Version 5.0.1 of mod_wsgi now out just in time for Python 3.13. Still need to finalize new release of wrapt code to ensure okay for next Python release.

But why is it this way?.

obj = C(). class M(type(obj)):. def __enter__(self):. return super().__enter__(). def __exit__(self, exc_type, exc_value, exc_tb):. return super().__exit__(exc_type, exc_value, exc_tb). obj.__class__ = M.

If monkey patching you can override “__class__” with a mocked type to get it to work.

In Python the context manager special methods have to be defined on the associated class type. Why can’t they be defined by attributes on the object instance?

Tests for wrapt under python 3.12-dev using GitHub actions (whatever actual version that is), still at least passing and it hasn’t broken anything that know of.

Worst case would need a major rewrite of mod_wsgi daemon mode, which although my actually improve performance for various reasons, not sure I have the stomach for it.

One problem have just heard about though is that mod_wsgi doesn’t compile against Apache httpd 2.5 trunk and from what have seen may not be simple fix.

Since looks like can no longer test with Python 2.7/3.5 under GitHub actions may finally be time to drop support for it in mod_wsgi and wrapt. 🤣.

Just wanting to make sure I understand what load timers may cause on REST API if used with quite short interval and you have a lot of resources being monitored.

One specific thing am concerned about is how can’t have multiple generators off same list from different threads at same time. Python will complain with an exception if this happens. So not sure to what degree need to rely on thread mutexes around index access.

Same deal with GitHub sponsors, how many would take the money there as well and feel they have no obligations in any way to users who are funding them.

With all these people grumbling about PyPi trying to impose requirements on them I really have to wonder how many would blindly accept Tidelift requirements and take the money anyway.

I personally could make a useful sum of money via Tidelift, but I don't feel comfortable agreeing to those requirements as my projects are very much less than part time best effort things at the moment and I wouldn't feel right about it.

Still reading stuff about PyPi 2FA discussion, but one thing that comes to my mind is Tidelift and the guarantees they expect of you to get money via them for working on your open source project.

@tidelift Oh weird. TideLift renders okay for me on Safari, but not on Chrome or Firefox.

As a result of some of the discussions around PyPi and 2FA, I wanted to go check something related to that on @tidelift which I thought was relevant to the discussion, but their site rendering is completely broken. Is TideLift still an active thing?.

When an open source maintainer is already not in the right frame of mind to handle a request from left field, this doesn't help. :-(.

Then what can happen is you start talking cross purposes with neither understanding the other party and what each wants to progress the dicussion and find the best outcome.

Please please please, if reporting issues against a project, don't just say "I need you to implement this solution". Without context of what the original problem is it is hard to understand the why part.