fix missing and extra patch_minor in ua tests

[masklinn]

patch_minor was added to regexes and some test_ua entries in #322.

Neither spec nor reference implementation were ever updated for it, so many regexes were merged (?) with a capture for patch_minor but without correctly asserting it, and a pair of cases specify a patch_minor which is not captured:

• the facebook regex¹ only has 4 capturing groups
• same for the AWS regex²

Footnotes

1. https://github.com/ua-parser/uap-core/blob/959e106754828ae557b0dbcfaf8eeee938d3c824/regexes.yaml#L176 ↩

2. https://github.com/ua-parser/uap-core/blob/959e106754828ae557b0dbcfaf8eeee938d3c824/regexes.yaml#L155 ↩

[migueldemoura]

Should we assert the key is present on all tests? Or at least for those where it's not empty.

[masklinn]

Should we assert the key is present on all tests? Or at least for those where it's not empty.

I'm not sure I quite understand the question.

This PR should have filled in all the entries where the regex has a 5th capturing group (so patch_minor ends up non-empty) as they were failing the test suite I was working on (and so were the two extras for which the regex has no matching group).

So all the entries with an unspecified patch_minor are those for which it would be empty / null. Whether it's easier for those to be present or not you'd have to ask the various maintainers. For the Python implementation it shouldn't matter as it's using a set comparison for the UA cases, which handles a missing entry fine, and the Rust implementation I'm working on already handles it since that's what I used to find the problematic test cases. So personally I don't mind either way.

Also this PR should probably be linked to #562, I'd completely forgotten I'd reported that issue but hadn't had the time to work on it.