Fix parsing of service provider file

[karelmaxa]

This PR fixes wrong parsing of the service provider configuration file in the OSGi environment.

The specification of the configuration file is the following (see ServiceLoader JavaDoc):

The provider-configuration file must be encoded in UTF-8. Space and tab characters surrounding each service provider's name, as well as blank lines, are ignored. The comment character is '#' (U+0023 NUMBER SIGN); on each line all characters following the first comment character are ignored

Check List:

• Unit tests: YES
• Documentation: NO

[filiphr]

Thanks @karelmaxa. Everything looks great, only one small change

[pavelhoral]

A bit off topic (explanation for the crashing tests):

The build crashes because of FlowableMyBatisRuntimeHints uses non-exported shaded (internal) Javassist class from mybatis dependency. Then the class FlowableAppRuntimeHints uses imports from spring-core dependency, which is not defined as OSGi at all. And there might be others (I did not dig deeper). I think that the current state of main branch is kind of totally broken from the OSGi point of view.

So this seems like it needs to be addressed at some architectural design level whether OSGi is a concern or not. I am just a "passer by", so I just wanted to share this knowledge without knowing what to do with this information.

[filiphr]

Thanks for the explanation @pavelhoral. Those hints were moved just today. We'll look into fixing those OSGi tests. Seems like you know your way around OSGi, do you perhaps know if there is a way to exclude some classes? e.g. those runtime hints are not needed in an OSGi world.

[pavelhoral]

I am just an OSGi user... but I can try to look into it later. Normally you would define the packages that should not be exported with an exclamation mark in front of it. But there is a conflict with the configuration in the root POM file that practically marks all module packages for export (so it can not be overridden).

[pavelhoral]

I have created dedicated issue to the wiring issue #3844. This PR has kind of nothing to do with it. Also apologies for my previous comment that contained a lot of nonsense (now decorated with "strikethrough").

[filiphr]

Thanks @pavelhoral for the OSGi fixes on main. I've re-run the tests here to see that everything is OK. Will merge after that. Thanks @karelmaxa

[karelmaxa]

I have rebased the branch with the main, so the OSGi test fix should now be included.

[filiphr]

Thanks @karelmaxa. The rebase wasn't entirely needed due to the way GitHub does PR checks. The tests are run by applying your changes on top of the target branch (main) and then running the tests.