Hi all,
I changed the bitwise and modulus operator functions to [try to] make them
more consistent with behavior on 64-bit platforms. By accepting double
operands and, if necessary, returning a double, their usable range is
greatly extended without running into overflow issues, etc. (Of course, at
some point, double precision would be a problem, but that's not till... the
trillion range/40-ish bits or so?) See also Bug #46579 (% integer
overflow - no error message).
That means anything like this previously only-64-bit code works as expected:
var_dump(((1 << 40) | (1 << 36)) % (10000000000 >> 1));
float(3231104512)
Fairly straightforward changes after adding a couple macros... Adding
zendi_convert_to_double() to the other zendi_* ones seemed to be the easiest
way to handle things. And, am I right to assume that most all modern
compilers have a 64-bit long type? If not, code is unchanged from before
(same for 64-bit).
This is mostly independent of whatever is done as far as general "double to
long conversion," but it does take care of the orginal issue I had. :-)
http://marc.info/?l=php-internals&m=120799720922202&w=2
Patches:
http://realplain.com/php/64bit_operators.diff
http://realplain.com/php/64bit_operators_5_3.diff
What do you all think?
Thanks,
Matt
--
PHP Internals - PHP Runtime Development Mailing List
To unsubscribe, visit: http://www.php.net/unsub.php
