IRC channel logs

***ChanServ sets mode: +o rekado_

***rekado_ is now known as rekado

<fossy>hey janneke does guix currently bootstrap bison?

<janneke>fossy: "currently?"

<fossy>janneke: yes, because I recall a bootstrap path for bison someone made hete

<fossy>here

<fossy>so does guix bootstrap bison?

<janneke>fossy: ah, i'm confusing myself

<janneke>fossy: yes, guix bootstraps bison

<fossy>janneke: good.

<gio>janneke: How does bison bootstrap work? How is bison's own grammar compiled before having bison on guix?

<civodul>janneke: does it? :-)

<OriansJ>actually I think guix just does the traditional gcc build of bison using the generated input file; which would mean that bison isn't properly bootstrapped.

<markjenkinsznc>Bison is also partially written in bison eh? We can fault them for not having a bootstrap, but I guess people would also fault them if they didn't eat their own pemmican

<janneke>fossy: oh wait, bison actually contains "generated code" aka an ascii bootstrap seed?

<janneke>i wasn't really thinking beyond: nope, no bison bootstrap binary in guix

<markjenkinsznc>Okay, I jumped to the wrong idea too fast, having taken a look I don't see evidence any parts of bison are written with bison. I guess OriansJ may have been referring to use of autoconf/automake outputs

<markjenkinsznc>Bison source does have files ./bootstrap and bootstrap.conf so that's a good sign

<dddddd>I guess this is what triggered the questions but, for those who missed it, gio recently linked to https://gitlab.com/giomasce/bison-bootstrap

<gio>markjenkinsznc: File parse-gram.y is written in Bison. The distributed tarball ship its compiled form parse-gram.{ch}, but that's not "source code", at least according to my definition of "source code".

<gio>File boostrap and boostrap.conf have nothing to do with source code bootstrapping.

<markjenkinsznc>Ahh thanks, https://git.savannah.gnu.org/cgit/bison.git/tree/src/parse-gram.y , Bison does eat it's own pemmican. Watch out for zombie cannibal buffalo with preserved meat/berry packs.

<markjenkinsznc>During a pandemic.

<gio>As dddddd just mentioned, however, I worked out a path around this, so guix can pick it up if they want.

<gio>And if someone wants to independently check it, that would be wonderful.

<gio>BTW, flex has the same problem. I also have a boostrapping path for flex, although that's less nice.

***lfam_ is now known as lfam

<OriansJ>gio: certainly appreciate the work on flex and bison bootstraps

***deesix_ is now known as deesix

***dddddd_ is now known as dddddd

<fossy>gio: I have done the bison bootstrap, it works.

<fossy>I would love to see the flex bootstrap :)

IRC channel logs

2020-04-08.log