"multiple compilers for a single language is bad because they aren't fully compatible" yea and making the code works on multiple of them, the ones that might be used, is your job as a developer

[srazkvt@tech.lgbt]

"multiple compilers for a single language is bad because they aren't fully compatible" yea and making the code works on multiple of them, the ones that might be used, is your job as a developer

[srazkvt@tech.lgbt]

anyone scream at web devs if they go "sorry we don't care about firefox, only chromium, they don't have perfect compatibility" but when it comes to literally any other codebase suddenly it's not expected anymore

like, webdevs write some utterly fucked up architecture, sure, and i'm not defending them for everything, but double standard much ?

[q66@gts.q66.moe]

@SRAZKVT people only care about portability when it affects their own usage, more at 11

it's not like it's not exactly like this with *anything*

[srazkvt@tech.lgbt]

@q66 yea but idk i was kinda hoping people were better, maybe that's hoping for too much though

[q66@gts.q66.moe]

@SRAZKVT

the number of times i've seen cases like someone whining that someone support their thing only to immediately deride someone else's case with "who cares noone even uses your shit"

[hypha@cafe.mycelium.locahlo.st]

@SRAZKVT @q66 i wonder if there is code only working on tcc and failing with gcc

[srazkvt@tech.lgbt]

@hypha @q66 well it isn't impossible, some feature might be easy to add to tcc, but significantly more complex to add to gcc, hence some version of tcc could have it but not gcc

generally though, if it works on tcc, it'll work on gcc

[q66@gts.q66.moe]

@hypha @SRAZKVT stuff containing UB that doesn't get wrecked by having a too advanced optimizer?

[srazkvt@tech.lgbt]

@q66 @hypha i wouldn't consider that to be working code, but code that fails succesfully with tcc

[q66@gts.q66.moe]

@SRAZKVT @hypha you can probably make even those work with gcc/clang using -O0 and/or other shenanigans anyway ig

[meluzzy@woof.tech]

@SRAZKVT "But you see there is this really fancy CSS thing they added that only works on chrome now and Firefox won't add it until they make a version that actually follows the specification unlike chromium. Sure we used to make websites compatible with IE6 , Chrome, Firefox, Safari and even old Opera but now we have gotten used to how comfy new coding features are that we are never going back to thinking in more than 1 browser, also we only support versions of that browser that are less than 1 year old because if you haven't updated you are just a security issue and thats the perfect excuse because we can make it look like we care about security and that's why we only allow access if you use the latest version of the browser. Isn't that neat?.
Here have a website full of brand new webgpu wasm code that definitely will not cause any security issues in the near future. Dont try to run it on Firefox because it doesn't support --pseudo-webkit content: content-fit( --deluxe ? auto : full-auto);"

[srazkvt@tech.lgbt]

@meluzzy yeah, webdevs writing non compatible websites that won't run anywhere other than chromium are shitty and not doing their job properly

but they are being told to fix it by people who wouldn't have done it either, just because they're in another branch, and that isn't fair, they should hold themselves to the same expectations as they have for others

[meluzzy@woof.tech]

@SRAZKVT Webdevs making the most convoluted frameworks possible that end up being used to make websites that work just like scrolling down a PDF with fancy animations inside.
If reinventing the wheel was a competition they would have won every single year. Not even Rust devs can compete against that.

And the whole "oh well my code doesn't work because you changed something so now you should fix it. However if it was me I would just ask you to fix it yourself instead because colaborating to achieve a compatible system is not a real thing in webdev"

[kirtai@tech.lgbt]

@q66 @SRAZKVT @hypha
This reminds me that gcc/clang will force enable --fast-math without warning f you enable -O3 which has issues for floating point maths.

[q66@gts.q66.moe]

@kirtai @SRAZKVT @hypha uh, no it won't? -ffast-math is never enabled unless explicitly, since it's not ieee754 compliant

under clang -O3 is identical to -O2 plus some optimizations with high compile time costs and low runtime benefits, under gcc -O3 is closer to clang -O2 but definitely no ffast-math either

[q66@gts.q66.moe]

@kirtai @SRAZKVT @hypha there is -Ofast which is basically -O3 plus non-compliant fp

[kirtai@tech.lgbt]

@q66 @SRAZKVT @hypha
I had an article about this but now I can't find it.
Dammit.

[q66@gts.q66.moe]

@kirtai @SRAZKVT @hypha well if you read an article about it that was definitely misinfo

neither of the major compilers enable it for any normal opt levels

[kirtai@tech.lgbt]

@q66 @SRAZKVT @hypha
I stand corrected then

[noisytoot@berkeley.edu.pl]

@hypha @SRAZKVT @q66 anything with a shebang line, which tcc ignores (to allow for scripting using tcc) but gcc does not