On 4/19/22 17:36, Zebediah Figura wrote:
On 4/19/22 08:56, Rémi Bernon wrote:
I now think that, though there's no guarantee the modules won't change, the transform / dmo classes and their tests should perhaps better live in the same modules where native has them. It is in theory possible that some applications load them and instantiate the classes directly, and I don't think it is likely that windows will remove these modules.
Maybe, although Windows has moved or deleted codecs in the past. (And I expect that if an application really does try to manually instantiate a class, it'll be easy to debug and may not even mean moving the entire implementation out of winegstreamer.)
Beyond that, though, why should the tests live in the same directory as the implementation? At best this helps make sure that tests are automatically run by the testbot, but as the person ensuring that all tests are manually run, I don't particularly see a need to change this (e.g. I'm still going to need to run tests manually anyway.) If we really care about this, we should come up with a solution that'll allow us to automatically run tests in other places where we have strong inter-module dependencies.
The main blocker for that is winegstreamer being a dll and the classes being registered in it. Perhaps it should be a static lib instead, or expose the unixlib interface as its PE entry points.
I don't think we can make winegstreamer a static library, but moving one or more frontends out of it is possible. It doesn't seem necessary yet, though.
I don't know, I'm trying to understand why these patches often stall and to figure a way to make it go smoother.
I'm not pretending the tests are all good and should be accepted right away but I hope they aren't completely nonsensical either. I still have plenty of tests to add, and I'm happy to maintain them if it's a burden for the modules maintainers.