Pekka Klärck

XML processing is already now performance optimized. There may be something special in your case, but that's something only you can investigate. This is what you can do: - Use...

if you run out of memory, the machine starts swapping which is slow.

Sounds good to me. There's actually a comment in the code about adding a separate `TaskSection` and that ought to be pretty easy. For backwards compatibility it probably should extend...

Each test gets a new instance, but the problem is that variables in arguments are resolved only when the library is imported. When instances are created later, changes to variables...

I was able to get the example working as expected, but several acceptance tests failed after that. I'm sure this could be fixed properly, but that isn't important enough to...

We are just discussing about RF 7 scope on the #roadmap channel on our Slack. So far nobody has mentioned this issue, but if others agree this should be fixed...

We refactored code related to test libraries heavily as part of RF 7.0 but this behavior wasn't changed. I thought about that for some time, but it would have required...

Sounds reasonable. This would require inspecting all module attributes to check are they classes that have the special `ROBOT_AUTO_KEYWORDS` attribute set. It wouldn't be hard and I believe it shouldn't...

Some further comments about design: - It should be an error if there are multiple classes with `ROBOT_AUTO_KEYWORS`. - We shouldn't search for library classes if the module itself has...

I'm not sure about all PythonLibCore usages, but in general the `@library` decorator should only be used with the main library and even with that it's optional. Do you @Snooz82...