IRC channel logs

<wingo>good morning guilefolk

<manumanumanu>Hi there! How is the licensing situation when using guile libraries? Say, ice-9 match?

<manumanumanu>I'm releasing it as GPL anyway, but for future projects... All the libraries seem to be (l)gplv3

<wingo>manumanumanu: guile itself is lgplv3+

<wingo>so a work that is based on guile can only be redistributed in ways provided for by that license.

<wingo>section 4d of https://www.gnu.org/licenses/lgpl-3.0.en.html has the important bits

<manumanumanu>so I don't have to comply with the gpl of ice-9 match, but only with the lgplv3?

<wingo>either you distribute combined work as source, so that the combined work can be rebuilt with a modified guile, or you dynamically link to Guile somehow, so that the user can use an updated guile with your combined work

<wingo>ice-9 match is lgplv3+

<wingo>not gplv3+

<wingo>though, any lgplv3+ may be (at the user's choice) redistributed under the gplv3+

<manumanumanu>oh... I have misread

<manumanumanu>i read gplv3+ and thought "hmh... that's odd"

<wingo>it is a little confusing :)

<wingo>and even if you have been working in this area for years sometimes we forget what the licenses actually say :)

<manumanumanu>Anyway, what is the best way to get guile-2.2 without having to wait for my distro to package it?

<manumanumanu>is it a painful buuild?

<wingo>for me i think the easiest way as just a one-off thing is to use the guix pack

<wingo>see end of https://lists.gnu.org/archive/html/guile-devel/2017-03/msg00095.html

<wingo>it's a simple binary install

<wingo>second easiest would be to install guix as a user (as opposed to making guixsd your OS) then do guix package -i guile; guix is really worth looking at. it is great stuff. https://www.gnu.org/software/guix/manual/html_node/Introduction.html#Introduction

<wingo>finally you can just build it yourself from a tarball

<wingo>which is fine too

<wingo>this is the build dir on a server where i installed it:

<wingo>gc-7.6.0.tar.gz gmp-6.1.2.tar.lz guile-2.1.6 libatomic_ops-7.4.4 libffi-3.2.1 libtool-2.4.6 libunistring-0.9.7 readline-7.0

<wingo>gc-7.6.0 gmp-6.1.2 guile guile-2.1.6.tar.xz libatomic_ops-7.4.4.tar.gz libffi-3.2.1.tar.gz libtool-2.4.6.tar.xz libunistring-0.9.7.tar.xz readline-7.0.tar.gz

<wingo>well, that was with a prerelease, you get the idea anyway

<manumanumanu>wingo: I'll go with guix whenever I reinstall my already very botched up debian installation. I have had it running for over 6 years and doing just about anything will break the system in weird places.

<wingo>heh i understand

<mrnugget>Is there some documentation that describes how to compile Guile locally? I cloned the repository and want to build the master branch, but `./configure` doesn't exist.

<lloda>mrnugget: see HACKING

<lloda>start with ./autogen.sh

<lloda>could be clearer, I think

<ArneBab>manumanumanu: To test something, I simply install a release tarball and then use ./configure --prefix ~/.local

<ArneBab>(get the tarball, configure, make install)

<wingo>mrnugget: tarballs build faster than the git checkout fwiw

<mrnugget>lloda: thanks! Huh, don't know how I missed that.

<mrnugget>wingo: thanks for the hint!

<wingo>np :) tarballs also already have ./configure

<wingo>which is helpful :)

<mrnugget>Yep, that's what I noticed and then I was wondering why ./configure doesn't exist in the repo. But I guess the tarballs are real "releases" and not just packaged "git tags"

<manumanumanu>Got it up and running. There has been magic optimization dust sprinkled over this release. My lazy prime generator runs 2x faster.

<ArneBab>they build *much, much, much* faster ← thanks to pre-generation tricks from wingo

<wingo>manumanumanu: good to hear :)

<manumanumanu>And that is using full call/cc. Might dive in to the delimited continuations tonight

<manumanumanu>using chez's call/1cc (one-off continuations) made it about 15% faster. Ah! A night of continuation fun! What more can one ask?

<nalaginrut>wingo: is there any way to get the current linux version in Guile? Or I have to parse the result from uname from pipe?

<wingo>manumanumanu: if those are escape continuations then consider let/ec from (ice-9 control)

<wingo>nalaginrut: uname and pipe i think

<nalaginrut>ok

<nalaginrut>oh, there's uname in Guile, but I still have to parse

<wingo>interesting, didn't know about that

<nalaginrut>me too

<nalaginrut>I have to make sure the epoll features in Artanis is supported for users

<nalaginrut>it'll be checked when artanis read config

<nalaginrut>of course, not only epoll

<wingo>morning civodul

<civodul>hello wingo!

***Guest45045 is now known as nikita3

<manumanumanu>wingo: beautiful, thank you

<nikita3>nalaginrut: how to make Artanis serve multiple requests in parallel (having pool of db connections)?

<wingo>civodul: i fixed the constant literals thing!

<wingo>v pleased about that

<civodul>wingo: the readonly-ness of literal pairs & co.?

<wingo>civodul: yeah and vectors bitvectors bytevectors and arrays

<civodul>awesome!

<wingo>regardless of whether they contain relocatable data or not

<wingo>yeah, neat stuffs.

<lloda>wingo: is there a way to know if a value is mutable?

<lloda>other than trying I mean

<wingo>lloda: no. why would you want that?

<wingo>big arrays i guess?

<wingo>i just think it's a responsibility of the programmer who makes literal arrays to not mutate them

<wingo>i would like to eventually provide a facility to mark an existing value as immutable

<wingo>so that we could apply the same provisions about mutability to the evaluator, for example, or otherwise when you want to "freeze" data before passing it to an untrusted party

<wingo>it would be possible to do this for every value type except pairs :P

<wingo>being able to have a cons vs mcons would be nice.

<lloda>I have to fix my C++ wrappers. In some places I want to check if SCM can-be-converted, so T const vs T didn't matter before.

<lloda>doesn't mean I need mutable?, hopefully I don't

<lloda>there's lots of breakage

<wingo>2.2 breakage? some other kind?

<lloda>it was a necessary change of course, no question

<wingo>or breakage related to this specific thing

<lloda>yeah the read-only vars

<wingo>before it would segfault tho, right?

<lloda>sure, I'm 100% behind this fix

<wingo>ok, just wondering if it broke something else :)