Amber 
(kemona_halftau) 
kemona_halftau@sharkey.skydevs.me
@hexaheximal this is something you would do
1
0
0
@kemona_halftau it gets better.
I wrote the lexer, parser, and codegen code entirely from scratch. in C. it only depends on a host C compiler and libc.
1
0
0
@kemona_halftau also it is currently less than 60kb for the entire compiler program which is pretty cool, although it will definitely increase over time.
at least with the current (prototype) c implementation of the compiler, I don't think the compiler will ever be larger than 500kb.
1
0
0
hexaheximal
hexaheximal@mstdn.social@noisytoot @kemona_halftau I was referring to the compiler executable, but the source code is very small too.
however, once I rewrite the compiler to make the language self-hosting, the source code and binary size will definitely grow quite a bit, as that's when I will implement all of the heavyweight features such as native compiler backends - the c implementation of the compiler just transpiles code into c.
1
0
1
Noisytoot
noisytoot
1
0
1
@noisytoot @kemona_halftau I'm actually going to take an approach similar to zig, but without webassembly. the c implementation of the compiler will be left unmaintained with no compatibility guarantees once the self-hosted compiler is ready, but the self-hosted compiler will be able to emit standard C code that you can use to bootstrap the self-hosted compiler from.
(in practice, this means that you would be able to bootstrap the latest version directly from any standard c toolchain.)
1
0
1
Noisytoot
noisytootI prefer hare's approach of having a maintained bootstrap compiler written in c and ensuring that the self-hosted compiler remains compilable with it
1
0
1
hexaheximal
hexaheximal@mstdn.social@noisytoot @kemona_halftau you will (obviously) still be able to bootstrap the compiler from the c implementation by compiling every intermediate compiler version with major breaking changes, but you won't have to, and I'd argue you probably shouldn't.
because the compiler is deterministic, trust is not a practical issue - it's reproducibly emitting c code, so it would not be very difficult to read it.
1
0
1
Noisytoot
noisytootShow content
@hexaheximal @kemona_halftau guix takes the long (full source bootstrap) path for zig, it probably would here too
> grep '(define zig' gnu/packages/zig.scm
(define zig-0.10-libc-abi-tools
(define zig-0.10
(define zig-0.10.0-538-source
(define zig-0.10.0-539-patch
(define zig-0.10.0-542-patch
(define zig-0.10.0-610
(define zig-0.10.0-675
(define zig-0.10.0-722
(define zig-0.10.0-747
(define zig-0.10.0-748
(define zig-0.10.0-851
(define zig-0.10.0-853
(define zig-0.10.0-961
(define zig-0.10.0-962
(define zig-0.10.0-1027
(define zig-0.10.0-1073
(define zig-0.10.0-1497
(define zig-0.10.0-1505-source
(define zig-0.10.0-1506
(define zig-0.10.0-1637-source
(define zig-0.10.0-1638
(define zig-0.10.0-1657
(define zig-0.10.0-1681
(define zig-0.10.0-1712
(define zig-0.10.0-1713
(define zig-0.10.0-1888
(define zig-0.10.0-1891
(define zig-0.10.0-2558
(define zig-0.10.0-2565-source
(define zig-0.10.0-2566
(define zig-0.10.0-2571
(define zig-0.10.0-2796-source
(define zig-0.10.0-2797
(define zig-0.10.0-2824-source
(define zig-0.10.0-2838
(define zig-0.10.0-3660
(define zig-0.10.0-3726
(define zig-0.10.0-3728
(define zig-0.10.0-3807
(define zig-0.10.0-3813
(define zig-0.10.0-3980
(define zig-0.10.0-3985
(define zig-0.11.0-149
(define zig-0.11.0-384
(define zig-0.11.0-494
(define zig-0.11.0-587
(define zig-0.11.0-631
(define zig-0.11.0-638
(define zig-0.11.0-702
(define zig-0.11.0-761
(define zig-0.11.0-1967
(define zig-0.11.0-3245
(define zig-0.11.0-3344
(define zig-0.11.0-3501
(define zig-0.11.0-3503
(define zig-0.11.0-3506
(define zig-0.11.0-3604
(define zig-0.12.0-109
(define zig-0.13.0-286
(define zig-0.13.0-503
(define zig-0.13.0-1323
(define zig-0.13.0-1528
(define zig-0.13.0-1951
(define zig-0.13.0-1952
(define zig-0.13.0-2795
(define zig-0.13.0-2899
(define zig-0.13.0-2924
(define zig-0.13.0-2925
(define zig-0.13.0-3252
(define zig-0.14.0-0
(define zig-0.14.0-687
(define zig-0.14.0-877
(define zig-0.14.0-930
(define zig-0.14.0-934
(define zig-0.14.0-1091
(define zig-0.14.0-1197
(define zig-0.15.0-0
(define zig-0.15.0-1447
(define zig-0.15.0-1459
(define zig-0.15.0-2738
(define zig-0.15.0-2807
(define zig-0.15.0-2821
(define zig-0.15.0-2876
(define zig-0.11-libc-abi-tools
(define zig-0.12-libc-abi-tools
(define zig-0.13-libc-abi-tools
(define zig-0.14-libc-abi-tools
(define zig-0.15-libc-abi-tools
(define zig-0.16-libc-abi-tools
hexaheximal
hexaheximal@mstdn.social@noisytoot @kemona_halftau in my opinion, transpiling it to C would still be a form of source code bootstrap - it's just a different language.
specific distributions may choose the more painful route if desired, but that's their problem, not mine. my goal is to make this a language that is relatively easy to use, and part of that is making it easy to bootstrap the compiler.
1
0
1
Noisytoot
noisytoot
1
0
1
hexaheximal
hexaheximal@mstdn.social@noisytoot that is true, I'll admit. I didn't think of that.
I think it will probably just be like zig then - quick bootstrap path for most users, and a chain of historical versions for full source bootstrap for those who want it.
0
0
1
MuhammadFree🇵🇸
MuhammadFreeSoftware@fosstodon.org@hexaheximal I saw it yesterday while browsing your codeberg profile. I'm working on an interpreter at the moment
0
0
0