Drop misleading comments about undefined behavior

[cmb69]

There is no undefined behavior here. If BitScan*() returns zero, the value written to the first parameter is undefined, but we return a reasonable value.

---

This came up in #17527.

While having a closer look, I found that zend_ulong_ntz() and zend_ulong_nlz() are documented to behave differently if the argument is zero. While the latter states that for zero the result is undefined, the former states that for zero, LEN would be returned:

php-src/Zend/zend_bitset.h

Line 47 in ce53dab

/* Number of trailing zero bits (0x01 -> 0; 0x40 -> 6; 0x00 -> LEN) */

While this is actually true on Windows, and for the fallback implementation, it is apparently not the case when __builtin_ctzl() or __builtin_ctzll() are used. From https://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html:

Returns the number of trailing 0-bits in x, starting at the least significant bit position. If x is 0, the result is undefined.

So at least the comment should be fixed, and I would put in assert regarding the pre-condition, but see #17506 (comment) (I'm not sure whether all callers use the function accordingly). Also, in that case the Windows implementation would not need to check the return value.

Possible alternative: define the functions for num == 0.

[nielsdos]

Right, the comment likely meant "written value to index is undefined" but confused this with UB...