test_worker_timeout fails on s390x

[mcepl]

When running the test suite while packaging for openSUSE, I get this test failing:

[  155s] =================================== FAILURES ===================================
[  155s] ______________________ test_worker_timeout[run_in_thread] ______________________
[  155s] 
[  155s] kernel = <curio.kernel.Kernel object at 0x3ff8675bd10>
[  155s] runner = <function run_in_thread at 0x3ff84bcd1c0>
[  155s] 
[  155s]     @pytest.mark.parametrize('runner', [run_in_thread, run_in_process])
[  155s]     def test_worker_timeout(kernel, runner):
[  155s]         results = []
[  155s]     
[  155s]         async def spin(n):
[  155s]             while n > 0:
[  155s]                 results.append(n)
[  155s]                 await sleep(0.1)
[  155s]                 n -= 1
[  155s]     
[  155s]         async def blocking(n):
[  155s]             try:
[  155s]                 result = await timeout_after(0.55, runner(time.sleep, n))
[  155s]             except TaskTimeout:
[  155s]                 results.append('cancel')
[  155s]     
[  155s]         async def main():
[  155s]             async with TaskGroup() as g:
[  155s]                 await g.spawn(spin, 10)
[  155s]                 await g.spawn(blocking, 5)
[  155s]     
[  155s]         kernel.run(main())
[  155s]     
[  155s] >       assert results == [
[  155s]             10, 9, 8, 7, 6, 5, 'cancel', 4, 3, 2, 1
[  155s]         ]
[  155s] E       AssertionError: assert [10, 9, 8, 7, 6, 'cancel', 5, 4, 3, 2, 1] == [10, 9, 8, 7, 6, 5, 'cancel', 4, 3, 2, 1]
[  155s] E         At index 5 diff: 'cancel' != 5
[  155s] E         Full diff:
[  155s] E         - [10, 9, 8, 7, 6, 5, 'cancel', 4, 3, 2, 1]
[  155s] E         ?                 ---
[  155s] E         + [10, 9, 8, 7, 6, 'cancel', 5, 4, 3, 2, 1]
[  155s] E         ?                            +++
[  155s] 
[  155s] tests/test_workers.py:143: AssertionError

I thought that curio is pure Python, so big endian/little endian issues (which is 90% of all such s390x-only bugs) should not affect it?

Complete build log with all packages used and steps taken to reproduce.

[dabeaz]

No idea. Curio is pure Python although there are a few minor differences to account for the Windows/Unix split.

[mcepl]

On Fri Nov 17, 2023 at 12:52 PM CET, David Beazley wrote: No idea. Curio is pure Python although there are a few minor differences to account for the Windows/Unix split.

I saw it is around the code deals with some bytestrings … is there some in fact binary material which can be influenced by endinnaness? Best, Matěj

…

-- http://matej.ceplovi.cz/blog/, @***@***.*** GPG Finger: 3C76 A027 CA45 AD70 98B5 BC1D 7920 5802 880B C9D8 Though boys throw stones at frogs in sport, the frogs do not die in sport but in earnest. -- Bion of Borysthenes as quoted by Plutarch