Fortran : ICE for division by zero in declaration PR95882

Message ID fbff9340-1b2f-a020-c191-92a295fd7343@codethink.co.uk
State New
Headers show
Series
  • Fortran : ICE for division by zero in declaration PR95882
Related show

Commit Message

Mark Eggleston Aug. 24, 2020, 7:03 a.m.
Please find attached a fix for PR95882.

Tested on x86_64 with bootstrap.

OK to commit and backport?

Fortran  : ICE for division by zero in declaration PR95882

A length expression containing a divide by zero in a character
declaration will result in an ICE if the constant is anymore
complicated that a contant divided by a constant.

The cause was that char_len_param_value can return MATCH_YES
even if a divide by zero was seen.  Prior to returning check
whether a divide by zero was seen and if so set it to MATCH_ERROR.

2020-08-24  Mark Eggleston <markeggleston@gcc.gnu.org>

gcc/fortran

     PR fortran/95882
     * decl.c (char_len_param_value): Check gfc_seen_div0 and
     if it is set return MATCH_ERROR.

2020-08-24  Mark Eggleston <markeggleston@gcc.gnu.org>

gcc/testsuite/

     PR fortran/95882
     * gfortran.dg/pr95882_1.f90: New test.
     * gfortran.dg/pr95882_2.f90: New test.
     * gfortran.dg/pr95882_3.f90: New test.
     * gfortran.dg/pr95882_4.f90: New test.
     * gfortran.dg/pr95882_5.f90: New test.

-- 
https://www.codethink.co.uk/privacy.html

Comments

Richard Sandiford via Gcc-patches Aug. 24, 2020, 4:42 p.m. | #1
Hi Mark,

> OK to commit and backport?


The test cases mentioned in the ChangeLog are not in the
patch, instead there is the test case for PR 96624.

Could you correct that?

Best regards

	Thomas
Mark Eggleston Aug. 25, 2020, 6:13 a.m. | #2
On 24/08/2020 17:42, Thomas Koenig wrote:
> Hi Mark,

>

>> OK to commit and backport?

>

> The test cases mentioned in the ChangeLog are not in the

> patch, instead there is the test case for PR 96624.

>

> Could you correct that?

Whoops, yes I'll fix that.
>

> Best regards

>

>     Thomas

>

-- 
https://www.codethink.co.uk/privacy.html
Mark Eggleston Aug. 25, 2020, 6:29 a.m. | #3
On 25/08/2020 07:13, Mark Eggleston wrote:
>

> On 24/08/2020 17:42, Thomas Koenig wrote:

>> Hi Mark,

>>

>>> OK to commit and backport?

>>

>> The test cases mentioned in the ChangeLog are not in the

>> patch, instead there is the test case for PR 96624.

>>

>> Could you correct that?

> Whoops, yes I'll fix that.

It is actually the wrong attachment, I'll try again.
>>

>> Best regards

>>

>>     Thomas

>>

-- 
https://www.codethink.co.uk/privacy.html

Patch

From ab94bb744a7d64751f6b93cc56ad3ed5fe5cfc81 Mon Sep 17 00:00:00 2001
From: Mark Eggleston <markeggleston@gcc.gnu.org>
Date: Mon, 17 Aug 2020 13:50:28 +0100
Subject: [PATCH] Fortran  : Runtime error, reshape constant array assignment
 PR96624

When assigning a reshaped constant array of shape [2,0] to a
variable fails with an invalid memory access.  If a varibale
with the parameter attribute is initialised with the same reshape
there is no runtime error.

2020-08-20  Steven G. Kargl  <kargl@gcc.gnu.org>

gcc/fortran/

	PR fortran/96624
	* simplify.c (gfc_simplifiy_reshape): Add new variable "zerosize".
	Set zerosize if any of the result shape ranks are zero.  After
	setting the result shapes, if zerosize is set jump to new label
	"sizezero".  Add label "sizezero" just before clearing index and
	returning result.

2020-08-20  Mark Eggleston  <markeggleston@gcc.gnu.org>

gcc/testsuite/

	PR fortran/96624
	*gfortran/pr96624.f90: New test.
---
 gcc/fortran/simplify.c                | 11 ++++++++++-
 gcc/testsuite/gfortran.dg/pr96624.f90 | 10 ++++++++++
 2 files changed, 20 insertions(+), 1 deletion(-)
 create mode 100644 gcc/testsuite/gfortran.dg/pr96624.f90

diff --git a/gcc/fortran/simplify.c b/gcc/fortran/simplify.c
index eb8b2afeb29..0d77d289651 100644
--- a/gcc/fortran/simplify.c
+++ b/gcc/fortran/simplify.c
@@ -6721,6 +6721,7 @@  gfc_simplify_reshape (gfc_expr *source, gfc_expr *shape_exp,
   unsigned long j;
   size_t nsource;
   gfc_expr *e, *result;
+  bool zerosize = false;
 
   /* Check that argument expression types are OK.  */
   if (!is_constant_array_expr (source)
@@ -6843,7 +6844,14 @@  gfc_simplify_reshape (gfc_expr *source, gfc_expr *shape_exp,
   result->rank = rank;
   result->shape = gfc_get_shape (rank);
   for (i = 0; i < rank; i++)
-    mpz_init_set_ui (result->shape[i], shape[i]);
+    {
+      mpz_init_set_ui (result->shape[i], shape[i]);
+      if (shape[i] == 0)
+	zerosize = true;
+    }
+ 
+  if (zerosize)
+    goto sizezero;
 
   while (nsource > 0 || npad > 0)
     {
@@ -6893,6 +6901,7 @@  inc:
       break;
     }
 
+sizezero:
   mpz_clear (index);
 
   return result;
diff --git a/gcc/testsuite/gfortran.dg/pr96624.f90 b/gcc/testsuite/gfortran.dg/pr96624.f90
new file mode 100644
index 00000000000..a4cfe5c3279
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/pr96624.f90
@@ -0,0 +1,10 @@ 
+! { dg-do run }
+
+program test
+  integer :: a(2,0)
+  character(4) :: buffer
+  a = reshape([1,2,3,4], [2,0])
+  write(buffer,"(2a1)") ">", "<"
+  if (trim(buffer).ne."><") stop 1
+end
+
-- 
2.11.0