IRC channel logs

2024-10-13.log

back to list of logs

<stikonas>oriansj: can you please merge https://github.com/oriansj/bootstrap-seeds/pull/45
<stikonas>I've reviewed new bootstrap seed, looks fine appart from some comments which I have now fixed
<stikonas>some of that stuff in this hex0 seed can be ported to other arches...
<stikonas>I was aware of a couple of possible tricks before but this new seed goes far beyond that...
<stikonas>well, 190 bytes on x86 means only 106 bytes of actual code
<stikonas>though some of the tricks are x86 specific there
<stikonas>in a couple of places some registers just have required values for x86 syscall numbers
<oriansj>stikonas: reviewed and merged.
<stikonas>thanks
<oriansj>well; there are a great many potential optimizations that could be done for the bootstrap-seeds (kaem-optional-seed could be shrunk down to under 256bytes if one got creative) I've opted for easy to reason about by default and left the clever optimizations to people better at code golf than I.
<stikonas>well, probably hard to reduce hex0_x86 seed further
<stikonas>but yes, kaem is not super optimized
<pabs3> https://pileofhacks.dev/post/the-optimised-version-of-7-zip-cant-be-built-from-source/ https://news.ycombinator.com/item?id=41824784
<pabs3>asmc isn't immediately bootstrappable
<matrix_bridge><Andrius Štikonas> It's assembler though ...
<matrix_bridge><Andrius Štikonas> How hard can it be to compile with another one
<matrix_bridge><Andrius Štikonas> Probably need to adjust makefiles and perhaps some asm syntax
<pabs3>from #debian-devel <helmut> I happen to have looked into nim packaging. My understanding is that nim can be compiled to C and the nim source package includes the compilied C sources, builds nim from that and then uses nim to build nim. Is that actually a good way to bootstrap it?
<matrix_bridge><Andrius Štikonas> Though yes, it is an annoying case of poor documented build
<pabs3>ACTION mentioned https://bootstrapping.miraheze.org/wiki/Bootstrapping_Nim to helmut
<stikonas>I'm not on that channel (#debian-devel) but I guess folks here would not consider that a good way of bootstrappng
<stikonas>thanks
<nimaje>that blog post says that asmc is a nasm compatible assembler, but only tries to replace it with jwasm, from which asmc was apparently forked from, not idea why nasm wasn't tried, shouldn't that be the first idea there?
<stikonas>yeah, my thought was to use nasm too before I've read that blog
<stikonas>so no idea either why it wasn't tried
<stikonas>dozer seems to be progressing nicerly with more than one contributor now: https://codeberg.org/notgull/dozer/commits/branch/main
<oriansj>pabs3: yeah Nim still requires a great deal more work before it can be properly bootstrapped. As generated C sources are clearly not human written nor the preferred form for modification.
<oriansj>stikonas: thanks for the link