Hacker News Reader

I encountered problems on some platforms in the 90s where zeroing the bits of an array of pointers or floats/doubles did not give you an array of NULL pointers or 0.0 values. The comp.lang.c FAQ file discussed this in detail.

View on HN

Yeah, C doesn't require null pointers to be represented as 0, even if its constant is written 0, nor does it require floats/doubles to be represented as per IEEE-754. Or at least it didn't use to.

I don't think that's standards compliant.

At least for floating-point, not entirely sure about pointers.

floats being IEEE compliant is optional by the language standards. Otherwise the bit representation is not specified.

It still requires 0 to be represented as all bits set to zero.

Where does the ISO C standard say so?

edit:

https://cigix.me/c17#6.2.6.1.p1 seems relevant, I did not find anything for floating point types.

edit2:

Notes are probably not normative, but still:

https://cigix.me/c17#footnote301

That footnote confirms the traditional position that a floating point zero need NOT have an all-zero bit representation.

Yes. I was porting code to new machines and programmers had assumed it was compliant but it wasn’t.

Reply on news.ycombinator.com