vitaut 🤍❤️🤍 🇺🇦Feb 22
So one side effect of C being a mess is that bool and char constants are formatted as integers by default. You have to either cast them to the correct type or use an explicit format specifier which is a bit annoying.
Show threadThiago AdamsFeb 22
@vitaut do you mean true/false constants?
They are _Bool type in C23
Char constants still int .
Show threadAndre WeissflogFeb 22
@thradams @vitaut _Bool goes all the way back to C99, C23 actually deprecated it in favour of 'bool' (eg. stdbool.h no longer needed), both are "proper" bools, but AFAIK the C standard doesn't enforce a specific size
Show threadAndre WeissflogFeb 22
@thradams @vitaut also I suspect that the problem might be that _Bool/bool implicitly converts to int (eg I guess in vararg lists, bools are passed as int, not sure though)
Show threadvitaut 🤍❤️🤍 🇺🇦Feb 22
@floooh @thradams I’m not using varargs, the problem is that true/false has int type in C11
Show threadvitaut 🤍❤️🤍 🇺🇦Feb 22
@floooh @thradams so it doesn’t work with _Generic
Show threadJoe Groff (1M Context)
@vitaut @floooh @thradams for back compat, how about your own <fmtbool.h> that re-#defines true and false as (_Bool)1 and (_Bool)0
Feb 22 at 2:49pmWeb
Show threadAndre WeissflogFeb 22

@joe @vitaut @thradams hmm, clang (and gcc) seem to get it right?

https://www.godbolt.org/z/1xd83zbeW

Compiler Explorer - C

void bla_bool(bool b) { printf("A boolean: %d\n", b); } void bla_int(int i) { printf("An int: %d\n", i); } #define bla(x) _Generic((x), \ bool: bla_bool, \ int: bla_int \ )(x) int main() { bla(true); bla(false); bla(123); return 0; }

Show threadAndre WeissflogFeb 22
@joe @vitaut @thradams ooopsi, wait Clang gets it wrong! One sec...
Show threadAndre WeissflogFeb 22

@joe @vitaut @thradams ah ok, apparently it has been fixed in C23, but C11 gets it wrong:

https://www.godbolt.org/z/1TW3Kndb7

Compiler Explorer - C

void bla_bool(bool b) { printf("A boolean: %d\n", b); } void bla_int(int i) { printf("An int: %d\n", i); } #define bla(x) _Generic((x), \ bool: bla_bool, \ int: bla_int \ )(x) int main() { bla(true); bla(false); bla(123); return 0; }

Show threadAlexander MonakovFeb 22

@floooh @joe @vitaut @thradams true/false are keywords in C only since C23 (macros before)

incidentally, sizeof(true) is 1 in C23, 4 before

Show threadMartin UeckerFeb 22
@floooh @joe @vitaut @thradams You need to pass -std=c23
Show threadJoe Groff (1M Context)Feb 22
@floooh @vitaut @thradams if you’re compiling as standard C99 or C11, it is “right” for true and false to be ints, since they are specified to be #define-d as 1 and 0 in stdbool.h
Show threadAndre WeissflogFeb 22
@joe @vitaut @thradams hmm, but `_Bool` was already introduced in C99, I would expect that true/false are of type _Bool, but maybe I'm wrong...
Show threadJoe Groff (1M Context)Feb 22

@floooh @vitaut @thradams until C23, true and false were just stdlib defines

#define bool _Bool
#define true 1
#define false 0

Show threadJoe Groff (1M Context)Feb 22
@floooh @vitaut @thradams the other annoying thing is that the comparison operators also all produce results of type `int` in C. I forget whether c23 fixes that as well
Show threadAndre WeissflogFeb 22
@joe @vitaut @thradams ...and then the next problem is that MSVC will probably never get C23 support (they should just be honest and declare MSVC as deprecated and switch Visual Studio to Clang as default compiler)
Show threadJonathan MüllerFeb 22
@joe @floooh @vitaut @thradams IIRC they didn't.
Show threadJonathan MüllerFeb 22
@joe @floooh @vitaut @thradams A character literal is also still int.
Show threadThiago AdamsFeb 22
@foonathan @joe @floooh @vitaut c compilers can track that by adding an extra type info “int” but “ essential bool”
Show threadvitaut 🤍❤️🤍 🇺🇦Feb 22
@thradams @foonathan @joe @floooh won't help with _Generic
Show threadMartin UeckerFeb 22

@vitaut @thradams @foonathan @joe @floooh
@JensGustedt

Yeah, we should just change this. But a compiler maintainer or some company sitting on legacy code that it wants remain conforming without maintenance effort will likely complain.

Show threadJoe Groff (1M Context)Feb 22
@uecker @vitaut @thradams @foonathan @floooh @JensGustedt it seems like wg14 is more willing now to accept lightly-breaking changes like new keywords. would this be the sort of change you all would accept under a new language version?
Show threadMartin UeckerFeb 22
@joe @vitaut @thradams @foonathan @floooh @JensGustedt Unclear. I try sometimes but some compiler vendors often protest vehemently against such changes. For example, I wanted to make null pointer constants have pointer types (which is a much more serious issue than this boolean issue), but there was a lot of opposition.