``test_ioctl`` fails on macOS

[Eclips4]

Bug report

Bug description:

./python.exe -m test -v test_ioctl
== CPython 3.14.0a0 (heads/main:a5fef800d3, May 30 2024, 09:48:26) [Clang 15.0.0 (clang-1500.3.9.4)]
== macOS-14.5-arm64-arm-64bit-Mach-O little-endian
== Python build: debug
== cwd: /Users/admin/Projects/cpython/build/test_python_worker_4187æ
== CPU count: 8
== encodings: locale=UTF-8 FS=utf-8
== resources: all test resources are disabled, use -u option to unskip tests

Using random seed: 2309191984
0:00:00 load avg: 63.19 Run 1 test sequentially
0:00:00 load avg: 63.19 [1/1] test_ioctl
test_ioctl (test.test_ioctl.IoctlTests.test_ioctl) ... ok
test_ioctl_mutate (test.test_ioctl.IoctlTests.test_ioctl_mutate) ... ok
test_ioctl_mutate_1024 (test.test_ioctl.IoctlTests.test_ioctl_mutate_1024) ... ok
test_ioctl_mutate_2048 (test.test_ioctl.IoctlTests.test_ioctl_mutate_2048) ... ok
test_ioctl_signed_unsigned_code_param (test.test_ioctl.IoctlTests.test_ioctl_signed_unsigned_code_param) ... ERROR

======================================================================
ERROR: test_ioctl_signed_unsigned_code_param (test.test_ioctl.IoctlTests.test_ioctl_signed_unsigned_code_param)
----------------------------------------------------------------------
Traceback (most recent call last):
 File "/Users/admin/Projects/cpython/Lib/test/test_ioctl.py", line 85, in test_ioctl_signed_unsigned_code_param
 new_winsz = fcntl.ioctl(mfd, set_winsz_opcode_maybe_neg, our_winsz)
OSError: [Errno 25] Inappropriate ioctl for device

----------------------------------------------------------------------
Ran 5 tests in 0.003s

FAILED (errors=1)
test test_ioctl failed
test_ioctl failed (1 error)

== Tests result: FAILURE ==

1 test failed:
 test_ioctl

Total duration: 96 ms
Total tests: run=5
Total test files: run=1/1 failed=1
Result: FAILURE

CPython versions tested on:

CPython main branch

Operating systems tested on:

macOS

Linked PRs

• gh-119770: Remove buggy ioctl test case #119818
• gh-119770: Make termios ioctl() constants positive #119840
• [3.13] gh-119770: Make termios ioctl() constants positive (GH-119840) #119862
• [3.12] gh-119770: Make termios ioctl() constants positive (GH-119840) #119863

[Eclips4]

Bisected to 92fab33
cc @vstinner

[sobolevn]

Yes, the same happens for me on m2

[Eclips4]

Unfortunately, this commit breaks backward compatibility. Previously, it was allowed to pass negative values for the request parameter and then they were converted to positive values. However, now, due to a change in the type from unsigned int to unsigned long, this is not working.

[vstinner]

This test is strange. Why passing negative request if its type is unsigned long? The value should be positive.

test_ioctl was always skipped until recently. I fixed regrtest to test it: #119275

[Eclips4]

This test is strange. Why passing negative request if its type is unsigned long? The value should be positive.

test_ioctl was always skipped until recently. I fixed regrtest to test it: #119275

Yes, I agree with you that this is strange. Our documentation does not mention that the ioctl request parameter can be negative. I think this test was broken by design.

[Eclips4]

This line from the test_ioctl_signed_unsigned_code_param is confused me:

if termios.TIOCSWINSZ < 0:

Is there any chance that TIOCSWINSZ can be negative?

[vstinner]

Is there any chance that TIOCSWINSZ can be negative?

darwin-xnu uses uint32_t with the MSB bit set:

#define IOC_IN (__uint32_t)0x80000000
#define _IOC(inout, group, num, len) \
	(inout | ((len & IOCPARM_MASK) << 16) | ((group) << 8) | (num))
#define _IOW(g, n, t) _IOC(IOC_IN,	(g), (n), sizeof(t))
#define TIOCSWINSZ _IOW('t', 103, struct winsize) /* set window size */

Source:

• https://github.com/apple/darwin-xnu/blob/2ff845c2e033bd0ff64b5b6aa6063a1f8f65aa32/bsd/sys/ioccom.h#L85-L95
• https://github.com/apple/darwin-xnu/blob/2ff845c2e033bd0ff64b5b6aa6063a1f8f65aa32/bsd/sys/ttycom.h#L154

Python termios module stores the constant as long in termios_constants array, and then create an object using PyLong_FromLong()`.

The value may be negative or not depending on how C treats the integers. I would prefer to have a more reliable value.