Re: [Mingw-users] Strange snprintf behaviour

[Danny S. <dan...@cl...>]

Greg Chicares
Wednesday, 18 July 2007 4:19 a.m.

> On 2007-07-12 23:30Z, Danny Smith wrote:
> > 
> > Recently added __mingw_snprintf and __mingw_vsnprintf  work 
> correctly,
> > per C99.
> 
> Thanks for this. I've tested it as follows, and it looks great.
> 
> (1) My production code has unit tests to validate snprintf(),
> because msvcrt isn't the only runtime that gets it wrong (e.g.,
> borland has its own errors). All those tests pass now that I've
> built libmingwex with CFLAGS='-DEXPORT_WEAK_SNPRINTF_ALIAS' and
> (as a temporary expedient) added

Weak alias support 'seems' to work most times, but I would not count on
it.  The PE_COFF characteristics for the weak symbol are not set
correctly in object file, so it is bound to fail wih MS linker.

>   extern "C" int __mingw_snprintf(char *s, size_t n, const 
> char *fmt, ...);
>   #define snprintf __mingw_snprintf
> to my code. I'm sure you have a much stronger testsuite already,
> but it never hurts to run a few more tests.
> 
> (2) As for speed, we don't want another "pow() is too slow"
> controversy, so I ran this function a hundred times:
>   void time0()
>   {
>     static char buf[100] = "zzzzzzzzzzzzz";
>     snprintf(buf, 12, "%11.6f", 1234.56789);
>   }
> and found that __mingw_snprintf took 30% longer; and it took
> 60% longer for this more-precise variant:
>     static char buf[100] = "zzzzzzzzzzzzzzzzzzzzz";
>     snprintf(buf, 20, "%19.14f", 1234.5678911111111);

Thanks for that.  I observed similar hit from using the gdtoa float
conversion
routine.


> That seems to be an acceptable price to pay for correctness.
> OTOH, using a 'd' conversion:
>     static char buf[100] = "zzzzzzzzzzzzzzzzzzzzz";
>     snprintf(buf, 20, "%19.14d", 1234.5678911111111);
> the new code ran in 40% less time than the msvcrt version.
> 
> So I look forward to seeing this in an official release whenever
> you get around to it. If msvcrt's _snprintf() remains available
> with its underscore-prefixed name, then is there any reason not
> to make the C99-conformant libmingwex version the default when
> the standard name snprintf() is used?
> 

IMHO, no.  

Danny

> --------------------------------------------------------------
> -----------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> MinGW-users mailing list
> Min...@li...
> 
> You may change your MinGW Account Options or unsubscribe at:
> https://lists.sourceforge.net/lists/listinfo/mingw-users
>