[llvm-dev] RFC: On removing magic numbers assuming 8-bit bytes

JF Bastien via llvm-dev llvm-dev at lists.llvm.org
Thu May 2 10:25:09 PDT 2019

> On May 2, 2019, at 10:21 AM, paul.robinson at sony.com wrote:
>> -----Original Message-----
>> From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of JF
>> Bastien via llvm-dev
>> I’m not a fan of C and C++ supporting anything but 8 bits per byte.
>> Realistically, C and C++ on such targets are different languages from 8-
>> bit-per-byte C and C++, and therefore code isn’t portable from one to the
>> other.
> Having done it, I promise you that it's reasonable to write portable C
> targeting both 7-bit and 8-bit 'char'.  It was too long ago to remember
> anything in detail, but the brain cells still remaining from that era
> believe it was pretty clean.

I agree it’s *possible*, same way I’ve seen some correct uses of volatile, and the same way I’m sure some code supported non-two’s-complement machines correctly. What I’m saying is that most code simply isn’t, often in subtle ways. The code you wrote is therefore a subset of C which is incompatible with C at large.

>> I intend to propose that C++23 support only 8 bits per byte, ditto
>> C. I’m therefore not a fan of teaching clang about this.
> My impression is that non-8-bit-byte machines are (these days) generally 
> small and likely for embedded or other special purposes, so a proposal 
> to stop trying to squeeze the bloated monster that C++ has become onto 
> those is probably fine. C, on the other hand, appears to be the language
> of choice for all sorts of weird things, and that's less likely to fly.
> (Just sayin'. I'm not on either committee and have no vested interest.)
> --paulr

More information about the llvm-dev mailing list