IRC channel logs
2014-04-25.log
back to list of logs
<civodul>mark_weaver: did you have chance to look at my gexp prose? ;-) <mark_weaver>civodul: yes, I did. I don't fully understand it, and I think there may be better ways of doing it, but I'll have to write about it later. about to catch a train. *civodul goes for #~ and #$ in gexps <Fulax>so #~ is derivation-computation time, while #$ refers to guix-evaluation time ? <civodul>well i'd say "build side" for #~ vs. "host side" for #$ <Fulax>civodul: seems strangly familiar <civodul>there's a pattern here: R6 syntax-quote, HOP <Fulax>I know too litle about the build side, not enough guix usage on my side I thinkl <Fulax>do you think to allow #~ escapes within #$ escapes ? <bavier>civodul: there are some packages that need to reference particular versions of packages as inputs, but can keep that version reference confined to the inputs list, with a generic reference to the name later. With gexps, would one need to use the versioned package name everwhere it is referenced? <civodul>Fulax: i haven't thought much about it, but i don't see any use case <civodul>bavier: gexps shouldn't change anything in the basic paradigm <civodul>they won't be used for packages in the short term <civodul>but in the longer term, that would allow you to do things like this in 'arguments': <civodul>#:configure-flags #~((string-append "--with-foo=" #$foo)) <civodul>instead of doing (assoc-ref %build-inputs "foo") <civodul>and also that would work even if 'foo' is not listed in 'inputs' <civodul>and you'd get a compile-time warning if 'foo' is unbound <bavier>I see, because currently, you'd only get an error once the build is running <bavier>would the packages referenced in a gexp be implicitely added to inputs? <bavier>interesting, ok. thanks for the clarifications