[PATCH] PR fortran/69395 -- don't exceed max allowed array dimensions

[PATCH] PR fortran/69395 -- don't exceed max allowed array dimensions

[Steve Kargl]

[-- Attachment #1: Type: text/plain, Size: 470 bytes --]

The attachedi patch detects situations where the sum of
an array's rank and corank exceeds the maximum allowed
by the Standard.  Regression tested on x86_64-*-freebsd.

2018-03-14  Steven G. Kargl  <kargl@gcc.gnu.org>

	PR fortran/69395
	* decl.c (merge_array_spec): Limit the merging to maximum allowed
	dimensions, and issue error message if limit is exceeded.

2018-03-14  Steven G. Kargl  <kargl@gcc.gnu.org>

	PR fortran/69395
	* gfortran.dg/pr69395.f90

-- 
Steve

[-- Attachment #2: pr69395.diff --]
[-- Type: text/x-diff, Size: 2442 bytes --]

Index: gcc/fortran/decl.c
===================================================================
--- gcc/fortran/decl.c	(revision 258537)
+++ gcc/fortran/decl.c	(working copy)
@@ -804,7 +804,7 @@ cleanup:
 static bool
 merge_array_spec (gfc_array_spec *from, gfc_array_spec *to, bool copy)
 {
-  int i;
+  int i, j;
 
   if ((from->type == AS_ASSUMED_RANK && to->corank)
       || (to->type == AS_ASSUMED_RANK && from->corank))
@@ -822,8 +822,14 @@ merge_array_spec (gfc_array_spec *from, gfc_array_spec
 
       for (i = 0; i < to->corank; i++)
 	{
-	  to->lower[from->rank + i] = to->lower[i];
-	  to->upper[from->rank + i] = to->upper[i];
+	  /* Do not exceed the limits on lower[] and upper[].  gfortran
+	     cleans up elsewhere.  */
+	  j = from->rank + i;
+	  if (j >= GFC_MAX_DIMENSIONS)
+	    break;
+
+	  to->lower[j] = to->lower[i];
+	  to->upper[j] = to->upper[i];
 	}
       for (i = 0; i < from->rank; i++)
 	{
@@ -846,19 +852,33 @@ merge_array_spec (gfc_array_spec *from, gfc_array_spec
 
       for (i = 0; i < from->corank; i++)
 	{
+	  /* Do not exceed the limits on lower[] and upper[].  gfortran
+	     cleans up elsewhere.  */
+	  j = to->rank + i;
+	  if (j >= GFC_MAX_DIMENSIONS)
+	    break;
+
 	  if (copy)
 	    {
-	      to->lower[to->rank + i] = gfc_copy_expr (from->lower[i]);
-	      to->upper[to->rank + i] = gfc_copy_expr (from->upper[i]);
+	      to->lower[j] = gfc_copy_expr (from->lower[i]);
+	      to->upper[j] = gfc_copy_expr (from->upper[i]);
 	    }
 	  else
 	    {
-	      to->lower[to->rank + i] = from->lower[i];
-	      to->upper[to->rank + i] = from->upper[i];
+	      to->lower[j] = from->lower[i];
+	      to->upper[j] = from->upper[i];
 	    }
 	}
     }
 
+  if (to->rank + to->corank >= GFC_MAX_DIMENSIONS)
+    {
+      gfc_error ("Sum of array rank %d and corank %d at %C exceeds maximum "
+		 "allowed dimensions of %d",
+		 to->rank, to->corank, GFC_MAX_DIMENSIONS);
+      to->corank = GFC_MAX_DIMENSIONS - to->rank;
+      return false;
+    }
   return true;
 }
 
Index: gcc/testsuite/gfortran.dg/pr69395.f90
===================================================================
--- gcc/testsuite/gfortran.dg/pr69395.f90	(nonexistent)
+++ gcc/testsuite/gfortran.dg/pr69395.f90	(working copy)
@@ -0,0 +1,5 @@
+! { dg-do compile }
+! { dg-options "-fcoarray=single" }
+program p
+real, dimension(1,2,1,2,1,2,1,2), codimension[1,2,1,2,1,2,1,*] :: z  ! { dg-error "allowed dimensions" }
+end

Re: [PATCH] PR fortran/69395 -- don't exceed max allowed array dimensions

[Thomas Koenig]

Hi Steve,


> The attachedi patch detects situations where the sum of
> an array's rank and corank exceeds the maximum allowed
> by the Standard.  Regression tested on x86_64-*-freebsd.

OK for trunk.  Thanks for the patch!

Regards, Thomas

Re: [PATCH] PR fortran/69395 -- don't exceed max allowed array dimensions

[Jerry DeLisle]

On 03/14/2018 06:23 PM, Steve Kargl wrote:
> The attachedi patch detects situations where the sum of
> an array's rank and corank exceeds the maximum allowed
> by the Standard.  Regression tested on x86_64-*-freebsd.
> 
> 2018-03-14  Steven G. Kargl  <kargl@gcc.gnu.org>
> 
> 	PR fortran/69395
> 	* decl.c (merge_array_spec): Limit the merging to maximum allowed
> 	dimensions, and issue error message if limit is exceeded.
> 
> 2018-03-14  Steven G. Kargl  <kargl@gcc.gnu.org>
> 
> 	PR fortran/69395
> 	* gfortran.dg/pr69395.f90
> 

This looks OK.

Thanks,
Jerry

Re: [PATCH] PR fortran/69395 -- don't exceed max allowed array dimensions

[Steve Kargl]

On Fri, Mar 16, 2018 at 11:46:24AM +0100, Dominique d'Humières wrote:
> 
> With one of my variants I see
> 
> pr69395_3_db.f90:2:63:
> 
>     real, dimension(1,2,1,2,1,2,1,2), codimension[1,2,1,2,1,2,*] :: z
>                                                                1
> Error: Sum of array rank 8 and corank 7 at (1) exceeds maximum allowed dimensions of 15
> 
> Is this expected or should
> 
> 	  if (j >= GFC_MAX_DIMENSIONS)
> 
> be replaced with
> 
> 	  if (j > GFC_MAX_DIMENSIONS)
> 
> (twice)?
> 

Well, the above is legal, so there is probably an off-by-one
issue.  Note, that there are a few other places where
"rank + corank >= GFC_MAX_DIMENSIONS" checks are done.
I simply copied a nearby if ().

      array.c:  if (as->rank + as->corank >= GFC_MAX_DIMENSIONS)
      array.c:  if (as->rank >= GFC_MAX_DIMENSIONS)
      array.c:  if (as->rank + as->corank >= GFC_MAX_DIMENSIONS)
      check.c:  if (source->rank >= GFC_MAX_DIMENSIONS)
mine  decl.c:   if (j >= GFC_MAX_DIMENSIONS)
mine  decl.c:   if (j >= GFC_MAX_DIMENSIONS)
      decl.c:   if (to->rank + to->corank >= GFC_MAX_DIMENSIONS)

My checks are breaking out of existing for-loops and protecting
against 

    as->lower[j] = ....
    as->upper[j] = ....

where

gfortran.h:  struct gfc_expr *lower[GFC_MAX_DIMENSIONS], *upper[GFC_MAX_DIMENSIONS];

I'll take a look at the code tomorrow.


-- 
Steve

Re: [PATCH] PR fortran/69395 -- don't exceed max allowed array dimensions

[Dominique d'Humières]

Steve,

With one of my variants I see

pr69395_3_db.f90:2:63:

    real, dimension(1,2,1,2,1,2,1,2), codimension[1,2,1,2,1,2,*] :: z
                                                               1
Error: Sum of array rank 8 and corank 7 at (1) exceeds maximum allowed dimensions of 15

Is this expected or should

	  if (j >= GFC_MAX_DIMENSIONS)

be replaced with

	  if (j > GFC_MAX_DIMENSIONS)

(twice)?

Cheers,

Dominique