IRC channel logs


back to list of logs

<zimoun>rekado_: hi! I am working on the R update. But after 1 day fighting with Cuirass, chats with mothacehe to understand what I am missing, and then the all IO issues (package reported as failed but just because IO), I gave up. And I did on my local computer. Building is longer…
<zimoun>Moreover, I do not know how I did my count but I created I mess with missing deps. So I am fixing. And I do my way, i.e., I rewrite the history otherwise I am lost.
<zimoun>Well, the lesson I am learning is: it is hard nor impossible to do a big update such Bioconductor using Cuirass to check the error without commit access. :-)
<rekado_>I never used Cuirass to work on the CRAN/Bioconductor upgrades
<rekado_>the iteration cycle is just too slow
<rekado_>I prefer to build it all on my laptop and fix things as I encounter them
<zimoun>it is what I am doing now. But my machine is a bit slow…
<zimoun>that’s why I am roaming on old bugs :-)
<zimoun>rekado_, civodul: about security, previous year, I was at and I had a quick chat with people from They presented their distro CLIP OS based on Gentoo. I do not know if my words make sense. :-)
<civodul>zimoun: they do :-)
<civodul>i've heard about CLIP OS, would be nice to chat with these folks someday
<rekado_>bah, I’ve spent the whole day reading RStudio code
<rekado_>I’m trying to make it spawn *any* R version for any user, including Guix-installed Rs.
<rekado_>multiple R versions are only available in their Pro (for “proprietary”) spin-off, but I can’t imagine it to be too difficult to implement.
<rekado_>turns out that they do a weird hybrid thing: some parts talk to an inferior R process, others only talk to a custom C++ wrapper around libR
<rekado_>the former is easy to override but the latter is seriously limiting due to the need for matching ABI
<PurpleSym>rekado_: My condolences. RStudio’s code is a terrible mess…
<rekado_>I think that for what it does it seems terribly complicated
<rekado_>and the inconsistency irks me
<PurpleSym>Yeah, and as a result they apparently break things regularly.
<PurpleSym>How do you spawn RStudio? Do you run the full server with PAM and everything?
<rekado_>while I’m hacking on it I just do “./server/rserver --config-file conf/rserver-dev.conf --auth-none=1”
<rekado_>the next RStudio project will be to figure out how to make it authenticate users via the system’s PAM
<PurpleSym>I don’t really need that (we’re just using per-user --auth-none instances) but I’d pull it into guix-science if you figure out how, rekado_.