[patch, fortran] Fix PR 85387

[patch, fortran] Fix PR 85387

[Thomas König]

[-- Attachment #1: Type: text/plain, Size: 521 bytes --]

Hello world,

the attached patch fixes the PR, an 8 regression caused by
trying to convert a nested implied DO loop to an array
for a case where this was not possible.

Regression-tested. OK for trunk?

Regards

	Thomas

2018-04-14  Thomas Koenig  <tkoenig@gcc.gnu.org>

	PR fortran/85387
	* frontend-passes.c (traverse_io_block): Check for start, end or
	stride being defined by an outer implied DO loop.

2018-04-14  Thomas Koenig  <tkoenig@gcc.gnu.org>

	PR fortran/85387
	* gfortran.dg/implied_do_io_5.f90: New test.

[-- Attachment #2: p2.diff --]
[-- Type: text/x-patch, Size: 823 bytes --]

Index: frontend-passes.c
===================================================================
--- frontend-passes.c	(Revision 259222)
+++ frontend-passes.c	(Arbeitskopie)
@@ -1237,6 +1237,23 @@ traverse_io_block (gfc_code *code, bool *has_reach
 	}
     }
 
+  /* Check for cases like ((a(i, j), i=1, j), j=1, 2). */
+  for (int i = 1; i < ref->u.ar.dimen; i++)
+    {
+      if (iters[i])
+	{
+	  gfc_expr *var = iters[i]->var;
+	  for (int j = i - 1; j < i; j++)
+	    {
+	      if (iters[j]
+		  && (gfc_check_dependency (var, iters[j]->start, true)
+		      || gfc_check_dependency (var, iters[j]->end, true)
+		      || gfc_check_dependency (var, iters[j]->step, true)))
+		  return false;
+	    }		  
+	}
+    }
+
   /* Create new expr.  */
   new_e = gfc_copy_expr (curr->expr1);
   new_e->expr_type = EXPR_VARIABLE;

[-- Attachment #3: implied_do_io_5.f90 --]
[-- Type: text/x-fortran, Size: 414 bytes --]

! { dg-do  run }
! { dg-additional-options "-ffrontend-optimize" }
! PR fortran/85387 - incorrect output
! Original test case by Vittorio Zecca
program main
  real :: efg_pw(2,2)
  character (len=80) :: c1, c2
  efg_pw(1,1)=1
  efg_pw(2,1)=2
  efg_pw(1,2)=3
  efg_pw(2,2)=4
  write (unit=c1,fmt='(3F12.5)') ((efg_pw(i, j), i=1, j), j=1, 2)
  write (unit=c2,fmt='(3F12.5)') 1.0, 3.0, 4.0
  if (c1 /= c2) stop 1
end

Re: [patch, fortran] Fix PR 85387

[Andre Vehreschild]

Hi Thomas,

this looks good. Ok for trunk. Thanks for the patch.

- Andre

On Sat, 14 Apr 2018 13:35:37 +0200
Thomas König <tk@tkoenig.net> wrote:

> Hello world,
> 
> the attached patch fixes the PR, an 8 regression caused by
> trying to convert a nested implied DO loop to an array
> for a case where this was not possible.
> 
> Regression-tested. OK for trunk?
> 
> Regards
> 
> 	Thomas
> 
> 2018-04-14  Thomas Koenig  <tkoenig@gcc.gnu.org>
> 
> 	PR fortran/85387
> 	* frontend-passes.c (traverse_io_block): Check for start, end or
> 	stride being defined by an outer implied DO loop.
> 
> 2018-04-14  Thomas Koenig  <tkoenig@gcc.gnu.org>
> 
> 	PR fortran/85387
> 	* gfortran.dg/implied_do_io_5.f90: New test.


-- 
Andre Vehreschild * Email: vehre ad gmx dot de 

Re: [patch, fortran] Fix PR 85387

[Thomas König]

Hi Andre,

> this looks good. Ok for trunk. Thanks for the patch.

Committed as r259384. Thanks for the quick review!

Looking at the serious regressions from the gcc home page,
we are fast approaching the gcc 8 release. Serious regressions
are below 100, and there are currently only

So, if anybody has any bugs that should urgently be fixed
for the release, please go ahead :-)

Regards

	Thomas