> bba131e48852 causes crashes on Windows.
>
> The attached patch fixes the crash and makes test_os pass for me.
>
> However, using "PyErr_ExceptionMatches(PyExc_RuntimeError)" to check
> whether to try again using narrow strings is ugly.  Maybe
> utime_read_time_arguments() should be changed to have three possible
> return values.

I appreciate the feedback, and the patch.  And I agree--we should be able to find a better fix than that particular band-aid.  Can we hold off on checking in a patch for now?

TBH I don't understand why it should crash, and therefore how your patch helps.  Trying again using narrow strings should always work; indeed, the code did that before I touched it.  Can you describe how it crashes?

(p.s. Considering that I can't test on Windows myself, I'm pretty happy that the code works as well as it does!)