IRC channel logs
2023-10-19.log
back to list of logs
<rekado>finished the build of jaxlib (the bazel thing actually just builds the wheel; I’m compiling the Python files in a separate package). <rekado>there’s a problem with protobuf, though, when importing jax in a Python session. <rekado>will follow nixpkgs’s lead and link jax with the static lib for protobuf <rekado>[libprotobuf ERROR google/protobuf/descriptor_database.cc:642] File already exists in database: xla/xla_data.proto <rekado>I may need to rebuild everything with the bundled protobuf <rekado>computing the tarball of all source code for jaxlib takes about 10mins <rekado>most of the time is spent running tar on the results <rekado>it’s compressing about 3GB of data <rekado>feels awfully slow and I wish we didn’t have to do this <rekado>because the next step is to unpack it all again (in a different derivation) <rekado>woo, jax can now be imported. What a mess. <alxsim>I'm an academic in population genetics (both simulation and bioinformatics space) and I'm exploring Guix's potential in this space. <alxsim>Any reason htslib and samtools are stuck at v1.14 from 2 years ago? Is it just due to a scarcity of contributors? <civodul>alxsim: hi! scarcity is a strong word (we’re talking about ~100 contributors per months), but presumably the bioinfo subset of users didn’t need the newer version so far <civodul>getting the new version is often quite easy: by running ‘guix build htslib --with-latest=htslib’ or, to update the package definition, by running ‘./pre-inst-env guix refresh -u htslib’ <alxsim>Thanks, just wondered if it was technical incompatibilities <alxsim>@civodul, thanks for the pointer, quite useful. I haven't had time to explore everyting in the manual yet. <rekado>alxsim: we’re at 1.16 for htslib <rekado>we also have packages for 1.14, 1.12, 1.10, 1.9, 1.3.1, and 1.2.1 <rekado>it’s a bit of an outlier in Guix, because normally we just keep one version. <rekado>I haven’t updated htslib in a while because there was no urgent need for it. <rekado>I usually keep all the R stuff up-to-date, but the bioinfo tools only when requests come in <alxsim>Ah thanks didn't see that, sounds good. Anyway, I'll take some time to learn a few things and then try contributing some packages (mainly python) in the pop gen field.