fix: ensure proper order of signal emission

[Czaki]

I have noticed that If some callback triggers emission of an event during an existing loop, the events may come in incorrect order, leading to a corrupted state.

I think that it is a bug, but I'm not sure if my approach to solve is correct.

[codspeed-hq]

CodSpeed Performance Report

Merging #281 will degrade performances by 10.47%

_{Comparing Czaki:proper_signal_order (eccef1a) with main (4aa62b0)}

Summary

❌ 1 (👁 1) regressions
✅ 65 untouched benchmarks

Benchmarks breakdown

	Benchmark	main	Czaki:proper_signal_order	Change
👁	test_emit_time[setattr-50]	200.7 µs	224.1 µs	-10.47%

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (4aa62b0) to head (eccef1a).

Additional details and impacted files

@@            Coverage Diff            @@
##              main      #281   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           22        22           
  Lines         1880      1883    +3     
=========================================
+ Hits          1880      1883    +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[tlambert03]

minor modification suggested in Czaki#1
it's not perfect (see note about recursion error), but curious if you have any thoughts on that approach vs this current one

[tlambert03]

please merge main here and make sure all the tests we've discussed are added, then ping for review

[Czaki]

@tlambert03 compiled test are ending with Segfault. Did you meet something like this before:

  File "/Users/runner/work/psygnal/psygnal/tests/test_psygnal.py", line 990 in test_recursion_error

[tlambert03]

no, I haven't seen that before. I can reproduce locally as well though (was just trying as you wrote), but not on main. (btw, to test locally, you can run make build, and make clean to undo ...)

[Czaki]

Ok. The problem is that hatch/mypyc performs compilation in tmp directory (unlike cython that places c/cpp files in the structure of packages. So the debugger cannot find the source files.

I will try to dig, but maybe you know how to change this behavior or disable cleaning tmp

#7  0x00007ffff508f242 in CPyDef__signal___SignalInstance____run_emit_loop ()
    at /tmp/tmpcocqfl4i/build/__native_ac51d50a4f4b6d748b8c.c:37311
37311   /tmp/tmpcocqfl4i/build/__native_ac51d50a4f4b6d748b8c.c: No such file or directory

[tlambert03]

good sleuthing. i also don't know, but happy to look into it if you can't find it easily

[Czaki]

Im going to Warsaw Python meeting so will not be active today. Will sit to this at evening or toomorow

[Czaki]

This simulated stack is 2-3x times bigger than the original one, but maybe it not a problem.

[tlambert03]

Works for me! Thanks.

I'm not worried about the very small benchmark decreases on emit to setattr and setitem for large numbers of callbacks.

Thanks for working on this