[PATCH] Fortran: out of bounds access with nested implied-do IO [PR111837]

[PATCH] Fortran: out of bounds access with nested implied-do IO [PR111837]

[Harald Anlauf]

[-- Attachment #1: Type: text/plain, Size: 394 bytes --]

Dear All,

the attached patch fixes a dependency check in frontend optimzation
for nested implied-do IO.  The problem appeared for >= 3 loops only
as the check considered dependencies to be only of band form instead
of triangular form.

Regtested on x86_64-pc-linux-gnu.  OK for mainline?

As this fixes a regression since 8-release, I plan to backport
to all active branches.

Thanks,
Harald


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: pr111837.diff --]
[-- Type: text/x-patch, Size: 2125 bytes --]

From 43ec8b856a67a1b70744e5c0d50ea7fa2dd9a8ee Mon Sep 17 00:00:00 2001
From: Harald Anlauf <anlauf@gmx.de>
Date: Mon, 16 Oct 2023 21:02:20 +0200
Subject: [PATCH] Fortran: out of bounds access with nested implied-do IO
 [PR111837]

gcc/fortran/ChangeLog:

	PR fortran/111837
	* frontend-passes.cc (traverse_io_block): Dependency check of loop
	nest shall be triangular, not banded.

gcc/testsuite/ChangeLog:

	PR fortran/111837
	* gfortran.dg/implied_do_io_8.f90: New test.
---
 gcc/fortran/frontend-passes.cc                |  2 +-
 gcc/testsuite/gfortran.dg/implied_do_io_8.f90 | 18 ++++++++++++++++++
 2 files changed, 19 insertions(+), 1 deletion(-)
 create mode 100644 gcc/testsuite/gfortran.dg/implied_do_io_8.f90

diff --git a/gcc/fortran/frontend-passes.cc b/gcc/fortran/frontend-passes.cc
index 136a292807d..536884b13f0 100644
--- a/gcc/fortran/frontend-passes.cc
+++ b/gcc/fortran/frontend-passes.cc
@@ -1326,7 +1326,7 @@ traverse_io_block (gfc_code *code, bool *has_reached, gfc_code *prev)
       if (iters[i])
 	{
 	  gfc_expr *var = iters[i]->var;
-	  for (int j = i - 1; j < i; j++)
+	  for (int j = 0; j < i; j++)
 	    {
 	      if (iters[j]
 		  && (var_in_expr (var, iters[j]->start)
diff --git a/gcc/testsuite/gfortran.dg/implied_do_io_8.f90 b/gcc/testsuite/gfortran.dg/implied_do_io_8.f90
new file mode 100644
index 00000000000..c66a0f6fde6
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/implied_do_io_8.f90
@@ -0,0 +1,18 @@
+! { dg-do run }
+! { dg-additional-options "-fcheck=bounds" }
+! PR fortran/111837 - out of bounds access with front-end optimization
+
+program implied_do_bug
+  implicit none
+  integer :: i,j,k
+  real    :: arr(1,1,1)
+  integer :: ni(1)
+  ni(1) = 1
+  arr = 1
+  write(*,*) (((arr(i,j,k), i=1,ni(k)), k=1,1), j=1,1)
+  write(*,*) (((arr(i,j,k), i=1,ni(k)), j=1,1), k=1,1)
+  write(*,*) (((arr(k,i,j), i=1,ni(k)), k=1,1), j=1,1)
+  write(*,*) (((arr(k,i,j), i=1,ni(k)), j=1,1), k=1,1)
+  write(*,*) (((arr(j,k,i), i=1,ni(k)), k=1,1), j=1,1)
+  write(*,*) (((arr(j,k,i), i=1,ni(k)), j=1,1), k=1,1)
+end
--
2.35.3

Re: [PATCH] Fortran: out of bounds access with nested implied-do IO [PR111837]

[Jerry D]

On 10/16/23 12:11 PM, Harald Anlauf wrote:
> Dear All,
> 
> the attached patch fixes a dependency check in frontend optimzation
> for nested implied-do IO.  The problem appeared for >= 3 loops only
> as the check considered dependencies to be only of band form instead
> of triangular form.
> 
> Regtested on x86_64-pc-linux-gnu.  OK for mainline?
> 
> As this fixes a regression since 8-release, I plan to backport
> to all active branches.
> 
> Thanks,
> Harald
> 

OK for Mainline and backport

Thanks Harald

Jerry