
>        open(unit=10,status='SCRATCH')
>        write(10,'(A)') "&mynml1 ,num1= 16,num2=32,&end"
> !
> ! write mynml2
> !
>        write(10,mynml2)
>        rewind(10)
> !
> ! now read back
> !
>        num1 = -1
>        num2 = -1
>        read(10,mynml1)
>        if (num1.eq.16.and.num2.eq.32) then
>           if (dbg) write(*,mynml1)
>        else
>           if (dbg) print *, 'expected 16 32 got ',num1,num2
>           call abort
>        endif
>        num3 = -1
>        num4 = -1
>        read(10,mynml2)
>        if (num3.eq.42.and.num4.eq.56) then
>           if (dbg) write(*,mynml2)
>        else
>           if (dbg) print *, 'expected 42 56 got ',num3,num4
>           call abort
>        endif
>
>        close(10)
>        end
>   

Re: [patch, libfortran] PR33039 Read NAMELIST: reads wrong namelist name

[Tobias Burnus]

Jerry DeLisle wrote:
> This patch fixes this by checking for the required space after a
> namelist name.  I also allow end of line characters, mostly to avoid
> rewriting a bunch of test cases.  It seems reasonable.  The space is a
> requirement on the user, not the compiler.

I just tried using a tab character and all my compilers accepted it - I
think one could/should allow a '\t' as extension.

Tobias