4:52 p.m.
I have code that trips the warning:
warning: function might be possible candidate for ‘printf’ format attribute
Unfortunately, mingw uses a different declaration for saying something is printf like than
gcc does running on MacOS and Linux.
Is there a way to do the declaration in a portable manner?
5:26 p.m.
On 10/25/2011 03:52 PM, Simson Garfinkel wrote:
I have code that trips the warning:
warning: function might be possible candidate for ‘printf’ format attribute
Unfortunately, mingw uses a different declaration for saying something is printf like
than gcc does running on MacOS and Linux.
Is there a way to do the declaration in a portable manner?
Projects that use gnulib don't have a problem. Gnulib solved this issue
with the following preprocessor goop in a self-built <stdio.h> file that
it builds to wrap the system <stdio.h>:
/* _GL_ATTRIBUTE_FORMAT_PRINTF
indicates to GCC that the function takes a format string and arguments,
where the format string directives are the ones standardized by ISO C99
and POSIX. */
#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 4)
# define _GL_ATTRIBUTE_FORMAT_PRINTF(formatstring_parameter,
first_argument) \
_GL_ATTRIBUTE_FORMAT ((__gnu_printf__, formatstring_parameter,
first_argument))
#else
# define _GL_ATTRIBUTE_FORMAT_PRINTF(formatstring_parameter,
first_argument) \
_GL_ATTRIBUTE_FORMAT ((__printf__, formatstring_parameter,
first_argument))
#endif
/* _GL_ATTRIBUTE_FORMAT_PRINTF_SYSTEM is like _GL_ATTRIBUTE_FORMAT_PRINTF,
except that it indicates to GCC that the supported format string
directives
are the ones of the system printf(), rather than the ones
standardized by
ISO C99 and POSIX. */
#define _GL_ATTRIBUTE_FORMAT_PRINTF_SYSTEM(formatstring_parameter,
first_argument) \
_GL_ATTRIBUTE_FORMAT ((__printf__, formatstring_parameter,
first_argument))
then it makes configure-time decisions on whether to use
_GL_ATTRIBUTE_FORMAT_PRINTF or _GL_ATTRIBUTE_FORMAT_PRINTF according to
whether the package requested the gnulib modules to override printf to
fix implementation bugs, or whether the package just wants the system
functionality (and is steering clear of the buggy constructs).
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library http://libvirt.org
5:30 p.m.
On 10/25/2011 04:26 PM, Eric Blake wrote:
then it makes configure-time decisions on whether to use
_GL_ATTRIBUTE_FORMAT_PRINTF or _GL_ATTRIBUTE_FORMAT_PRINTF according to
whether the package requested the gnulib modules to override printf to
fix implementation bugs, or whether the package just wants the system
functionality (and is steering clear of the buggy constructs).
By the way, the number of buggy printf implementations is appalling,
with the mingw implementations tending to have the most flaws. Also
from gnulib:
dnl 1 = checking whether printf supports size specifiers as in C99...
dnl 2 = checking whether printf supports 'long double' arguments...
dnl 3 = checking whether printf supports infinite 'double' arguments...
dnl 4 = checking whether printf supports infinite 'long double' arguments...
dnl 5 = checking whether printf supports the 'a' and 'A' directives...
dnl 6 = checking whether printf supports the 'F' directive...
dnl 7 = checking whether printf supports the 'n' directive...
dnl 8 = checking whether printf supports the 'ls' directive...
dnl 9 = checking whether printf supports POSIX/XSI format strings with
positions...
dnl 10 = checking whether printf supports the grouping flag...
dnl 11 = checking whether printf supports the left-adjust flag correctly...
dnl 12 = checking whether printf supports the zero flag correctly...
dnl 13 = checking whether printf supports large precisions...
dnl 14 = checking whether printf survives out-of-memory conditions...
dnl 15 = checking for snprintf...
dnl 16 = checking whether snprintf truncates the result as in C99...
dnl 17 = checking whether snprintf returns a byte count as in C99...
dnl 18 = checking whether snprintf fully supports the 'n' directive...
dnl 19 = checking whether snprintf respects a size of 1...
dnl 20 = checking whether vsnprintf respects a zero size as in C99...
dnl
dnl . = yes, # = no.
dnl
dnl 1 2 3 4 5 6 7 8 9 10 11 12
13 14 15 16 17 18 19 20
dnl glibc 2.5 . . . . . . . . . . . .
. . . . . . . .
dnl glibc 2.3.6 . . . . # . . . . . . .
. . . . . . . .
dnl FreeBSD 5.4, 6.1 . . . . # . . . . . . #
. # . . . . . .
dnl MacOS X 10.5.8 . . . # # . . . . . . #
. . . . . . . .
dnl MacOS X 10.3.9 . . . . # . . . . . . #
. # . . . . . .
dnl OpenBSD 3.9, 4.0 . . # # # # . # . # . #
. # . . . . . .
dnl Cygwin 1.7.0 (2009) . . . # . . . ? . . . .
. ? . . . . . .
dnl Cygwin 1.5.25 (2008) . . . # # . . # . . . .
. # . . . . . .
dnl Cygwin 1.5.19 (2006) # . . # # # . # . # . #
# # . . . . . .
dnl Solaris 11 2010-11 . . # # # . . # . . . #
. . . . . . . .
dnl Solaris 10 . . # # # . . # . . . #
# . . . . . . .
dnl Solaris 2.6 ... 9 # . # # # # . # . . . #
# . . . # . . .
dnl Solaris 2.5.1 # . # # # # . # . . . #
. . # # # # # #
dnl AIX 7.1 . . # # # . . . . . . #
# . . . . . . .
dnl AIX 5.2 . . # # # . . . . . . #
. . . . . . . .
dnl AIX 4.3.2, 5.1 # . # # # # . . . . . #
. . . . # . . .
dnl HP-UX 11.31 . . . . # . . . . . . #
. . . . # # . .
dnl HP-UX 11.{00,11,23} # . . . # # . . . . . #
. . . . # # . #
dnl HP-UX 10.20 # . # . # # . ? . . # #
. . . . # # ? #
dnl IRIX 6.5 # . # # # # . # . . . #
. . . . # . . .
dnl OSF/1 5.1 # . # # # # . . . . . #
. . . . # . . #
dnl OSF/1 4.0d # . # # # # . . . . . #
. . # # # # # #
dnl NetBSD 5.0 . . . # # . . . . . . #
. # . . . . . .
dnl NetBSD 4.0 . ? ? ? ? ? . ? . ? ? ?
? ? . . . ? ? ?
dnl NetBSD 3.0 . . . . # # . ? # # ? #
. # . . . . . .
dnl Haiku . . . # # # . # . . . .
. ? . . ? . . .
dnl BeOS # # . # # # . ? # . ? .
# ? . . ? . . .
dnl old mingw / msvcrt # # # # # # . . # # . #
# ? . # # # . .
dnl MSVC 9 # # # # # # # . # # . #
# ? # # # # . .
dnl mingw 2009-2011 . # . # . . . . # # . .
. ? . . . . . .
dnl mingw-w64 2011 # # # # # # . . # # . #
# ? . # # # . .
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library http://libvirt.org
6:11 p.m.
Oh my goodness. Okay. Let me clarify my question.
1. What should I put in configure.ac?
2. Can you show me specifically how to instrument my function, for example,:
void my_crazy_error_printer(void *random_arg,const char *format,...)
void my_other_crazy_printer(const char *fmt,...)
3. I've tried this, but it doesn't work:
void error(const char *fmt, ...) __attribute__((format(printf, 2, 0)));
Thanks!
On Oct 25, 2011, at 6:30 PM, Eric Blake wrote:
On 10/25/2011 04:26 PM, Eric Blake wrote:
> then it makes configure-time decisions on whether to use
> _GL_ATTRIBUTE_FORMAT_PRINTF or _GL_ATTRIBUTE_FORMAT_PRINTF according to
> whether the package requested the gnulib modules to override printf to
> fix implementation bugs, or whether the package just wants the system
> functionality (and is steering clear of the buggy constructs).
By the way, the number of buggy printf implementations is appalling, with the mingw
implementations tending to have the most flaws. Also from gnulib:
dnl 1 = checking whether printf supports size specifiers as in C99...
dnl 2 = checking whether printf supports 'long double' arguments...
dnl 3 = checking whether printf supports infinite 'double' arguments...
dnl 4 = checking whether printf supports infinite 'long double' arguments...
dnl 5 = checking whether printf supports the 'a' and 'A' directives...
dnl 6 = checking whether printf supports the 'F' directive...
dnl 7 = checking whether printf supports the 'n' directive...
dnl 8 = checking whether printf supports the 'ls' directive...
dnl 9 = checking whether printf supports POSIX/XSI format strings with positions...
dnl 10 = checking whether printf supports the grouping flag...
dnl 11 = checking whether printf supports the left-adjust flag correctly...
dnl 12 = checking whether printf supports the zero flag correctly...
dnl 13 = checking whether printf supports large precisions...
dnl 14 = checking whether printf survives out-of-memory conditions...
dnl 15 = checking for snprintf...
dnl 16 = checking whether snprintf truncates the result as in C99...
dnl 17 = checking whether snprintf returns a byte count as in C99...
dnl 18 = checking whether snprintf fully supports the 'n' directive...
dnl 19 = checking whether snprintf respects a size of 1...
dnl 20 = checking whether vsnprintf respects a zero size as in C99...
dnl
dnl . = yes, # = no.
dnl
dnl 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
19 20
dnl glibc 2.5 . . . . . . . . . . . . . . . . . .
. .
dnl glibc 2.3.6 . . . . # . . . . . . . . . . . . .
. .
dnl FreeBSD 5.4, 6.1 . . . . # . . . . . . # . # . . . .
. .
dnl MacOS X 10.5.8 . . . # # . . . . . . # . . . . . .
. .
dnl MacOS X 10.3.9 . . . . # . . . . . . # . # . . . .
. .
dnl OpenBSD 3.9, 4.0 . . # # # # . # . # . # . # . . . .
. .
dnl Cygwin 1.7.0 (2009) . . . # . . . ? . . . . . ? . . . .
. .
dnl Cygwin 1.5.25 (2008) . . . # # . . # . . . . . # . . . .
. .
dnl Cygwin 1.5.19 (2006) # . . # # # . # . # . # # # . . . .
. .
dnl Solaris 11 2010-11 . . # # # . . # . . . # . . . . . .
. .
dnl Solaris 10 . . # # # . . # . . . # # . . . . .
. .
dnl Solaris 2.6 ... 9 # . # # # # . # . . . # # . . . # .
. .
dnl Solaris 2.5.1 # . # # # # . # . . . # . . # # # #
# #
dnl AIX 7.1 . . # # # . . . . . . # # . . . . .
. .
dnl AIX 5.2 . . # # # . . . . . . # . . . . . .
. .
dnl AIX 4.3.2, 5.1 # . # # # # . . . . . # . . . . # .
. .
dnl HP-UX 11.31 . . . . # . . . . . . # . . . . # #
. .
dnl HP-UX 11.{00,11,23} # . . . # # . . . . . # . . . . # #
. #
dnl HP-UX 10.20 # . # . # # . ? . . # # . . . . # #
? #
dnl IRIX 6.5 # . # # # # . # . . . # . . . . # .
. .
dnl OSF/1 5.1 # . # # # # . . . . . # . . . . # .
. #
dnl OSF/1 4.0d # . # # # # . . . . . # . . # # # #
# #
dnl NetBSD 5.0 . . . # # . . . . . . # . # . . . .
. .
dnl NetBSD 4.0 . ? ? ? ? ? . ? . ? ? ? ? ? . . . ?
? ?
dnl NetBSD 3.0 . . . . # # . ? # # ? # . # . . . .
. .
dnl Haiku . . . # # # . # . . . . . ? . . ? .
. .
dnl BeOS # # . # # # . ? # . ? . # ? . . ? .
. .
dnl old mingw / msvcrt # # # # # # . . # # . # # ? . # # #
. .
dnl MSVC 9 # # # # # # # . # # . # # ? # # # #
. .
dnl mingw 2009-2011 . # . # . . . . # # . . . ? . . . .
. .
dnl mingw-w64 2011 # # # # # # . . # # . # # ? . # # #
. .
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library http://libvirt.org
8:47 p.m.
On 10/25/2011 05:11 PM, Simson Garfinkel wrote:
Oh my goodness. Okay. Let me clarify my question.
1. What should I put in configure.ac?
2. Can you show me specifically how to instrument my function, for example,:
void my_crazy_error_printer(void *random_arg,const char *format,...)
void my_other_crazy_printer(const char *fmt,...)
Well, that all depends. Is your function my_crazy_error_printer going
to call the system printf (where it matters which parameters you pass,
and for cross-platform compatibility to mingw, you can't use %lld, %zu,
%1$s, or a number of other useful constructs), or is it only going to be
used with a printf replacement (such as gnulib) that guarantees POSIX
compliance?
If the answer is the former, then use:
#if __GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 7)
# define ATTR_FORMAT(spec) __attribute__ ((__format__ spec))
#else
# define ATTR_FORMAT(spec) /* empty */
#endif
#define ATTR_FORMAT_PRINTF_SYSTEM(param, arg) \
ATTR_FORMAT ((__printf__, param, arg))
void my_other_crazy_printer(const char *fmt,...)
ATTR_FORMAT_PRINTF_SYSTEM(1,0)
and fix the warnings, because they are real bugs in the code (mingw
_will_ mishandle the format strings that gcc warned about).
If the answer is the latter, then use:
#if __GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 7)
# define ATTR_FORMAT(spec) __attribute__ ((__format__ spec))
#else
# define ATTR_FORMAT(spec) /* empty */
#endif
#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 4)
# define ATTR_FORMAT_PRINTF(param, arg) \
ATTR_FORMAT ((__gnu_printf__, param, arg))
#else
# define ATTR_FORMAT_PRINTF(param, arg) \
ATTR_FORMAT ((__printf__, param, arg))
#endif
void my_other_crazy_printer(const char *fmt,...)
ATTR_FORMAT_PRINTF(1,0)
and because you promised gcc that your printer obeys POSIX rules,
whether compiled on Linux or mingw, gcc won't blink an eye at constructs
that mingw normally chokes on. On the other hand, it _will_ warn if you
try %I64d, as that is a non-portable mingw extension not present in glibc.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library http://libvirt.org
4574
days inactive
4575
days old
4 comments
3 participants
participants (3)
-
Eric Blake -
Simson Garfinkel -
Simson Garfinkel