If this is in a test doesn't it mean that it only affects people doing source installs? Binary installs will work fine. For source installs is the problem that if your user name happens to have "fr" in it the test fails? (most other workarounds would involve telling people to use a different build directory).

For my final comment I was trying to say that it's easier to tell people to use a different build directory than it is to tell them they have to change their user name.

+1 I'm all for fixing known bugs. It potentially saves a lot of time and frustration.

Would it make sense to include the other patch that people talked about including if we are going to go to the effort of a new release? I'm not entirely convinced that this edge case (supporting source builds for people who can't avoid having "fr" in their path) is enough to justify the work.

Could someone please explain why this was withdrawn?

The DM-CCB has discussed the progress on this RFC in the past weeks and took the decision to withdraw it because the cost of implementation was not worth it for the changes to be included.

The main reason is that, since the release 19.0.0 has been done, a breaking change in the build system has been introduced (conda compilers and moving 3rd party libraries into conda-forge). This change makes it very difficult to use the release tooling to issue a patch for releases prior to the change.

This is an exceptional situation that we should be experiencing very seldom, but we cannot exclude a priori. In principle, we should be able to do patch releases on the next major 20.0.0 release and future ones, unless another breaking change in the build infrastructure and tooling is implemented.

We are working to refactor the release tooling so that future changes will not cause this problem.