run-tests: refactor worker globals into global WorkerContext value-object

[staabm]

slowly reducing the number of global vars in run-tests.php to make it easier to reason about it.

if you like this direction, I would send more small PRs like that with the goal to reduce global variables.

proposed as suggested in #16675 (comment)

[iluuu1994]

I don't object to this change, but it likely only makes sense as part of a larger change. In that case, I would suggest consulting the list on whether large refactorings are something we want to do for this file. The file is already big and convoluted. Making more changes can easily make it worse if we're not careful.

[iluuu1994]

Please use native property types.

[staabm]

thanks for taking a look.

I don't object to this change, but it likely only makes sense as part of a larger change.

I totally agree. I was doing this change as small as possible to not waste time and get the conversation started.
I hoped for multiple small PRs to de-global run-tests.php

In that case, I would suggest consulting the list on whether large refactorings are something we want to do for this file.

please take this as a constructive comment:

hopefully some day the php-project is able to get a more modern mean of communication.
I think php is loosing lots of contributors/contributions as soon as it requires people to connect with the list.

It would be way easier, when the php-project would use what the major part of the php community is using - whatever github.com provides.

I know this was discussed elsewhere and we don't need to argument about it here.

The file is already big and convoluted. Making more changes can easily make it worse if we're not careful.

Totally agree.

[iluuu1994]

It would be way easier, when the php-project would use what the major part of the php community is using - whatever github.com provides.

@Crell, @pronskiy and I were looking into this a while ago, but unfortunately momentum has died down. I think GitHub is a bad medium for such discussions. Discourse is not perfect, but significantly better, so it might be a good option. We had two primary ideas:

• Improving the RFC process itself, i.e. by making it less likely for much time to be invested in an idea that is doomed to fail.
• Switch to a better medium, one that reduces noise, improves notifications, and is overall more enjoyable to use.

We have, for now, focused on the former. I don't think there's anything stopping us from working on these things in parallel either, so if that's something you're interested in, please let me know.

Also, please don't let the ML stop you from helping improve PHP! It's really not that bad. Essentially, you can send out an initial e-mail and read the general perception, and then we can decide how to proceed.