Another test script:

Test script:
---------------

php -r '$a=( 103.24*6*10**2); echo $a,"=",intval($a),"=",intval(61944.0),"\n";'

Expected result:
----------------

61944=61944=61944

Actual result:
--------------

61944=61943=61944

that's indeed a joke

php > echo 103.24 * 6 * 10**2;
61944

php > echo round(( 103.24 * 6 * 10**2));
61944

php > echo (int)( 103.24 * 6 * 10**2);
61943

I have no idea what you tried in Python and Java, but it was clearly something else, because these are standard floating point semantics that are consistent across all languages. I've just checked `(int)(103.24 * 6 * 10**2)` on Python 3.6 and it indeed gives 61943, as expected.

If you don't want truncating conversion ("round down" for positive numbers), don't use integer casts, or ensure that the floating point number represents an exact integer beforehand, by using round().

> If you don't want truncating conversion

where are the decimal parts which could be truncated?

php > echo 103.24 * 6 * 10**2;
61944

I do not want to round() but only truncate decimal part, if one exists.

You're example well show that integer casting make lost one unit on a number without decimal :

php > echo (int)( 103.24 * 6 * 10**2);
61943

Perhaps it's not a bug, I've just to learn the well use of float.

Thanks for your answer.
Cheers.

In Java

    float a = (float) (103.24 * 6 * Math.pow(10,2)) ;
    System.out.println( a ); // print 61944.0
    System.out.println( (int) a ); // print 61944, not 61943

@cyrille: The reason it "works" in Java is that you're using the float type instead of the double type. If you use the double type instead, you will receive 61943 as the result. Floating-point numbers in PHP are always double-precision, not single-precision.

@rhsoft: Floating-point to string conversion is controlled by the "precision" ini settings. If you would like to see all round-trip relevant digits, set precision=-1. In this case, this will display "61943.99999999999" rather than "61944".

Thanks à lot @nikic

And sorry for noise.