IRC channel logs

<mid-kid>scheme's syntax is exacly what makes it so easy to implement

<mid-kid>you can't have everything

<Googulator>I get the underlying idea, but please, break compatibility with emacs's lisp mode :(

<Googulator>use anything other than ()

<Googulator>S-expressions are not the problem, the culture around them is

<Googulator>tbh, this would be perfectly readable:

<Googulator>(a (b c) (d

<Googulator> (e f)

<Googulator> )

<Googulator>)

<Googulator>this isn't:

<Googulator>(a (b c) (d

<Googulator> (e f)))

<Googulator>(toy example, so it doesn't really come across)

<Googulator>imagine a similar scenario where the ending would have 13 parens rather than 3

<Googulator>and then try to figure out where the form beginning with (d ends

<Googulator>as in, how many parens do you need to select at the end to correctly copy the form to a different context

<Googulator>pure pain

<Googulator>vs if everything were aligned K&R-style: easy

<andydude>Googulator: you would probably like Wren

<andydude>Also, you would probably like this https://gist.github.com/andydude/50c81f693cc74c366070be62149aad02

<lanodan>stikonas: Last time I did an strace on tcc it was in fact generating assembly for some reason (and I guess just pipes to itself).

<stikonas>lanodan: no, it really does not

<stikonas>you can easily see it in the source

<stikonas>let me find

<stikonas>e.g. https://repo.or.cz/tinycc.git/blob/HEAD:/i386-gen.c#l822

<stikonas>but it's full of these, this is just one example

<stikonas>its assembly backend is separate

<stikonas>so there is C->bin and asm->bin parts

<stikonas>so there is some duplication...

<lanodan>Uh indeed doesn't for at least simple C files, I wonder if I misread strace or was a case of like inline assembly

<oriansj>well M2-Planet's output is just a direct 1:1 state machine translation from C to assembly; so we do miss out on many potential optimizations but honestly running at 70% of TCC's compiled output isn't half bad for a self-hosting compiler that can do 1MLoC/second

<oriansj>but yeah, fixing TCC's poor readability could take years of work but would give a path forward to making it a much better compiler.

<andydude>I like improving readability

<andydude>what is constraint_priority

<andydude>the phrase "constraint priority" appears zero times in the tcc docs

<andydude>I found something, but not in the tcc docs, it was in the GCC docs

<andydude> https://gcc.gnu.org/onlinedocs/gcc/Simple-Constraints.html

<matrix_bridge><cosinusoidally> tbh I don't think tcc's poor readability will ever be fixed. Unless there is a hard fork under a completely different name.

<matrix_bridge><cosinusoidally> The current maintainer of tcc doesn't seem to have the time or motivation to fix the readability issues.

<matrix_bridge><cosinusoidally> Which is fair enough if they have other priorities.

<matrix_bridge><cosinusoidally> (and they are volunteering)

<matrix_bridge><cosinusoidally> a volunteer I mean

<matrix_bridge><cosinusoidally> Interestingly the pnut devs are working on the ability to compile tcc https://github.com/udem-dlteam/pnut/issues/63#issuecomment-2294154308

<matrix_bridge><cosinusoidally> so that could eventually be another compiler capable of compiling tcc

<lrvick>Has anyone managed to bootstrap haskell yet?

<lrvick>I think it is going to need some mrustc heroics last I checked.

<oriansj>lrvick: only a subset of the haskell language thus far (as far as I am aware)

<oriansj>with the blynn-compiler being the current leader (but very far from GHC compatibility and able to be used to bootstrap GHC)

<lrvick>Shame. It seems like the only major language we -can't- bootstrap right now.

<lrvick>but we have a hard requirement for bootstrappability/reproducibility in stagex, so will just tell people to file issues with the haskell team, not me. ha ha.

<lrvick>I thought guix had a similar requirement, which is why I was suprised to see it in guix.

IRC channel logs

2024-08-18.log