thephd, 4 months ago

If you were wondering it was the most inflammatory line possible and tbh I've been building up a lot of gasoline-on-fire blogposts for a while so like, not only absolutely expected but totally deserved:

https://thephd.dev/intmax_t-hell-c++-c

(It was the conclusion header: "C++ is a better C")

slyecho, 4 months ago

@thephd ```
typedef __int128_t longmax_t;
typedef __int256_t longlongmax_t;

:welp:

muvlon, 4 months ago

@slyecho @thephd intmax_t_final_2_FINAL.pdf

slyecho, 4 months ago

@muvlon @thephd All this time we've been thinking about length-prefixed arrays and strings, why have we not thought of length-prefixed integers???

resuna, 4 months ago

@slyecho @muvlon @thephd

I had some experience with PLI before C and wondered why C didn't use bitsize types. Then I had todeal with porting software from 36 bit PLI to 32 bit PLI and had my answer. SO MANY "DCL FIXED (35)" that had to be sorted out.

slyecho, 4 months ago

@resuna @muvlon @thephd any language after 1999 can't do int, all they know is uint32_t, i32, f64, make chip hot & overflow

resuna, 4 months ago

@slyecho @muvlon @thephd

If you don't have 64 bits you're not playing with a full DEC.

resuna, 4 months ago

@thephd

If the ABI is broken then the major version number of the library must be changed, so existing code doesn't link to it. Change the version number on libc and put a new imaxabs in the new library.

If you're on Windows and this requires changing the name of the C runtime, that's on you.

If the FSF or whoever has broken the ability to change the version number on glibc, then glibc needs to burn.

Edit: that glibc needs to burn can be taken as a given.