Petr Pisar

This is triggered by https://github.com/mojolicious/mojo/commit/52a237a813e778a998caa99e71dd8c8d5fd0cd2a.

The Mojolicious change adds use 5.016; that enables unicode_eval feature. The Mojolicious scope somehow leaks into Mango::BSON::_decode_value() that compiles a regular expression: ~~~ # Regex if ($type eq REGEX) {...

Actually it's simple. Mojo::Base that is used from both /lib/Mango/BSON.pm a t/base.t does use 5.016; in the top-level scope. Thus it's in the same scope as the test code. This...

A plain top-level use feature ':5.16'; is not enough. The explanation is a bit longer: ~~~ package Mojo::Base; sub import { require feature; feature->import(':5.16'); } 1; ~~~

There is a patch in fortl's fork in commit . It's mostly a white-space shuffle. Without it, it boils down to: ~~~~ --- a/lib/Mango/Auth/SCRAM.pm +++ b/lib/Mango/Auth/SCRAM.pm @@ -34,11 +34,9 @@...

Modules are always built against a specific Fedora release and they run-require the release. Hence it's not possible to inherit F35 modular builds for F36. Currently all non-Rawhide modular builds...

Any progress? Perl maintainers are going to add a new perl:5.34 stream into Fedora in a few weeks and then they will need to deliver multiple contexts for other perl-*...

Any progress?

I think your approach with not specifying a TLS method is better.

Thanks for the report. I will try to investigate what's wrong with `Base.download_packages()` method. However, for issues in RHEL software please contact Red Hat support . This repository on Github...