Do not use const attribute for nan functions (bug 23277) [committed]

Message ID alpine.DEB.2.20.1806121657420.18047@digraph.polyomino.org.uk
State New
Headers show
Series
  • Do not use const attribute for nan functions (bug 23277) [committed]
Related show

Commit Message

Joseph Myers June 12, 2018, 4:57 p.m.
As in https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86113 for
__builtin_nan, bits/mathcalls.h wrongly declares the nan function with
the __const__ attribute.  Because the function reads memory pointed to
by an argument, it's only pure, not const.  This patch removes the
incorrect attribute and adds a testcase for the bug.  No __pure__
attribute is added to replace the incorrect __const__ one, since that
would introduce problems when using GCC versions that have the
incorrect built-in __const__ attribute and warn for the combination of
those two attributes.

Tested for x86_64.  Committed.

2018-06-12  Joseph Myers  <joseph@codesourcery.com>

	[BZ #23277]
	* math/bits/mathcalls.h [__USE_ISOC99] (nan): Do not use __const__
	attribute.
	* math/test-nan-const.c: New file.
	* math/Makefile (tests): Add test-nan-const.
	(CFLAGS-test-nan-const.c): New variable.

Comments

Florian Weimer June 12, 2018, 6:15 p.m. | #1
* Joseph Myers:

> +static int
> +do_test (void)
> +{
> +  char buf[2] = { '2', 0 };
> +  float a = nanf (buf);
> +  buf[0] = '3';
> +  float b = nanf (buf);
> +  return memcmp (&a, &b, sizeof (float)) == 0;
> +}

By the way, there's TEST_COMPARE_BLOB now, which can simplify
debugging of test failures because the failure message includes the
blobs that differ.
Joseph Myers June 12, 2018, 7:33 p.m. | #2
On Tue, 12 Jun 2018, Florian Weimer wrote:

> * Joseph Myers:
> 
> > +static int
> > +do_test (void)
> > +{
> > +  char buf[2] = { '2', 0 };
> > +  float a = nanf (buf);
> > +  buf[0] = '3';
> > +  float b = nanf (buf);
> > +  return memcmp (&a, &b, sizeof (float)) == 0;
> > +}
> 
> By the way, there's TEST_COMPARE_BLOB now, which can simplify
> debugging of test failures because the failure message includes the
> blobs that differ.

In this case, the requirement is actually that the blobs must differ, with 
a failure if they are the same (if the compiler optimized out the second 
nanf call and so produced the same NaN for both calls).
Florian Weimer June 13, 2018, 4:55 a.m. | #3
* Joseph Myers:

> On Tue, 12 Jun 2018, Florian Weimer wrote:
>
>> * Joseph Myers:
>> 
>> > +static int
>> > +do_test (void)
>> > +{
>> > +  char buf[2] = { '2', 0 };
>> > +  float a = nanf (buf);
>> > +  buf[0] = '3';
>> > +  float b = nanf (buf);
>> > +  return memcmp (&a, &b, sizeof (float)) == 0;
>> > +}
>> 
>> By the way, there's TEST_COMPARE_BLOB now, which can simplify
>> debugging of test failures because the failure message includes the
>> blobs that differ.
>
> In this case, the requirement is actually that the blobs must differ, with 
> a failure if they are the same (if the compiler optimized out the second 
> nanf call and so produced the same NaN for both calls).

Oh, right.  I was confused.

Patch

diff --git a/math/Makefile b/math/Makefile
index ea141cb..335b1fb 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -235,7 +235,7 @@  tests = test-matherr-3 test-fenv basic-test \
 	test-iseqsig-excess-precision test-flt-eval-method \
 	test-fp-ilogb-constants test-fp-llogb-constants \
 	test-fe-snans-always-signal test-finite-macros test-narrow-macros \
-	$(tests-static)
+	test-nan-const $(tests-static)
 tests-static = test-fpucw-static test-fpucw-ieee-static \
 	       test-signgam-uchar-static test-signgam-uchar-init-static \
 	       test-signgam-uint-static test-signgam-uint-init-static \
@@ -442,6 +442,8 @@  CFLAGS-test-fe-snans-always-signal.c += -fsignaling-nans
 
 CFLAGS-test-finite-macros.c += -ffinite-math-only
 
+CFLAGS-test-nan-const.c += -fno-builtin
+
 include ../Rules
 
 gen-all-calls = $(gen-libm-calls) $(gen-calls)
diff --git a/math/bits/mathcalls.h b/math/bits/mathcalls.h
index 9200363..cf87313 100644
--- a/math/bits/mathcalls.h
+++ b/math/bits/mathcalls.h
@@ -198,7 +198,7 @@  __MATHCALLX (copysign,, (_Mdouble_ __x, _Mdouble_ __y), (__const__));
 
 #ifdef __USE_ISOC99
 /* Return representation of qNaN for double type.  */
-__MATHCALLX (nan,, (const char *__tagb), (__const__));
+__MATHCALL (nan,, (const char *__tagb));
 #endif
 
 
diff --git a/math/test-nan-const.c b/math/test-nan-const.c
new file mode 100644
index 0000000..f7bba9e
--- /dev/null
+++ b/math/test-nan-const.c
@@ -0,0 +1,32 @@ 
+/* Test nan functions do not have const attribute.  Bug 23277.
+   Copyright (C) 2018 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <math.h>
+#include <string.h>
+
+static int
+do_test (void)
+{
+  char buf[2] = { '2', 0 };
+  float a = nanf (buf);
+  buf[0] = '3';
+  float b = nanf (buf);
+  return memcmp (&a, &b, sizeof (float)) == 0;
+}
+
+#include <support/test-driver.c>