Subject: | Hash-Randomization causes 10_analyse.t to fail |
As I reported here https://rt.cpan.org/Ticket/Display.html?id=82938 on Test-CPAN-Meta-YAML , the order of ->error() is not consistent between runs, and MCA 's tests ( 10_analyse.t ) inherently rely on one specific ordering.
I have extracted the broken parts of 10_analyse.t into a seperate test case that spawns a child perl instance with predetermined seed values to replicate the failure consistently: https://github.com/kentfredric/Module-CPANTS-Analyse/commit/84e402baa24fffa287ac5c559febf7e38fd5b5d8
And to remedy this issue, either a) Test::CPAN::Meta::YAML needs to change their code to be predictable , or , b) MCA's tests should sort the values returned from ->error() to enforce some testable consistency.