halcy 
okay but has someone written like. a contrarian c compiler. a c compiler that on purpose, whenever anything is left up to the implementation, does whatever GCC and LLVM don't?
Jason Bowen 🇺🇦@halcy I've wanted to do this. Just do absolutely absurd, horrible things with lots of side effects.
Matt Might
byte 
-> @@halcy what a horrible idea. I like it
Mrs Beanbag@halcy yeah, microsoft
Alaric Snell-Pym@halcy I think @womble once told me about doing something similar in order to torture students. Maybe involving a 31-bit ones-complement processor with downright creepy memory layout, implemented in an FPGA, or am I conflating memories? 🤔
Petr Tesarik@kitten_tech @halcy And CHAR_BIT != 8 and sizeof(void*) != sizeof(long), of course… All of this does exist somewhere, but AFAIK nobody has done all this in one system.
If you want to get real fancy, make an implementation for Setun (a ternary computer):
https://en.wikipedia.org/wiki/Setun
Emelia/Emi@halcy Like when Clang goes "I'mma optimize this null check away because spec says I can assume all uses are non-null, and you accidentally 'used' it three lines earlier"?
Personally I'd go for the "chaos compiler" where whenever something is left to the implementation it just picks a random behavior, including intentionally doing something absolutely nonsensical if the given input results in undefined behavior (it's not against spec for a compiler to insert bounds-checks if it knows the size of the array for certain... it's just that C's ABI makes it next to impossible for the compiler to know it)
Martin Uecker@becomethewaifu @halcy I would say the insistence of programmers to use an ancient style of C instead of something as young as 2 decades often makes it next to impossible for the compiler to know the bounds:
https://godbolt.org/z/5hsTe8bP1
(but I agree there also some language problems which prevent one from doing this consistently)
https://godbolt.org/z/5hsTe8bP1
(but I agree there also some language problems which prevent one from doing this consistently)
Ridley @ WATCH LYCORECO
Officially Tara 🕷️ 
🌹@halcy malicious C compiler that does the worst possible thing with every unspecified behavior
Félix@halcy I wanted to write an adversarial C compiler that tries to screw with you whenever the standard says it can. I believe this will be the final stage of compilers anyways,
Drahflow@halcy c.f.
https://web.archive.org/web/20080419132129/https://en.wikipedia.org/wiki/DeathStation_9000 (but doesn't actually exist)
https://web.archive.org/web/20080419132129/https://en.wikipedia.org/wiki/DeathStation_9000 (but doesn't actually exist)
ewhac@halcy
You could fsck a lot of code over by making `sizeof (int) == 2`
That was a lot of fun on the Amiga (SAS vs. Manx compilers). To this day, I still think carefully before using `int`.
PoroCYon
Matteꙮ Italia
🔏 Matthias Wiesmann@halcy if someone writes such a compiler, here is a proposal for complex number implementation…. https://wiesmann.codiferes.net/wordpress/archives/36840
K. Reid Wightman 
🌻 
@halcy gccc. "whenever there are cases of undefined behaviour in the C standard, we repeatedly write 0x41 to the target of the operation..."
Sven Dowideit@halcy the short time I played with it, the Unisys 2200 c compiler felt like it was written in a contrarian way.
Pretty much nothing could be ported to it
Pretty much nothing could be ported to it
none gender with left politics
The Penguin of Evil@halcy gcc 1.34 interpreted #pragma as "play hack or rogue"
There are some good real world ones - pointer operations that can fault if the pointer is not valid, ones complement, twos complement but overflows cause an exception. For pre ANSI C the Honeywell L66 was fun. One config was 5 7bit chars and a spare bit per machine word, which sure made "sizeof()" interesting
When gcc first started using all the pointer aliasing opportunities that also created mayhem.
Exa (compte bientôt supprimé)@halcy I remember something like this but I don't remember if it was a C compiler Basically all undefined behaviour was random.
Sam Wilson@halcy I've wanted to build that, but even further: an asshole C compiler on top of an asshole POSIX architecture.
Len
KeyJ@halcy Microsoft did!
Syd Polk (he/his)