Pollita <[EMAIL PROTECTED]> wrote:
> log(0) in any base (except 0, which would be silly) is an undefined
> number.
Yes, that's what the teachers told us before they admitted the existance
of infinity.
> the libc log() function will return an exceedingly small number
> to avoid causing widepsread panic when log(0) is attempted
> (-1.7976931348623E+308 shown below).
Are you sure?
% cat log.c
#include <stdio.h>
#include <math.h>
int main()
{
printf("%g\n",log(0));
}
% gcc log.c -o log -lm
% ./log
-inf
That's on Linux. On Tru64, it prints:
-1.79769e+308
> The test function for log:
>
> $x2 = (int) pow($base, log($x, $base));
>
> attempts to check that $x2 is close to $x, but since log(0,*) will return
> an excessively small number, pow() winds up choking.
>
> Suggest: $x2 = (int) log(pow($x,base),$base); instead.
Thanks for the suggestion, but it dumps core nevertheless:
% sapi/cli/php -r 'echo log(pow(0,2),2),"\n";'
Floating point exception (core dumped)
% sapi/cli/php -r 'echo log(0,2),"\n";'
Floating point exception (core dumped)
> P.S. example below using log in base 1 is also nonsensical... only
> log(1,1) is a valid use of log with a base parameter of 1, any other value
> results in "undefined"
Umm, yes, maybe - but my point was that PHP should not dump core in such
situations, and it certainly should not do so in a "make test" ("Wait a
minute, what's that core file here? And now you want me to 'make
install'?!").
Regards...
Michael
--
PHP Development Mailing List <http://www.php.net/>
To unsubscribe, visit: http://www.php.net/unsub.php
