Re: Declaration of isspace in C/C++ not consistent?

[Mike Frysinger <vapier at gentoo dot org>]

On Sunday 02 June 2013 20:40:35 Hongxu Chen wrote:
> OndÅej BÃlka <neleai@seznam.cz> writes:
> > On Sun, Jun 02, 2013 at 11:42:11PM +0800, Hongxu Chen wrote:
> >> Mike Frysinger <vapier@gentoo.org> writes:
> >> > On Sunday 02 June 2013 10:45:21 Hongxu Chen wrote:
> >> >> Hi list,
> >> >> 
> >> >> Maybe this question is a bit silly, but I just cannot understand why
> >> >> `isspace` seems not consistent for C and C++(I have put this question
> >> >> in stackoverflow but no satisfactory answer has been given yet).
> >> >> 
> >> >> I am using *clang* analyzer to get the definition information and I
> >> >> know quite little about the mechanism behind it, so the declaration
> >> >> result might not be accurate; but I am just confused.
> >> >> 
> >> >> For c code like this:
> >> >>     // test.c
> >> >>     #include <ctype.h>
> >> >>     int main(int argc, char *argv[]) {
> >> >>     
> >> >>       isspace('a');
> >> >>       return 0;
> >> >>     
> >> >>     }
> >> >> 
> >> >> clang reports below as the declaration of isspace:
> >> >>     # define isspace(c)	__isctype((c), _ISspace)  // LINE 207 in
> >> >> 
> >> >> /usr/include/ctype.h
> >> >> 
> >> >> and when for this snippet:
> >> >>     // test.cpp
> >> >>     #include <cctype>
> >> >>     int main() {
> >> >>     
> >> >>       std::isspace('t');
> >> >>       return 0;
> >> >>     
> >> >>     }
> >> >> 
> >> >> clang reports the declaration here:
> >> >>     __exctype (isspace);  // LINE 120 in /usr/include/ctype.h
> >> >>     // #define	__exctype(name)	extern int name (int) __THROW
> >> >> 
> >> >> So why should there be such a difference?
> >> > 
> >> > glibc provides ctype.h which follows POSIX:
> >> > http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/ctype.h.html
> >> > we provide both real funcs and macros for each because the standard
> >> > allows it, and the macro ends up producing better code at runtime.
> >> 
> >> You mean that the c code is a macro implementation and it generates
> >> better runtime binary while C++ code uses the function one? By saying
> >> macro you mean `__ctype_b_loc'?(Actually I don't know what this symbol
> >> does)
> >> 
> >> # define __isctype(c, type) \
> >> 
> >>   ((*__ctype_b_loc ())[(int) (c)] & (unsigned short int) type)
> >> 
> >> Also there is another implementation called `__isctype_f', which is
> >> defined as:
> >> 
> >> # define __isctype_f(type) \
> >> 
> >>   __extern_inline int							      \
> >>   is##type (int __c) __THROW						      \
> >>   {									      \
> >>   
> >>     return (*__ctype_b_loc ())[(int) (__c)] & (unsigned short int)
> >>     _IS##type; \
> >>   
> >>   }
> >> 
> >> Then what's this supposed to be doing?
> > 
> > Replace function call by simple table lookup where table is 1 for
> > character inside class and 0 otherwise.
> 
> Would you please tell some more details? You mean __isctype_f
> implementation would actually lookup a _ISspacetype table and find
> whether `c' is 1 in the table? Or you are talking about __isctype?

__ctype_b_loc is a function call that returns the address of the data table.  
then the char is used an index into that table.
-mike

Attachment: signature.asc
Description: This is a digitally signed message part.