IRC channel logs

2018-06-14.log

back to list of logs

<chrislck>from comment a couple days ago - I peeked into SRFI-48. omg.
<chrislck>sneek: botsnack
<sneek>:)
<chrislck>further to a comment couple days ago I peeked at srfi-48... omg.
***Raimondii is now known as Raimondi
***rubdos_ is now known as rubdos
<wingo>good day guilers
<rain1>hello
<janneke>hey wingo
<wingo>heya janneke
<wingo>congrats on all that awesome mes work :)
<janneke>ow, thanks!
<janneke>i'm very pleased it's starting to land, phew :)
<wingo>i can understand the relief :-)
<janneke>you have some experience building compilers and interpreters...
<wingo>it's horrible when everything is in the air and you can't trust anything!
<wingo>only way to learn tho :/
<janneke>hehe, yeah -- apparently
<janneke>at one moment, i managed to break less_p (written in C of course)
<janneke>it took me quite some time to find that, though -- imagine the results you get on anything numeric :)
<janneke>and this was a pretty harmless, small thing when most was already working
<wingo>ACTION gets mad at srfi-1.c, a file that should not exist
<janneke>ACTION cheers!
<janneke>i've been wondering about some .c files that Mes "just" has in scheme
<janneke>one thing that has bothered me is to write a new (and rather crappy) Scheme implementation in Mes, just because the aim is easy bootstrapping
<janneke>wouldn't it be wiser and just about as much work to find a way to make Guile bootstrappable
<janneke>eg, starting by using lots of #if !BOOSTRAP or something, and indeed, moving other .c bits to scheme
<wingo>janneke: yeah working on moving more c to scheme. a continual though slow process :)
<wingo>it's essential to be able to rebase on top of any other target, e.g. webassembly...
<janneke>wingo: i'm still hoping in a way that Mes will have been an important, but temporary thing, and we'll be able to create a `bootstrappable-Guile'...but that may take some time...years?
<wingo>yes i think so. i think the things i am doing, moving more to scheme, go in that direction -- but it is taking time :)
<wingo>i am not directly focused on the minimal bootstrap thing tho, could be there's a more direct path; but i guess mes is a fairly direct path there, so, cool :)
<janneke>yeah, thanks. it seemed too big a stretch just now, but maybe some time later...
<janneke>also it may help if mes shows this bootstrap route is working, that might support an effort to make Guile more bootstrappable -- it could well be that we manage to remove Mes/Scheme from the bootstrap path though -- we'll see
<rekado_>I would expect to see Mes being used to be the bootstrap interpreter in a C-less Guile.
<rekado_>*being used as
<civodul>though it's good to keep libguile/eval.c IMO
<janneke>rekado_: yeah, you put some effort into that when i visited Berlin, right?
<janneke>at that time, Mes didn't have recursive macro expansion, I added that in 0.14
<rekado_>yes, I remember that we got stuck because of macro expansion.
<rekado_>ACTION gets distracted too quickly :(
<rain1>isn't guile already bootstrappable
<rekado_>rain1: yes, via C.
<janneke>rekado_: hehe, so that path may be worth exploring again, some time
<rekado_>rain1: an alternative route to get to Guile might be to use Mes as the bootstrap interpreter, not the Scheme interpreter that’s written in C.
<wingo>libguile 3.0 is about 6% smaller than 2.2, but will get smaller still i think
<wingo>incrementally :)
<wingo>anyway, sleepytime
<wingo>night folks :)
<janneke>night wingo
<cmaloney>night wingo.