IRC channel logs
2024-03-14.log
back to list of logs
<rekado>I don't understand the question. The bazel stuff is fixed output. <rekado>I don't see how the tensorflow package would be any different from other packages in that regard. <rekado>civodul: the printed report looks nice! <rekado>so... what's the effective page size? Is it slightly smaller than A5 due to cropping for the front page illustration? <rekado>ACTION just fixed python-tensorflow on guix-science <rekado>weird: python-tensorstore fails with a hash mismatch. Not good. <civodul>nice thing is i didn’t give folks any instructions and they just did the right thing <jonsger>civodul: can the CVE fix in guix-daemon (ff1251de0bc327ec478fc66a562430fbf35aef42) directly be backported to 1.4.0 or are there additionally patches required? <civodul>jonsger: the two commits can be applied as-is to 1.4.0 i believe <civodul>i think vagrantc already did that for Debian <rekado>the bazel-generated external/pypa_numpy/BUILD.bazel in the tensorstore sources contains references to /gnu/store/...-python-numpy-.../ <rekado>it wants to copy the numpy headers, so it generates a list of them all and copies them. <rekado>so the bazel output here depends on python-numpy, which makes this a non-fixed output derivation. <rekado>I can avoid the store locations, but ... the question is if I really should <rekado>it's copying a handful of numpy headers, and I'll just assume that they don't change over time <rekado>the *source* code bundle really should not include numpy headers. <civodul>rekado: going back to Tensorflow, ayatsfer mentioned that without ‘--no-grafts’, they’d get a different derivationm than that available on guix.bordeaux, and no substitutes <rekado>I don't know why that is, though. <rekado>no headers are actually included; only their absolute file names were embedded