* [PATCH] Fortran: error recovery on arithmetic overflow on unary operations [PR113799]
@ 2024-02-08 21:03 Harald Anlauf
2024-02-09 0:33 ` Jerry D
0 siblings, 1 reply; 2+ messages in thread
From: Harald Anlauf @ 2024-02-08 21:03 UTC (permalink / raw)
To: fortran, gcc-patches
[-- Attachment #1: Type: text/plain, Size: 538 bytes --]
Dear all,
the attached patch improves error recovery when we encounter an
array constructor where a unary operator (e.g. minus) is applied
and -frange-check is active. The solution is not to terminate
early in that case to avoid inconsistencies between check_result
and reduce_unary when such a situation occurs.
(There might be similar issues for binary operators, not treated
here.)
Regtested on x86_64-pc-linux-gnu. OK for mainline?
The ICE/memory corruption is actually a 10+ regression.
Do we need a backport?
Thanks,
Harald
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: pr113799.diff --]
[-- Type: text/x-patch, Size: 2686 bytes --]
From eec039211e396e35204b55588013d74289a984cd Mon Sep 17 00:00:00 2001
From: Harald Anlauf <anlauf@gmx.de>
Date: Thu, 8 Feb 2024 21:51:38 +0100
Subject: [PATCH] Fortran: error recovery on arithmetic overflow on unary
operations [PR113799]
PR fortran/113799
gcc/fortran/ChangeLog:
* arith.cc (reduce_unary): Remember any overflow encountered during
reduction of unary arithmetic operations on array constructors and
continue, and return error status, but terminate on serious errors.
gcc/testsuite/ChangeLog:
* gfortran.dg/arithmetic_overflow_2.f90: New test.
---
gcc/fortran/arith.cc | 11 ++++++++---
gcc/testsuite/gfortran.dg/arithmetic_overflow_2.f90 | 12 ++++++++++++
2 files changed, 20 insertions(+), 3 deletions(-)
create mode 100644 gcc/testsuite/gfortran.dg/arithmetic_overflow_2.f90
diff --git a/gcc/fortran/arith.cc b/gcc/fortran/arith.cc
index 0598f6ac51b..d17d1aaa1d9 100644
--- a/gcc/fortran/arith.cc
+++ b/gcc/fortran/arith.cc
@@ -1323,6 +1323,7 @@ reduce_unary (arith (*eval) (gfc_expr *, gfc_expr **), gfc_expr *op,
gfc_constructor *c;
gfc_expr *r;
arith rc;
+ bool ov = false;
if (op->expr_type == EXPR_CONSTANT)
return eval (op, result);
@@ -1336,13 +1337,17 @@ reduce_unary (arith (*eval) (gfc_expr *, gfc_expr **), gfc_expr *op,
{
rc = reduce_unary (eval, c->expr, &r);
- if (rc != ARITH_OK)
+ /* Remember any overflow encountered during reduction and continue,
+ but terminate on serious errors. */
+ if (rc == ARITH_OVERFLOW)
+ ov = true;
+ else if (rc != ARITH_OK)
break;
gfc_replace_expr (c->expr, r);
}
- if (rc != ARITH_OK)
+ if (rc != ARITH_OK && rc != ARITH_OVERFLOW)
gfc_constructor_free (head);
else
{
@@ -1363,7 +1368,7 @@ reduce_unary (arith (*eval) (gfc_expr *, gfc_expr **), gfc_expr *op,
*result = r;
}
- return rc;
+ return ov ? ARITH_OVERFLOW : rc;
}
diff --git a/gcc/testsuite/gfortran.dg/arithmetic_overflow_2.f90 b/gcc/testsuite/gfortran.dg/arithmetic_overflow_2.f90
new file mode 100644
index 00000000000..6ca27f74215
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/arithmetic_overflow_2.f90
@@ -0,0 +1,12 @@
+! { dg-do compile }
+! { dg-additional-options "-frange-check" }
+!
+! PR fortran/113799 - handle arithmetic overflow on unary minus
+
+program p
+ implicit none
+ real, parameter :: inf = real(z'7F800000')
+ real, parameter :: someInf(*) = [inf, 0.]
+ print *, -someInf ! { dg-error "Arithmetic overflow" }
+ print *, minval(-someInf) ! { dg-error "Arithmetic overflow" }
+end
--
2.35.3
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] Fortran: error recovery on arithmetic overflow on unary operations [PR113799]
2024-02-08 21:03 [PATCH] Fortran: error recovery on arithmetic overflow on unary operations [PR113799] Harald Anlauf
@ 2024-02-09 0:33 ` Jerry D
0 siblings, 0 replies; 2+ messages in thread
From: Jerry D @ 2024-02-09 0:33 UTC (permalink / raw)
To: Harald Anlauf, fortran, gcc-patches
On 2/8/24 1:03 PM, Harald Anlauf wrote:
> Dear all,
>
> the attached patch improves error recovery when we encounter an
> array constructor where a unary operator (e.g. minus) is applied
> and -frange-check is active. The solution is not to terminate
> early in that case to avoid inconsistencies between check_result
> and reduce_unary when such a situation occurs.
>
> (There might be similar issues for binary operators, not treated
> here.)
>
> Regtested on x86_64-pc-linux-gnu. OK for mainline?
>
> The ICE/memory corruption is actually a 10+ regression.
> Do we need a backport?
>
> Thanks,
> Harald
>
Hi Harald,
This patch looks OK.
Thanks,
Jerry
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2024-02-09 0:33 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-02-08 21:03 [PATCH] Fortran: error recovery on arithmetic overflow on unary operations [PR113799] Harald Anlauf
2024-02-09 0:33 ` Jerry D
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).