* [PATCH 1/3] Fix _GL_HAVE__STATIC_ASSERT typo
@ 2019-05-09 16:10 Paul Eggert
2019-05-09 16:10 ` [PATCH 2/3] Support C2X and C++17 static_assert Paul Eggert
2019-05-09 16:10 ` [PATCH 3/3] verify: remove verify_true Paul Eggert
0 siblings, 2 replies; 5+ messages in thread
From: Paul Eggert @ 2019-05-09 16:10 UTC (permalink / raw)
To: bug-gnulib; +Cc: Paul Eggert
* lib/verify.h (_Static_assert): For the FreeBSD workaround,
use _GL_HAVE__STATIC_ASSERT, not _GL_HAVE_STATIC_ASSERT.
---
ChangeLog | 6 ++++++
lib/verify.h | 6 +++---
2 files changed, 9 insertions(+), 3 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index 31e05adbd..12231d97b 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,9 @@
+2019-05-08 Paul Eggert <eggert@cs.ucla.edu>
+
+ Fix _GL_HAVE__STATIC_ASSERT typo
+ * lib/verify.h (_Static_assert): For the FreeBSD workaround,
+ use _GL_HAVE__STATIC_ASSERT, not _GL_HAVE_STATIC_ASSERT.
+
2019-05-05 Bruno Haible <bruno@clisp.org>
wcwidth: Ensure width 1, not 2, for ambiguous characters.
diff --git a/lib/verify.h b/lib/verify.h
index 6930645a3..eccd7e201 100644
--- a/lib/verify.h
+++ b/lib/verify.h
@@ -32,18 +32,18 @@
down we could also use it with other compilers, but since this
affects only the quality of diagnostics, why bother? */
#if (4 < __GNUC__ + (6 <= __GNUC_MINOR__) \
- && (201112L <= __STDC_VERSION__ || !defined __STRICT_ANSI__) \
+ && (201112L <= __STDC_VERSION__ || !defined __STRICT_ANSI__) \
&& !defined __cplusplus)
# define _GL_HAVE__STATIC_ASSERT 1
#endif
-#if (6 <= __GNUC__) && defined __cplusplus
+#if 6 <= __GNUC__ && defined __cplusplus
# define _GL_HAVE_STATIC_ASSERT 1
#endif
/* FreeBSD 9.1 <sys/cdefs.h>, included by <stddef.h> and lots of other
system headers, defines a conflicting _Static_assert that is no
better than ours; override it. */
-#ifndef _GL_HAVE_STATIC_ASSERT
+#ifndef _GL_HAVE__STATIC_ASSERT
# include <stddef.h>
# undef _Static_assert
#endif
--
2.21.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 2/3] Support C2X and C++17 static_assert
2019-05-09 16:10 [PATCH 1/3] Fix _GL_HAVE__STATIC_ASSERT typo Paul Eggert
@ 2019-05-09 16:10 ` Paul Eggert
2019-05-09 16:10 ` [PATCH 3/3] verify: remove verify_true Paul Eggert
1 sibling, 0 replies; 5+ messages in thread
From: Paul Eggert @ 2019-05-09 16:10 UTC (permalink / raw)
To: bug-gnulib; +Cc: Paul Eggert
C2X and C++17 finally added support for a simple, single-argument
‘static_assert’ that implements what the Gnulib ‘verify’ macro was
doing back in 2005. Implement static_assert on older platforms.
The only remaining advantage of ‘verify’ is a shorter name.
* doc/posix-headers/assert.texi (assert.h):
* doc/verify.texi (Compile-time Assertions):
Modernize for C2X and C++17.
* lib/verify.h (_GL_HAVE__STATIC_ASSERT1, _GL_HAVE_STATIC_ASSERT1):
New macros.
(_GL_HAVE__STATIC_ASSERT): Remove.
(_GL_HAVE__STATIC_ASSERT): Rely more heavily on __STDC_VERSION__.
(_GL_VERIFY_TRUE, _GL_VERIFY_TYPE): Remove 2nd arg, the diagnostic
string. All callers changed.
(_GL_VERIFY): Require 3 or more args, of which only the first 2
are used. All callers changed.
(_Static_assert): Allow either 1 or 2 args, and define if
!_GL_HAVE__STATIC_ASSERT1 instead of defining if
!_GL_HAVE__STATIC_ASSERT.
(static_assert): Define if !_GL_HAVE_STATIC_ASSERT1 instead
of defining if !_GL_HAVE_STATIC_ASSERT.
(verify_expr, verify): Don’t bother trying to copy the expression
into the diagnostic, since 1-argument static_assert doesn’t.
(verify): Prefer 1-argument _Static_assert if it works.
* m4/assert_h.m4 (gl_ASSERT_H): Check for 1-argument static_assert.
---
ChangeLog | 28 ++++++++++
| 21 +++++---
doc/verify.texi | 15 +++---
lib/verify.h | 97 +++++++++++++++++++----------------
m4/assert_h.m4 | 2 +
5 files changed, 105 insertions(+), 58 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index 12231d97b..760cde3c6 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,31 @@
+2019-05-09 Paul Eggert <eggert@cs.ucla.edu>
+
+ Support C2X and C++17 static_assert
+ C2X and C++17 finally added support for a simple, single-argument
+ ‘static_assert’ that implements what the Gnulib ‘verify’ macro was
+ doing back in 2005. Implement static_assert on older platforms.
+ The only remaining advantage of ‘verify’ is a shorter name.
+ * doc/posix-headers/assert.texi (assert.h):
+ * doc/verify.texi (Compile-time Assertions):
+ Modernize for C2X and C++17.
+ * lib/verify.h (_GL_HAVE__STATIC_ASSERT1, _GL_HAVE_STATIC_ASSERT1):
+ New macros.
+ (_GL_HAVE__STATIC_ASSERT): Remove.
+ (_GL_HAVE__STATIC_ASSERT): Rely more heavily on __STDC_VERSION__.
+ (_GL_VERIFY_TRUE, _GL_VERIFY_TYPE): Remove 2nd arg, the diagnostic
+ string. All callers changed.
+ (_GL_VERIFY): Require 3 or more args, of which only the first 2
+ are used. All callers changed.
+ (_Static_assert): Allow either 1 or 2 args, and define if
+ !_GL_HAVE__STATIC_ASSERT1 instead of defining if
+ !_GL_HAVE__STATIC_ASSERT.
+ (static_assert): Define if !_GL_HAVE_STATIC_ASSERT1 instead
+ of defining if !_GL_HAVE_STATIC_ASSERT.
+ (verify_expr, verify): Don’t bother trying to copy the expression
+ into the diagnostic, since 1-argument static_assert doesn’t.
+ (verify): Prefer 1-argument _Static_assert if it works.
+ * m4/assert_h.m4 (gl_ASSERT_H): Check for 1-argument static_assert.
+
2019-05-08 Paul Eggert <eggert@cs.ucla.edu>
Fix _GL_HAVE__STATIC_ASSERT typo
--git a/doc/posix-headers/assert.texi b/doc/posix-headers/assert.texi
index 785a07aa9..fa99d3b88 100644
--- a/doc/posix-headers/assert.texi
+++ b/doc/posix-headers/assert.texi
@@ -5,29 +5,34 @@ POSIX specification:@* @url{http://www.opengroup.org/onlinepubs/9699919799/based
Gnulib module: assert-h
-See also the Gnulib module @code{assert}.
+See also the Gnulib modules @code{assert} and @code{verify}.
Portability problems fixed by Gnulib:
@itemize
@item
-The C11 and C++11 @code{static_assert}, and the C11
-@code{_Static_assert}, are not supported by many platforms.
-For example, GCC versions before 4.6.0 do not support @code{_Static_assert},
-and G++ versions through at least 4.6.0 do not support @code{static_assert}.
+On older platforms @code{static_assert} and @code{_Static_assert} do
+not allow the second string-literal argument to be omitted. For
+example, GCC versions before 9.1 do not support the single-argument
+@code{static_assert} that was standardized by C2X and C++17.
+@item
+Even-older platforms do not support @code{static_assert} or
+@code{_Static_assert} at all. For example, GCC versions before 4.6 do
+not support @code{_Static_assert}, and G++ versions before 4.3 do not
+support @code{static_assert}, which was standardized by C11 and C++11.
@end itemize
Portability problems not fixed by Gnulib:
@itemize
@item
-C11 @code{_Static_assert} and C++11 @code{static_assert}
+C @code{_Static_assert} and C++ @code{static_assert}
are keywords that can be used without including @code{<assert.h>}.
The Gnulib substitutes are macros that require including @code{<assert.h>}.
@item
-The C11 @code{static_assert} and @code{_Static_assert} can also
+The C @code{static_assert} and @code{_Static_assert} can also
be used within a @code{struct} or @code{union} specifier, in place of
an ordinary declaration of a member of the struct or union. The
Gnulib substitute can be used only as an ordinary declaration.
@item
-In C99, @code{assert} can be applied to any scalar expression.
+In C99 and later, @code{assert} can be applied to any scalar expression.
In C89, the argument to @code{assert} is of type @code{int}.
@end itemize
diff --git a/doc/verify.texi b/doc/verify.texi
index 3a92f2e05..65aede210 100644
--- a/doc/verify.texi
+++ b/doc/verify.texi
@@ -52,15 +52,18 @@ integer constant expression, then a compiler might reject a usage like
@samp{verify (@var{V});} even when @var{V} is
nonzero.
-Although the standard @code{assert} macro is a runtime test, C11
-specifies a builtin @code{_Static_assert (@var{V},
-@var{STRING-LITERAL})}, its @file{assert.h} header has a similar macro
-named @code{static_assert}, and C++11 has a similar
+Although the standard @code{assert} macro is a runtime test, C2X
+specifies a builtin @code{_Static_assert (@var{V})},
+its @file{assert.h} header has a similar macro
+named @code{static_assert}, and C++17 has a similar
@code{static_assert} builtin. These builtins and macros differ
from @code{verify} in two major ways. First, they can also be used
within a @code{struct} or @code{union} specifier, in place of an
-ordinary member declaration. Second, they require the programmer to
-specify a compile-time diagnostic as a string literal.
+ordinary member declaration. Second, they allow the programmer to
+specify, as an optional second argument, a compile-time diagnostic as
+a string literal. If your program is not intended to be portable to
+compilers that lack C2X or C++17 @code{static_assert}, the only
+advantage of @code{verify} is that its name is a bit shorter.
The @file{verify.h} header defines one more macro, @code{assume
(@var{E})}, which expands to an expression of type @code{void}
diff --git a/lib/verify.h b/lib/verify.h
index eccd7e201..d7e15bc8a 100644
--- a/lib/verify.h
+++ b/lib/verify.h
@@ -21,23 +21,31 @@
#define _GL_VERIFY_H
-/* Define _GL_HAVE__STATIC_ASSERT to 1 if _Static_assert works as per C11.
- This is supported by GCC 4.6.0 and later, in C mode, and its use
- here generates easier-to-read diagnostics when verify (R) fails.
-
- Define _GL_HAVE_STATIC_ASSERT to 1 if static_assert works as per C++11.
- This is supported by GCC 6.1.0 and later, in C++ mode.
-
- Use this only with GCC. If we were willing to slow 'configure'
- down we could also use it with other compilers, but since this
- affects only the quality of diagnostics, why bother? */
-#if (4 < __GNUC__ + (6 <= __GNUC_MINOR__) \
- && (201112L <= __STDC_VERSION__ || !defined __STRICT_ANSI__) \
- && !defined __cplusplus)
-# define _GL_HAVE__STATIC_ASSERT 1
-#endif
-#if 6 <= __GNUC__ && defined __cplusplus
-# define _GL_HAVE_STATIC_ASSERT 1
+/* Define _GL_HAVE__STATIC_ASSERT to 1 if _Static_assert (R, DIAGNOSTIC)
+ works as per C11. This is supported by GCC 4.6.0 and later, in C
+ mode.
+
+ Define _GL_HAVE__STATIC_ASSERT1 to 1 if _Static_assert (R) works as
+ per C2X, and define _GL_HAVE_STATIC_ASSERT1 if static_assert (R)
+ works as per C++17. This is supported by GCC 9.1 and later.
+
+ Support compilers claiming conformance to the relevant standard,
+ and also support GCC when not pedantic. If we were willing to slow
+ 'configure' down we could also use it with other compilers, but
+ since this affects only the quality of diagnostics, why bother? */
+#ifndef __cplusplus
+# if (201112L <= __STDC_VERSION__ \
+ || (!defined __STRICT_ANSI__ && 4 < __GNUC__ + (6 <= __GNUC_MINOR__)))
+# define _GL_HAVE__STATIC_ASSERT 1
+# endif
+# if (202000L <= __STDC_VERSION__ \
+ || (!defined __STRICT_ANSI__ && 9 <= __GNUC__))
+# define _GL_HAVE__STATIC_ASSERT1 1
+# endif
+#else
+# if 201703L <= __cplusplus || 9 <= __GNUC__
+# define _GL_HAVE_STATIC_ASSERT1 1
+# endif
#endif
/* FreeBSD 9.1 <sys/cdefs.h>, included by <stddef.h> and lots of other
@@ -167,11 +175,9 @@
#define _GL_GENSYM(prefix) _GL_CONCAT (prefix, _GL_COUNTER)
/* Verify requirement R at compile-time, as an integer constant expression
- that returns 1. If R is false, fail at compile-time, preferably
- with a diagnostic that includes the string-literal DIAGNOSTIC. */
+ that returns 1. If R is false, fail at compile-time. */
-#define _GL_VERIFY_TRUE(R, DIAGNOSTIC) \
- (!!sizeof (_GL_VERIFY_TYPE (R, DIAGNOSTIC)))
+#define _GL_VERIFY_TRUE(R) (!!sizeof (_GL_VERIFY_TYPE (R)))
#ifdef __cplusplus
# if !GNULIB_defined_struct__gl_verify_type
@@ -181,40 +187,43 @@ template <int w>
};
# define GNULIB_defined_struct__gl_verify_type 1
# endif
-# define _GL_VERIFY_TYPE(R, DIAGNOSTIC) \
- _gl_verify_type<(R) ? 1 : -1>
-#elif defined _GL_HAVE__STATIC_ASSERT
-# define _GL_VERIFY_TYPE(R, DIAGNOSTIC) \
+# define _GL_VERIFY_TYPE(R) _gl_verify_type<(R) ? 1 : -1>
+#elif defined _GL_HAVE__STATIC_ASSERT1
+# define _GL_VERIFY_TYPE(R) \
struct { \
- _Static_assert (R, DIAGNOSTIC); \
+ _Static_assert (R); \
int _gl_dummy; \
}
#else
-# define _GL_VERIFY_TYPE(R, DIAGNOSTIC) \
+# define _GL_VERIFY_TYPE(R) \
struct { unsigned int _gl_verify_error_if_negative: (R) ? 1 : -1; }
#endif
/* Verify requirement R at compile-time, as a declaration without a
- trailing ';'. If R is false, fail at compile-time, preferably
- with a diagnostic that includes the string-literal DIAGNOSTIC.
+ trailing ';'. If R is false, fail at compile-time.
+
+ This macro requires three or more arguments but uses at most the first
+ two, so that the _Static_assert macro optionally defined below supports
+ both the C11 two-argument syntax and the C2X one-argument syntax.
Unfortunately, unlike C11, this implementation must appear as an
ordinary declaration, and cannot appear inside struct { ... }. */
-#ifdef _GL_HAVE__STATIC_ASSERT
-# define _GL_VERIFY _Static_assert
+#if defined _GL_HAVE__STATIC_ASSERT
+# define _GL_VERIFY(R, DIAGNOSTIC, ...) _Static_assert (R, DIAGNOSTIC)
#else
-# define _GL_VERIFY(R, DIAGNOSTIC) \
+# define _GL_VERIFY(R, DIAGNOSTIC, ...) \
extern int (*_GL_GENSYM (_gl_verify_function) (void)) \
- [_GL_VERIFY_TRUE (R, DIAGNOSTIC)]
+ [_GL_VERIFY_TRUE (R)]
#endif
/* _GL_STATIC_ASSERT_H is defined if this code is copied into assert.h. */
#ifdef _GL_STATIC_ASSERT_H
-# if !defined _GL_HAVE__STATIC_ASSERT && !defined _Static_assert
-# define _Static_assert(R, DIAGNOSTIC) _GL_VERIFY (R, DIAGNOSTIC)
+# if !defined _GL_HAVE__STATIC_ASSERT1 && !defined _Static_assert
+# define _Static_assert(...) \
+ _GL_VERIFY (__VA_ARGS__, "static assertion failed", -)
# endif
-# if !defined _GL_HAVE_STATIC_ASSERT && !defined static_assert
+# if !defined _GL_HAVE_STATIC_ASSERT1 && !defined static_assert
# define static_assert _Static_assert /* C11 requires this #define. */
# endif
#endif
@@ -235,22 +244,22 @@ template <int w>
verify_true is obsolescent; please use verify_expr instead. */
-#define verify_true(R) _GL_VERIFY_TRUE (R, "verify_true (" #R ")")
+#define verify_true(R) _GL_VERIFY_TRUE (R)
/* Verify requirement R at compile-time. Return the value of the
expression E. */
-#define verify_expr(R, E) \
- (_GL_VERIFY_TRUE (R, "verify_expr (" #R ", " #E ")") ? (E) : (E))
+#define verify_expr(R, E) (_GL_VERIFY_TRUE (R) ? (E) : (E))
/* Verify requirement R at compile-time, as a declaration without a
- trailing ';'. */
+ trailing ';'. verify (R) acts like static_assert (R) except that
+ it is portable to C11/C++14 and earlier, and its name is shorter
+ and may be more convenient. */
-#ifdef __GNUC__
-# define verify(R) _GL_VERIFY (R, "verify (" #R ")")
+#ifdef _GL_HAVE__STATIC_ASSERT1
+# define verify(R) _Static_assert (R)
#else
-/* PGI barfs if R is long. Play it safe. */
-# define verify(R) _GL_VERIFY (R, "verify (...)")
+# define verify(R) _GL_VERIFY (R, "verify (...)", -)
#endif
#ifndef __has_builtin
diff --git a/m4/assert_h.m4 b/m4/assert_h.m4
index cf11bae6a..ee278b206 100644
--- a/m4/assert_h.m4
+++ b/m4/assert_h.m4
@@ -14,9 +14,11 @@ AC_DEFUN([gl_ASSERT_H],
[AC_LANG_PROGRAM(
[[#include <assert.h>
static_assert (2 + 2 == 4, "arithmetic doesn't work");
+ static_assert (2 + 2 == 4);
]],
[[
static_assert (sizeof (char) == 1, "sizeof doesn't work");
+ static_assert (sizeof (char) == 1);
]])],
[gl_cv_static_assert=yes],
[gl_cv_static_assert=no])])
--
2.21.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 3/3] verify: remove verify_true
2019-05-09 16:10 [PATCH 1/3] Fix _GL_HAVE__STATIC_ASSERT typo Paul Eggert
2019-05-09 16:10 ` [PATCH 2/3] Support C2X and C++17 static_assert Paul Eggert
@ 2019-05-09 16:10 ` Paul Eggert
2019-06-18 22:26 ` Bruno Haible
1 sibling, 1 reply; 5+ messages in thread
From: Paul Eggert @ 2019-05-09 16:10 UTC (permalink / raw)
To: bug-gnulib; +Cc: Paul Eggert
* NEWS: Mention this.
* lib/verify.h (verify_true): Remove.
* tests/test-verify.c (item): Test verify_expr, not verify_true.
---
ChangeLog | 5 +++++
NEWS | 2 ++
lib/verify.h | 13 +++----------
tests/test-verify.c | 2 +-
4 files changed, 11 insertions(+), 11 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index 760cde3c6..10c513cb6 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,10 @@
2019-05-09 Paul Eggert <eggert@cs.ucla.edu>
+ verify: remove verify_true
+ * NEWS: Mention this.
+ * lib/verify.h (verify_true): Remove.
+ * tests/test-verify.c (item): Test verify_expr, not verify_true.
+
Support C2X and C++17 static_assert
C2X and C++17 finally added support for a simple, single-argument
‘static_assert’ that implements what the Gnulib ‘verify’ macro was
diff --git a/NEWS b/NEWS
index 77b9f24a7..2655d7fc9 100644
--- a/NEWS
+++ b/NEWS
@@ -55,6 +55,8 @@ User visible incompatible changes
Date Modules Changes
+2019-05-90 verify verify_true (deprecated 2011-06-15) is removed.
+
2019-02-02 c-strtod This and related modules no longer define
the HAVE_C99_STRTOLD macro. Programs requiring
standard strtold should use the strtold module.
diff --git a/lib/verify.h b/lib/verify.h
index d7e15bc8a..f8e4eff02 100644
--- a/lib/verify.h
+++ b/lib/verify.h
@@ -149,9 +149,9 @@
which do not support _Static_assert, also do not warn about the
last declaration mentioned above.
- * GCC warns if -Wnested-externs is enabled and verify() is used
+ * GCC warns if -Wnested-externs is enabled and 'verify' is used
within a function body; but inside a function, you can always
- arrange to use verify_expr() instead.
+ arrange to use verify_expr instead.
* In C++, any struct definition inside sizeof is invalid.
Use a template type to work around the problem. */
@@ -235,17 +235,10 @@ template <int w>
assert (R), there is no run-time overhead.
There are two macros, since no single macro can be used in all
- contexts in C. verify_true (R) is for scalar contexts, including
+ contexts in C. verify_expr (R, E) is for scalar contexts, including
integer constant expression contexts. verify (R) is for declaration
contexts, e.g., the top level. */
-/* Verify requirement R at compile-time, as an integer constant expression.
- Return 1. This is equivalent to verify_expr (R, 1).
-
- verify_true is obsolescent; please use verify_expr instead. */
-
-#define verify_true(R) _GL_VERIFY_TRUE (R)
-
/* Verify requirement R at compile-time. Return the value of the
expression E. */
diff --git a/tests/test-verify.c b/tests/test-verify.c
index e677d8aff..0afbcabde 100644
--- a/tests/test-verify.c
+++ b/tests/test-verify.c
@@ -39,7 +39,7 @@ verify (1 == 1); verify (1 == 1); /* should be ok */
enum
{
- item = verify_true (1 == 1) * 0 + 17 /* should be ok */
+ item = verify_expr (1 == 1, 10) * 0 + 17 /* should be ok */
};
static int
--
2.21.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH 3/3] verify: remove verify_true
2019-05-09 16:10 ` [PATCH 3/3] verify: remove verify_true Paul Eggert
@ 2019-06-18 22:26 ` Bruno Haible
2019-06-19 0:52 ` Paul Eggert
0 siblings, 1 reply; 5+ messages in thread
From: Bruno Haible @ 2019-06-18 22:26 UTC (permalink / raw)
To: bug-gnulib; +Cc: Paul Eggert
Hi Paul,
> 2019-05-09 Paul Eggert <eggert@cs.ucla.edu>
>
> + verify: remove verify_true
> + * NEWS: Mention this.
> + * lib/verify.h (verify_true): Remove.
> + * tests/test-verify.c (item): Test verify_expr, not verify_true.
> +
As a consequence of this change, the test-verify.c file does not compile
any more on AIX 7.1 with xlc:
xlc: 1501-230 (S) Internal compiler error; please contact your Service Representative. For more information visit:
http://www.ibm.com/support/docview.wss?uid=swg21110810
The machine is gcc111.fsffrance.org.
The compiler version is
IBM XL C for AIX, V12.1 (5765-J01, 5725-C71)
Copyright IBM Corp. 1991, 2012 and by others.
All Rights Reserved.
US Government Users Restricted Rights -
Use, duplication or disclosure restricted by
GSA ADP Schedule Contract with IBM Corp.
-----------------------------------------
Version: 12.01.0000.0000
Version: 12.01.0000.0000
Driver 120315
120321
Compiler Services 120315
120203
Date: Wed Mar 21 21:39:51 EDT 2012
-----------------------------------------
The following compiles fine:
enum
{
item = verify_true (1 == 1) * 0 + 17 /* should be ok */
};
int item1 = verify_expr (1 == 1, 10) * 0 + 17;
const int item2 = verify_expr (1 == 1, 10) * 0 + 17;
But the following does not:
enum
{
item = verify_expr (1 == 1, 10) * 0 + 17 /* should be ok */
};
So, it's apparently the use of verify_expr in an enum value.
We had a similar bug with HP-UX cc in the past.
Do you see other options than
(a) reintroduce verify_true and check that it's usable in enum.
(b) relax the check and only check that it's usable as a static initializer.
?
Bruno
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH 3/3] verify: remove verify_true
2019-06-18 22:26 ` Bruno Haible
@ 2019-06-19 0:52 ` Paul Eggert
0 siblings, 0 replies; 5+ messages in thread
From: Paul Eggert @ 2019-06-19 0:52 UTC (permalink / raw)
To: Bruno Haible; +Cc: bug-gnulib
[-- Attachment #1: Type: text/plain, Size: 863 bytes --]
On 6/18/19 3:26 PM, Bruno Haible wrote:
> Do you see other options than
> (a) reintroduce verify_true and check that it's usable in enum.
> (b) relax the check and only check that it's usable as a static initializer.
Also (c) relax the check so that the buggy compiler passes but continue
to check that verify_expr is usable in an enum. Sort of a hack, but
there it is - it's arguably good enough since the (now-removed)
verify_true macro can also make that buggy compiler crash if we
complicate the test a bit. I installed the attached, which worked for me
on gcc111.
> Version: 12.01.0000.0000
Quite possibly that compiler bug is fixed in xlc by now; IBM says the
current xlc 12.1 release is 12.1.0.21 dated January 2019. I don't know
who's maintaining gcc111 though.
https://www-01.ibm.com/support/docview.wss?uid=swg27038605#121021
[-- Attachment #2: 0001-verify-tests-work-around-xlc-bug.patch --]
[-- Type: text/x-patch, Size: 1430 bytes --]
From 75acd2ec14fcc0b0695e8a108167562d65417f79 Mon Sep 17 00:00:00 2001
From: Paul Eggert <eggert@cs.ucla.edu>
Date: Tue, 18 Jun 2019 17:41:46 -0700
Subject: [PATCH] verify-tests: work around xlc bug
Problem reported by Bruno Haible in:
https://lists.gnu.org/r/bug-gnulib/2019-06/msg00049.html
* tests/test-verify.c (item): Move the arithmetic inside the
verify_expr, to avoid tickling a bug in IBM AIX xlc V12.1.
---
ChangeLog | 8 ++++++++
tests/test-verify.c | 2 +-
2 files changed, 9 insertions(+), 1 deletion(-)
diff --git a/ChangeLog b/ChangeLog
index 3ab71a3d9..c76ac8852 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,11 @@
+2019-06-18 Paul Eggert <eggert@cs.ucla.edu>
+
+ verify-tests: work around xlc bug
+ Problem reported by Bruno Haible in:
+ https://lists.gnu.org/r/bug-gnulib/2019-06/msg00049.html
+ * tests/test-verify.c (item): Move the arithmetic inside the
+ verify_expr, to avoid tickling a bug in IBM AIX xlc V12.1.
+
2019-06-16 Bruno Haible <bruno@clisp.org>
Restore Emacs time-stamp hook applicability.
diff --git a/tests/test-verify.c b/tests/test-verify.c
index 0afbcabde..98983c3c4 100644
--- a/tests/test-verify.c
+++ b/tests/test-verify.c
@@ -39,7 +39,7 @@ verify (1 == 1); verify (1 == 1); /* should be ok */
enum
{
- item = verify_expr (1 == 1, 10) * 0 + 17 /* should be ok */
+ item = verify_expr (1 == 1, 10 * 0 + 17) /* should be ok */
};
static int
--
2.21.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
end of thread, other threads:[~2019-06-19 0:58 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-05-09 16:10 [PATCH 1/3] Fix _GL_HAVE__STATIC_ASSERT typo Paul Eggert
2019-05-09 16:10 ` [PATCH 2/3] Support C2X and C++17 static_assert Paul Eggert
2019-05-09 16:10 ` [PATCH 3/3] verify: remove verify_true Paul Eggert
2019-06-18 22:26 ` Bruno Haible
2019-06-19 0:52 ` Paul Eggert
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).