Speaking of Minitest, I think this is the culprit: https://github.com/minitest/minitest/commit/a2c6c18570f6f0a1bf6af70fe3b6d... IOW the many broken gems could temporary either: 1. set `MT_COMPAT` env variable 2. explicitly require "minitest/unit" Other issue is old Mocha: https://github.com/minitest/minitest/issues/960 Which I am trying to update, but the `MiniTest` constant pops-up on other places, e.g. in fog-libvirt: https://github.com/fog/fog-libvirt/blob/1c7cf79effde6f3432fcb1921554a8614... Vít

These are the related PRs: https://src.fedoraproject.org/rpms/rubygem-webrobots/pull-request/1 https://src.fedoraproject.org/rpms/rubygem-unf/pull-request/4 https://src.fedoraproject.org/rpms/rubygem-jekyll/pull-request/7 Vít Dne 05. 09. 23 v 12:14 Vít Ondruch napsal(a): > > Just FTR, the Minitest / Mocha issues has been mostly resolved by > now. I have left untouched a few leaf packages, which would be better > to be removed from Fedora. And a few others, which I believe will be > fixed by their packager, if they wish (but these kind of fall into > the first category anyway). > > I have also updated shoulda / shoulda-context and I'll take this > opportunity and try to remove all dependencies on rubygem-shoulda, > subsequently dropping it altogether (however leaving the > shoulda-context around). Let me explain. rubygem-shoulda is > metapackage and > > 1) As far as I remember, it is not always very well maintained > upstream and I think that upstream also suggested using the shoudla-* > packages instead. As an example, we have shoulda-matchers 5.x in > Fedora but the meta package still depends on version 4.x > > 2) They still don't support Rails 7 officially. > > 3) They have introduced new test dependency called `snowball`, which > is responsible for creating new Rails apps for their test suite. This > in theory is nice, but it introduces a lot of assumptions (e.g. gems > are installed from rubygems.org), which are hard to influence on the > consumer side. So the Shoulda test suite is essentially disabled now. > > 4) It seems that there is not much package using shoulda left around. > When I started this was the list: > > ~~~ > > rubygem-jekyll-0:4.3.2-2.fc39.src > rubygem-unf-0:0.1.4-22.fc39.src > rubygem-webrobots-0:0.1.2-16.fc39.src > > ~~~ > > > As it turns out, all these are fine with just shoulda-context. > > 5) We will also be able to drop shoulda-matchers later. > > 6) This will simplify dependency chain, improve package build speed a > bit. > > > > > Vít > > > Dne 02. 08. 23 v 12:53 Vít Ondruch napsal(a): >> >> >> Speaking of Minitest, I think this is the culprit: >> >> https://github.com/minitest/minitest/commit/a2c6c18570f6f0a1bf6af70fe3b6d... >> >> IOW the many broken gems could temporary either: >> >> 1. set `MT_COMPAT` env variable >> >> 2. explicitly require "minitest/unit" >> >> >> Other issue is old Mocha: >> >> https://github.com/minitest/minitest/issues/960 >> >> Which I am trying to update, but the `MiniTest` constant pops-up on >> other places, e.g. in fog-libvirt: >> >> https://github.com/fog/fog-libvirt/blob/1c7cf79effde6f3432fcb1921554a8614... >> >> >> >> Vít >>