From mboxrd@z Thu Jan 1 00:00:00 1970 From: ludo@gnu.org (Ludovic =?utf-8?Q?Court=C3=A8s?=) Subject: Re: Guile-Git, bytestructures, libgit2 and Guix 0.14.0 Date: Thu, 14 Dec 2017 10:00:23 +0100 Message-ID: <87d13hu1js.fsf@gnu.org> References: <201712082348.29845.paul@boddie.org.uk> <87r2rzoqo8.fsf@gnu.org> <201712121805.24675.paul@boddie.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Return-path: Received: from eggs.gnu.org ([2001:4830:134:3::10]:38809) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ePPNZ-0002r3-Ld for help-guix@gnu.org; Thu, 14 Dec 2017 04:00:43 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1ePPNU-0006AT-IF for help-guix@gnu.org; Thu, 14 Dec 2017 04:00:33 -0500 Received: from hera.aquilenet.fr ([2a0c:e300::1]:48796) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1ePPNU-00068F-3e for help-guix@gnu.org; Thu, 14 Dec 2017 04:00:28 -0500 In-Reply-To: <201712121805.24675.paul@boddie.org.uk> (Paul Boddie's message of "Tue, 12 Dec 2017 18:05:22 +0100") List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: help-guix-bounces+gcggh-help-guix=m.gmane.org@gnu.org Sender: "Help-Guix" To: Paul Boddie Cc: help-guix@gnu.org Paul Boddie skribis: > On Tuesday 12. December 2017 17.29.43 Ludovic Court=C3=A8s wrote: >>=20 >> Paul Boddie skribis: >> > I also found that the packaged version of libgit2 on Debian Jessie >> > doesn't export the git_libgit2_init symbol for some reason: >> >=20 >> > ice-9/boot-9.scm:106:20: In procedure dynamic-pointer: Symbol not foun= d: >> > git_libgit2_init >>=20 >> This suggests a bug in the Debian packaging (or an old version?). > > Yes, it looks pretty awkward. Debian packaging for Guile seems all over t= he=20 > place. I don't see the Guile-TLS package for Stretch (guile-gnutls), for= =20 > instance, which has kept me from just switching to that distribution vers= ion: > > https://packages.debian.org/search?keywords=3Dguile-gnutls IIRC they removed it because of a bug that didn=E2=80=99t get fix in time. >> But yeah, I can sympathize with all the pain you=E2=80=99ve gone through= to >> build all this by hand. That=E2=80=99s a chicken-and-egg problem: the p= ackage >> manager is here to help you sort this out, but you=E2=80=99re precisely = trying >> to build it. > > Right. If one is already on board, it's just a matter of staying on board= , but=20 > getting on board and also doing it from scratch gives us some opportuniti= es to=20 > see whether the approach still works conveniently. Yeah. >> Pjotr Prins reported earlier that we don=E2=80=99t have a good bootstrap= ping >> story here, and I come to realize what it means. ;-) > > Debian has a similar experience, which is why there are efforts to improv= e=20 > bootstrapping. As you know, I was also interested in cross-bootstrapping,= and=20 > there has also been some more interest in that within the Debian realm,=20 > perhaps because there have been a few new to-be-supported architectures=20 > recently and people have probably become more easily frustrated about the= =20 > problems they have encountered than they might have been before. > > I have some interest in building distributions from scratch, but I find t= he=20 > process rather opaque with Debian, which rather explains my interest in=20 > looking at Guix instead. Nice. Though to me there are really two sorts of bootstrapping issue: 1. Whole-distro bootstrapping as discussed at , and which is what Mes, stage0, and friends are attempting to solve (see ). I also think that the purely functional approach (Guix and Nix) makes the bootstrapping path much clearer than with the imperative approach (Debian, etc.). 2. Building Guix itself on a system that doesn=E2=80=99t have it yet, whi= ch is what we=E2=80=99re discussing here. Clearly we could solve #1 without solving #2 and vice versa. These are fairly orthogonal issues. >> > ERROR: In procedure memoize-variable-access!: >> > ERROR: gzip: unbound variable >> > Makefile:5295: recipe for target 'make-go' failed >>=20 >> I=E2=80=99m pretty sure we solved it recently but I forgot how. Do you = have the >> problem on current master? > > This is with 0.14.0. I could try with the current master when I'm done tr= ying=20 > to build 0.13.0 bootstrap tarballs, hopefully successfully. ;-) OK, let us know how it goes! Ludo=E2=80=99.