IRC channel logs
2014-01-28.log
back to list of logs
<mark_weaver>I'm afraid that setting --localstatedir=/var is a disaster. <mark_weaver>lots of builds are now failing, because they fail to create /var <mark_weaver>I'm not sure how to deal with this. "make install" tends to want to create things in localstatedir. <civodul>yeah i just noticed and reverted :-) <civodul>actually i remember having to patch packages precisely because of that <civodul>so i suppose we may have to keep doing it on a case-by-case basis <civodul>oh, gdb release in 1 day, glibc release in 2 days <civodul>jmd: i pushed a "fix" that should solve the problem you had with downloading apr <civodul>i think it's even a fix, without quotes <jmd>Oh ok. Had the url moved, or did we have it wrong in the first place? <civodul>they move thing eventually to archive.apache.org <civodul>so i just added that URL to the 'apache' mirror list <jmd>I'm concerned that if an upstream moves their download site or rearranges it, it might take a long time before guix notices - 12 months later it could cause a lot of trouble finding it again... <jmd>It might not be a bad idea to have a process which pings the download sites periodically. <sriharsha>jmd, openembedded also suffers from this issue of download sites going offline. If anyone is going to FOSDEM, please ask them how they cope with this... <viric>ah, references. (I looked at the backlog) <viric>hydra could include a torrent tracker <civodul>sriharsha: no it's not obsolete, why would it be? <sriharsha>because I could not find lsh-service anywhere.. <civodul>aah, that's because it's actually forward-looking :-) <viric>be it nar files, be it direct tarbals <civodul>'lsh-service' is not implemented :-) <civodul>sriharsha: but you can find pretty much all the rest under gnu/system <sriharsha>you lost me.. but doesn;t it complain of a missing variable? <civodul>the manual itself doesn't complain, no ;-) <sriharsha>can you write an example of how a service configuration is handled? I want to add nginx <sriharsha>but it doesn't specify how configuration for the service is handled.. <civodul>look at 'syslog-service', for instance, in gnu/system/dmd.scm <civodul>it produces a config file, and start syslogd to refer to that file <civodul>maybe that's a better example for the manual <civodul>i'm really glad you're looking into this <sriharsha>what does (package-file inetutils "libexec/syslogd") <civodul>it returns the file name /.../...-inetutils/libexec/syslogd <civodul>(there's a docstring for 'package-file' in monads.scm) <sriharsha>thanks; actually, we are using guix for deploying gnunet experiments over wide variety of machines.. so, the quicker I learn this, the quicker we deploy.. besides, this is a fun and functional project :-() <civodul>if you're not using Geiser, you really should ;-) <civodul>sriharsha: beware, this part is work in progress! <civodul>at this stage, we can only instantiate VMs <sriharsha>I am using Geiser, but it doesn't show me any doc or take be to the function definition <civodul>sriharsha: try C-c C-k while visiting dmd.scm, and then C-c C-d d when the cursor is under 'package-file' <viric>we need someone to shot a documentary, "Life in Guix" <viric>those nixers think only of kde <civodul>a package manager with an IDE, can you imagine? :-) <sriharsha>civodul, it does work! I was used to gtags and was pressing M-. all the while <civodul>sriharsha: ah that should work too :-) <sriharsha>oh yeah, that's working too! So its just that I have to compile with Geiser first <civodul>live tomorrow: TZ=Pacific/Kiritimati date <jmd>So I want to have a few "private" packages which I'm not going to publish. Where is there recommended place to put them? <viric>Better don't pay attention to me. <jmd>I meant in what directory? <jmd>Is there a site-local or something? <mark_weaver>set GUILE_LOAD_PATH, and arrange for the packages to be in <DIR>/gnu/packages/*.scm where DIR is a component of GUILE_LOAD_PATH. <mark_weaver>alternatively, make a new local branch in 'git', and run ./pre-inst-env guix ... from your git checkout. <jmd>ok. I'll try the GUILE_LOAD_PATH approach <mark_weaver>the branch method is more powerful, because it also allows you to make local changes to existing packages. <mark_weaver>and then git will automatically handle merging upstream changes into your local branch. <jmd>git doesn't automatically merge branches you have to do it manually. <jmd>Now if we were using aegis ... <mark_weaver>well, most of the time it's automatic. occasionally there's a merge conflict, but in practice it works pretty well. <jmd>But you have to manually instruct it "git merge master" or simular. <mark_weaver>personally, I wouldn't want it to be more automatic than that. because nothing short of A.I. could handle the merges correctly in all cases, and so I want to control when it happens, in case some fixups are needed. <jmd>That is true. But I like aegis' approach where "branches" really are true branches. Not just parallel lines of development like in git. <mark_weaver>I'll have to take a closer look at aegis at some point. but not today; I'm too overloaded with Guile work. <mark_weaver>jmd: sorry, I'm not sure that the GUILE_LOAD_PATH approach will quite work correctly at present. <mark_weaver>It could be made to work properly, but it seems that some code in Guix currently assumes that the packages will all be in just one component of the GUILE_LOAD_PATH <jmd>civodul must have started his pilgrimmage early. <mark_weaver>Specifically, 'package-files' in gnu/packages.scm only searches for packages in the first component of the load path that contains gnu/packages.csm. <mark_weaver>that should really be fixed though. it should search for packages in all components of the load path. <mark_weaver>until that's fixed, I guess you'll need to use the git branch approach. <zerwas>sriharsha: do you make progress in packaging GNUnet in Guix?