IRC channel logs

<zimoun>When building an old OpenBLAS, I hit the well-known error because unsupported hardware.

<zimoun>make: *** [Makefile.prebuild:66: getarch_2nd] Error 1

<zimoun>Makefile:135: *** OpenBLAS: Detecting CPU failed. Please set TARGET explicitly, e.g. make TARGET=your_cpu_target. Please read README for the detail.. Stop.

<zimoun>Would it possible to rebuild that derviation on Bordeaux or Berlin?

<zimoun>/gnu/store/1bl5swkx4f4s4z81jm2xyl23cmqc2sa8-openblas-0.3.5.drv

<civodul>hi zimoun!

<civodul>what commit is this?

<zimoun>civodul: A very old one. ;-) August 2019.

<zimoun>711cec9d15

<civodul>ACTION tries

<zimoun>civodul: thanks.

<civodul>seems like this is gonna take a while

<zimoun>Ah!

<zimoun>If it succeeds. ;-)

<zimoun>Because it could hit time-bombs on the road…

<zimoun>civodul: Well, I did some experiments about past revisions: Many Haskell, thus Pandoc, are not covered by either SWH or Disarchive. Idem for TeXLive. Therefore, building old packages often fails because documentation.

<civodul>uh, interesting

<civodul>TeX Live should be recoverable in general though, maybe with some manual tweaking?

<zimoun>For TeXLive, you need to convert the past fixed-output derivation to a modern one. Otherwise you have an ugly (expected) backtrace.

<zimoun>Using this modern fixed-output derivation, it fallbacks as expected. But it returns:

<zimoun>SWH: directory with nar-sha256 hash 6a0169f01a52bb6dd21061d9ce1fda20f4e5af8f4e3f4b5a7c18ba47057201aa not found

<civodul>that’s probably because those “combined svn checkout” nar-sha256 hashes were not computed retroactively

<civodul>we should ask the SWH folks

<civodul>in https://gitlab.softwareheritage.org/swh/infra/sysadm-environment/-/issues/5263

<civodul>ACTION has to go

<civodul>ttyl!

<zimoun>see you

<civodul>zimoun, if you read this: “guix time-machine --commit=711cec9d154786e995ba104bceb39a2564d5c72f -q -- build openblas --no-grafts” should now have substitutes on ci.guix

IRC channel logs

2024-06-10.log