[bug#27271] [PATCH 0/4] Catch collisions at profile creation time

[Ricardo Wurmus <rekado@elephly.net>]

Hi Ludo,

L> These patches allow us to catch problematic collisions when computing
L> a profile derivation.  As we know, the profile builder often spits out
L> a number of warnings about collisions but that is not very useful because
L> users cannot distinguish the problematic cases from the harmless cases
L> (an example of a harmless case is when GDB and Binutils provide an
L> almost-identical .info file twice).

[…]

R> I’ll give this a try soon and report my experience with it.

I just tried it and I didn’t work the way I thought it would.

Here’s what I did:

  # install old numpy
  guix package -p /tmp/test -i /gnu/store/s02iw98l234ngkcnxqi7jz54vqqgx6hj-python2-numpy-1.10.4

  # install a package depending on a later version of numpy
  guix package -p /tmp/test -i bamm

It built bamm and then proceeded to build a profile, while spitting out
hundreds of lines about conflicts between python2-numpy-1.10.4 and
python2-numpy-1.12.0.

This is the profile’s manifest:

--8<---------------cut here---------------start------------->8---
(manifest
  (version 3)
  (packages
    (("bamm"
      "1.7.3"
      "out"
      "/gnu/store/lcb2s2x3s50gmf24asl2mvv34jhx8n1x-bamm-1.7.3"
      (propagated-inputs
        (("python2-numpy"
          "1.12.0"
          "out"
          "/gnu/store/pzf5yszv5dlzmk71w7srdi2qdqh2j40a-python2-numpy-1.12.0"
          (propagated-inputs ())
          (search-paths ()))))
      (search-paths ()))
     ("python2-numpy"
      "1.10.4"
      "out"
      "/gnu/store/s02iw98l234ngkcnxqi7jz54vqqgx6hj-python2-numpy-1.10.4"
      (propagated-inputs ())
      (search-paths ())))))
--8<---------------cut here---------------end--------------->8---

Did I do something wrong?

I also wonder if we should add a way to force Guix to build the profile
despite the detected conflict.

--
Ricardo

GPG: BCA6 89B6 3655 3801 C3C6  2150 197A 5888 235F ACAC
https://elephly.net