IRC channel logs


back to list of logs

<efraim>I was thinking an interm method of dealing with it would be to download the "release" unminimized version and minimize it ourselves for the packages which want it
<efraim>there are already some packages which bundle jquery or other JS libraries
<rekado_>efraim: this is what I wanted to do since day one, but I didn’t find anything that was buildable and could be considered source code.
<rekado_>if you have a URL for a concatinated unminified version I’d gladly take it.
*rekado_ prepares to push CRAN updates to the r-updates branch
<efraim>I'm on vacation ATM but I'll keep my eyes open
<zimoun>civodul, argh local build of julia leads to error test/testdefs.jl:21. Because libgit2. The next commit remove libgit2. is it a fix for that? Or unrelated?
<civodul>zimoun: is self-explanatory; does Julia need a custom libgit2?
<civodul>actually are you on master?
<zimoun>I have read the message. I am on core-updates-frozen. About libgit2, I do not know, the error «could not load library "/gnu/store/…-libgit2-1.3.0/lib/"» which is really weird.
<civodul>so it's a regression compared to libgit2 1.2?
<civodul>if needed, we can reintroduce 1.2 and have Julia use it
<zimoun>I do not know, I am checking. Berlin has the substitutes for julia@1.6.3 and libgit2@1.1 if I am correct. Pre-e0e2316420 update to 1.3.
<zimoun>another data point, Julia for this test: usr/share/julia/stdlib/v1.6/LinearAlgebra/test/matmul.jl:155
<civodul>zimoun: ah yes, i think that's what i saw in the test failure logs
<zimoun>civodul: from IRClog you have seen test/math.jl:296 :-)
<zimoun>however, libgit seems an issue. And deterministic with the few tests I did. It is so long…
<civodul>zimoun: maybe you can build with -K, interrupt the build, and from there run tests manually and selectively
<zimoun>civodul, about non-deterministic ones. Yeah. But Julia is not reproducible so I propose to first see if it passes on Berlin. For instance, it does on roptat machine except libgit2.
<zimoun>and it does sometimes on mine too.
<zimoun>Otherwise, yes we could remove the 2-3 non-deterministic tests.
<civodul>i'd do like apteryx suggested: report the non-deterministic tests upstream, and skip them in the meantime
*civodul goes afk for a bit