test_clock_settime of test_time fails on macOS Tahoe developer preview 1 causing --enable-optimizations build to fail

[smasyutin]

Bug report

Bug description:

I'm giving macOS Tahoe a try and when installing Python 3.13 with macPorts or building from source it fails with

0:00:29 load avg: 13.06 [40/43] test_tabnanny
0:00:29 load avg: 13.06 [40/43] test_tabnanny passed
0:00:29 load avg: 13.06 [41/43] test_time
test test_time failed
0:00:32 load avg: 12.02 [41/43] test_time failed (1 error)

CPython versions tested on:

3.13

Operating systems tested on:

macOS

[StanFromIreland]

Can you run the test in verbose -v and post the output here.

This may simply be because of you timezone.

[markmentovai]

macOS 26.0db1 25A5279m has a broken version of clock_settime. This is filed with Apple as FB17944386 (OpenRadar copy).

When the user has no permission to set the clock, clock_settime is documented (by spec as well as Apple’s own documentation) to fail with EPERM. In Apple’s new OS beta, it’s incorrectly failing with ENOENT instead. This is a problem for Python’s test_time test_clock_settime, which either expects clock_settime to succeed, or to fail with PermissionError (the Python error mapped from EPERM).

As this is a bug in an OS beta, I don’t believe any action is necessary, other than to wait for Apple to fix it.

As a local fix, I’m using this:

diff --git Modules/timemodule.c Modules/timemodule.c
index b6d8bf3..9af859b 100644
--- Modules/timemodule.c
+++ Modules/timemodule.c
@@ -286,6 +286,8 @@ time_clock_settime(PyObject *self, PyObject *args)
 
     ret = clock_settime((clockid_t)clk_id, &tp);
     if (ret != 0) {
+        if (errno == ENOENT)
+            errno = EPERM;
         PyErr_SetFromErrno(PyExc_OSError);
         return NULL;
     }

I don’t believe that it’s appropriate to work around this in Python at this point. On the off chance that Apple doesn’t fix the bug by the time they release macOS 26, it would be appropriate to cook up something better than this rough patch.

[zware]

Thanks for the info, @markmentovai! I fully agree that this is a macOS bug and not something we should try to fix. Chances are that if we do need to make a change, it will be to accept the new error in the test.

If it's still an issue when macOS 26 is released, please ping this issue and we'll reopen it.

[smasyutin]

Thanks folks.
I submitted the same via Feedback Assistant, as failing Python 3.13 build actually blocks me from installation of any app that depends on Python 3.13 via MacPorts.

[ned-deily]

FTR, the build is failing because of the noted failure of test_time during the PGO profile execution. Until fixed by Apple, you can avoid the build failure by either patching test_time as suggested above or by avoiding using the configure option --enable-optimizations which runs test_time.

[markmentovai]

This has been fixed in macOS 26.0db2 25A5295e (2025-06-23).