Test against 3.10 beta #1921
Test against 3.10 beta #1921
Conversation
This is a requirement to achieve Python 3.10 compatibility.
Codecov Report
@@ Coverage Diff @@
## master #1921 +/- ##
=======================================
Coverage 99.56% 99.56%
=======================================
Files 114 114
Lines 14618 14634 +16
Branches 1117 1118 +1
=======================================
+ Hits 14554 14570 +16
Misses 43 43
Partials 21 21
|
In CPython 3.10.0a5 TracebackException.__init__() started to care about the difference between _seen=set() and _seen=None. As such we need to avoid turning None into set() prior to passing it to the real .__init__(). python/cpython@6dfd173
|
I'm not really sure why |
| threading.excepthook, prev = threading.__excepthook__, threading.excepthook | ||
| else: | ||
| threading.excepthook, prev = _noop, threading.excepthook | ||
|
|
There was a problem hiding this comment.
This is mysterious :-). What's going on here?
There was a problem hiding this comment.
I think this was for disabling the thread hooks that pytest started using in 6.2 or so to catch and complain about uncaught exceptions in threads. In certain cases, IIRC, we literally have explanations in the code, or even the error messages, saying that yes, these will fall out un-handled in the thread or whatever but that that's the way it is and we know. But @graingert did this bit. The added bit here is to use the 'original' excepthook instead of a noop when it is available (in 3.10). This additional feature in 3.10 let's this code be more consistent with the restore_unraisablehook() above.
Hopefully that's in the direction you were wondering about...
|
@agronholm, yeah, coding |
Draft for:
compact=keyword toTracebackException.__init__()in Python <3.10. Is this acceptable?python-version:insanity in the GHA workflow, just in case it doesn't get rejected...typed-astnot building on 3.10-devCheck out the 3.10-dev compilation failuremasterbranch but likely don't want to leave that laying around post-merge.Alljob:])