[Patch, fortran] PR65677 - Incomplete assignment on deferred-length character variable

[Patch, fortran] PR65677 - Incomplete assignment on deferred-length character variable

[Paul Richard Thomas]

[-- Attachment #1: Type: text/plain, Size: 578 bytes --]

This is yet another deferred character length problem that this time
is caused by a dependency in assignment between lhs and rhs
string_lengths. The comment in the testcase explains all.

Bootstraps and regtests on FC21/x86_64 - OK for trunk and 8-branch?

I cannot commit until next week.

Paul

2018-09-23  Paul Thomas  <pault@gcc.gnu.org>

    PR fortran/65667
    * trans-expr.c (gfc_trans_assignment_1): If there is dependency
    fix the rse stringlength.

2018-09-23  Paul Thomas  <pault@gcc.gnu.org>

    PR fortran/65667
    * gfortran.dg/dependency_52.f90 : New test.

[-- Attachment #2: submit.diff --]
[-- Type: text/x-patch, Size: 712 bytes --]

diff --git a/gcc/fortran/trans-expr.c b/gcc/fortran/trans-expr.c
index 1453828..6a05412 100644
--- a/gcc/fortran/trans-expr.c
+++ b/gcc/fortran/trans-expr.c
@@ -10187,7 +10187,11 @@ gfc_trans_assignment_1 (gfc_expr * expr1, gfc_expr * expr2, bool init_flag,
 	   || TREE_CODE (rse.string_length) == INDIRECT_REF))
     string_length = gfc_evaluate_now (rse.string_length, &rse.pre);
   else if (expr2->ts.type == BT_CHARACTER)
-    string_length = rse.string_length;
+    {
+      if (expr1->ts.deferred && gfc_check_dependency (expr1, expr2, false))
+	rse.string_length = gfc_evaluate_now (rse.string_length, &rse.pre);
+      string_length = rse.string_length;
+    }
   else
     string_length = NULL_TREE;
 

[-- Attachment #3: dependency_52.f90 --]
[-- Type: text/x-fortran, Size: 1188 bytes --]

! { dg-do run }
!
! Test the fix for PR65667, in which the dependency was missed and
! the string length of 'text' was decremented twice. The rhs string
! length is now fixed after the function call so that the dependency
! on the length of 'text' is removed for later evaluations.
!
!Contributed by John  <jwmwalrus@gmail.com>
!
module mod1
    implicit none
contains
    subroutine getKeyword(string, keyword, rest)
        character(:), allocatable, intent(IN) :: string
        character(:), allocatable, intent(OUT) :: keyword, rest
        integer :: idx
        character(:), allocatable :: text

        keyword = ''
        rest = ''
        text = string
        text = ADJUSTL(text(2:))    ! Note dependency.
        idx = INDEX(text, ' ')

        if (idx == 0) then
            keyword = TRIM(text)
        else
            keyword = text(:idx-1)
            rest = TRIM(ADJUSTL(text(idx+1:)))
        endif
    end subroutine
end module mod1

    use mod1
    implicit none

    character(:), allocatable :: line, keyword, rest

    line = '@HERE    IT IS'

    call getKeyword(line, keyword, rest)

    if (keyword .ne. 'HERE') stop 1
    if (rest .ne. 'IT IS') stop 2
end

Re: [Patch, fortran] PR65677 - Incomplete assignment on deferred-length character variable

[Janne Blomqvist]

On Sun, Sep 23, 2018 at 2:57 PM Paul Richard Thomas <
paul.richard.thomas@gmail.com> wrote:

> This is yet another deferred character length problem that this time
> is caused by a dependency in assignment between lhs and rhs
> string_lengths. The comment in the testcase explains all.
>
> Bootstraps and regtests on FC21/x86_64 - OK for trunk and 8-branch?
>

Ok, thanks.

-- 
Janne Blomqvist

Re: [Patch, fortran] PR65677 - Incomplete assignment on deferred-length character variable

[Paul Richard Thomas]

The attached fix breaks the very thing that I set out to fix in the
first place. I'm working on it.

Paul
On Mon, 1 Oct 2018 at 08:54, Paul Richard Thomas
<paul.richard.thomas@gmail.com> wrote:
>
> Hi Dominique,
>
> I seem to have got the wrong PR number right from the moment that I
> created the directory for it and kept copying it thereafter -
> apologies, I will put it right.
>
> As to the regression, the fix is simple and is attached. I had quite
> forgotten about that assertion, which I am not very convinced is
> necessary.
>
> Many thanks for alerting me to these problems.
>
> Paul
>
> On Sun, 30 Sep 2018 at 18:16, Dominique d'Humières <dominiq@lps.ens.fr> wrote:
> >
> > Hi Paul,
> >
> > First your patch has been committed with wrong entry: 65667 instead of 65677.
> >
> > Second, I suspect it is responsible of the following ICE:
> >
> > % gfc pr72755.f90 -fno-range-check
> > pr72755.f90:1485:0:
> >
> > 1485 |   this%buffer = transfer( c(start:iend), this%buffer )
> >      |
> > internal compiler error: in gfc_dep_resolver, at fortran/dependency.c:2257
> >
> > and
> >
> > % gfc pr64125_db.f90
> > pr64125_db.f90:20:0:
> >
> > 20 |   left%chars = transfer( right, left%chars )
> >    |
> > internal compiler error: in gfc_dep_resolver, at fortran/dependency.c:2257
> >
> > for
> >
> > module test
> > type t_string
> >   private
> >   character(len=:), allocatable :: chars
> > end type t_string
> >
> > contains
> >
> > pure subroutine string_assign_from_array( left, right )
> >
> > ! The target string
> >   type(t_string), intent(out) :: left
> >
> > ! The source string
> >   character, dimension(:), intent(in) :: right
> >
> >
> > ! Copy memory
> >   allocate( character(len=size(right)) :: left%chars )
> >   left%chars = transfer( right, left%chars )
> >
> > end subroutine string_assign_from_array
> >
> > end module test
> >
> > Cheers,
> >
> > Dominique
> >
>
>
> --
> "If you can't explain it simply, you don't understand it well enough"
> - Albert Einstein



-- 
"If you can't explain it simply, you don't understand it well enough"
- Albert Einstein

Re: [Patch, fortran] PR65677 - Incomplete assignment on deferred-length character variable

[Paul Richard Thomas]

[-- Attachment #1: Type: text/plain, Size: 1731 bytes --]

Hi Dominique,

I seem to have got the wrong PR number right from the moment that I
created the directory for it and kept copying it thereafter -
apologies, I will put it right.

As to the regression, the fix is simple and is attached. I had quite
forgotten about that assertion, which I am not very convinced is
necessary.

Many thanks for alerting me to these problems.

Paul

On Sun, 30 Sep 2018 at 18:16, Dominique d'Humières <dominiq@lps.ens.fr> wrote:
>
> Hi Paul,
>
> First your patch has been committed with wrong entry: 65667 instead of 65677.
>
> Second, I suspect it is responsible of the following ICE:
>
> % gfc pr72755.f90 -fno-range-check
> pr72755.f90:1485:0:
>
> 1485 |   this%buffer = transfer( c(start:iend), this%buffer )
>      |
> internal compiler error: in gfc_dep_resolver, at fortran/dependency.c:2257
>
> and
>
> % gfc pr64125_db.f90
> pr64125_db.f90:20:0:
>
> 20 |   left%chars = transfer( right, left%chars )
>    |
> internal compiler error: in gfc_dep_resolver, at fortran/dependency.c:2257
>
> for
>
> module test
> type t_string
>   private
>   character(len=:), allocatable :: chars
> end type t_string
>
> contains
>
> pure subroutine string_assign_from_array( left, right )
>
> ! The target string
>   type(t_string), intent(out) :: left
>
> ! The source string
>   character, dimension(:), intent(in) :: right
>
>
> ! Copy memory
>   allocate( character(len=size(right)) :: left%chars )
>   left%chars = transfer( right, left%chars )
>
> end subroutine string_assign_from_array
>
> end module test
>
> Cheers,
>
> Dominique
>


-- 
"If you can't explain it simply, you don't understand it well enough"
- Albert Einstein

[-- Attachment #2: regression.diff --]
[-- Type: text/x-patch, Size: 1002 bytes --]

Index: gcc/fortran/trans-expr.c
===================================================================
*** gcc/fortran/trans-expr.c	(revision 264724)
--- gcc/fortran/trans-expr.c	(working copy)
*************** gfc_trans_assignment_1 (gfc_expr * expr1
*** 10208,10214 ****
      string_length = gfc_evaluate_now (rse.string_length, &rse.pre);
    else if (expr2->ts.type == BT_CHARACTER)
      {
!       if (expr1->ts.deferred && gfc_check_dependency (expr1, expr2, false))
  	rse.string_length = gfc_evaluate_now (rse.string_length, &rse.pre);
        string_length = rse.string_length;
      }
--- 10208,10216 ----
      string_length = gfc_evaluate_now (rse.string_length, &rse.pre);
    else if (expr2->ts.type == BT_CHARACTER)
      {
!       if (expr1->ts.deferred
! 	  && rss != gfc_ss_terminator && lss != gfc_ss_terminator
! 	  && gfc_check_dependency (expr1, expr2, false))
  	rse.string_length = gfc_evaluate_now (rse.string_length, &rse.pre);
        string_length = rse.string_length;
      }

Re: [Patch, fortran] PR65677 - Incomplete assignment on deferred-length character variable

[Dominique d'Humières]

Hi Paul,

First your patch has been committed with wrong entry: 65667 instead of 65677.

Second, I suspect it is responsible of the following ICE:

% gfc pr72755.f90 -fno-range-check
pr72755.f90:1485:0:

1485 |   this%buffer = transfer( c(start:iend), this%buffer )
     | 
internal compiler error: in gfc_dep_resolver, at fortran/dependency.c:2257

and

% gfc pr64125_db.f90
pr64125_db.f90:20:0:

20 |   left%chars = transfer( right, left%chars )
   | 
internal compiler error: in gfc_dep_resolver, at fortran/dependency.c:2257

for

module test
type t_string
  private
  character(len=:), allocatable :: chars
end type t_string

contains

pure subroutine string_assign_from_array( left, right )

! The target string
  type(t_string), intent(out) :: left

! The source string
  character, dimension(:), intent(in) :: right


! Copy memory
  allocate( character(len=size(right)) :: left%chars )
  left%chars = transfer( right, left%chars )

end subroutine string_assign_from_array

end module test

Cheers,

Dominique