IRC channel logs
2024-12-04.log
back to list of logs
<aggi>finally, got the o1 scheduler and nptl patchset rebased onto linux-2.4.37.11, and it seems to work, musl-1.1 seems satisfied; so i can throw tcc-devutils.squashfs at it <aggi>smp isn't stable yet, but nosmp is stable, and too ioapic and usbcore get attached as these should <matrix_bridge><cosinusoidally> yep was that like 3 hours before the messages appeared on irc <matrix_bridge><cosinusoidally> seems quicker now. aggi congratulations on getting the backports working with the 2.4 kernel. <mid-kid>significant improvement over the -2024 version <mid-kid>but won't work as-is due to some sources having become unavailable in the gentoo distfiles mirrors, and the https support isn't quite there <mid-kid>you'll have to download some manually <mid-kid>In the long run I want my bootstrap to be future-proof <homo>oriansj: hi, is your fork, blynn-compiler, intended to bootstrap ghc? btw, the license of the original has changed from GPL to MIT <homo>sorry, not MIT, but 3-clause BSD <stikonas>homo: that fork is not actively developed right now <stikonas>it was mostly worked on before M2-Planet was able to build mes <stikonas>some people thought that M2-Planet -> blynn-compiler -> scheme interpreter in haskell -> mes might be easier than M2-Planet -> mes <stikonas>original repo might be targeting bootstrapping ghc, but I guess they are nowhere close? <matrix_bridge><googulator> IMO for bootstrapping Haskell, nhc98 or Yale Haskell will be a lot more useful - Haskell, in the form that it's used in modern GHC, is not well-defined enough to write an alternative implementation for ("Haskell" is basically "whatever the latest GHC implements", and that's also what GHC's own source code targets), and I doubt blynn is targeting early Haskell versions relevant to bootstrapping old GHC <matrix_bridge><googulator> It's a very similar problem to Rust, except that the language itself is a lot more difficult to understand <matrix_bridge><googulator> but worst of all, GHC has a quite recent gap in its bootstrap story between 7.4 and 7.6 <matrix_bridge><googulator> not counting intermediate Git commits, nothing older than 7.6 can build 7.6 or newer, at least not in a way that results in a self-hosting compiler <matrix_bridge><googulator> 7.2 and 7.4 can build 7.6, but the resulting binary can't recompile itself <matrix_bridge><googulator> and already by the time of 7.x, Haskell has long left behind the idea of a standardized language with a full specification and potential for multiple independent, conformant implementations <matrix_bridge><googulator> so, one needs to find a powerful enough, bootstrappable compiler for Haskell 1.2, 1.3, 1.4 or 98 that can compile old GHCs written in those Haskell versions - and then find a list of intermediate Git revisions between 7.4 and 7.6 that can bridge the gap <matrix_bridge><googulator> And that's hoping that it's not an Autoconf-esque breaking commit (one that depends on its own changes for successfully compiling itself) <stikonas>well, one can still target some specific and newish version of GHC for bootstrapping... <matrix_bridge><googulator> Given the immense complexity of what needs to be implemented to compile a newish GHC, it seems infeasible <homo>googulator well, guix comes with hugs and nhc98, but those are not enough, for example, to run darcs :( <homo>on top of that hugs is not available because for some reason gcc 4.9 fails to build, and nhc98 is built from generated C files, as nhc98 is also written in Haskell <homo>also why target bootstrapping old GHC instead of the most recent one? doesn't it become much harder to bootstrap on new arch and os combination when there are so many versions? <fossy>ahhh, that is what i was looking for mid-kid, thank you :D