Terry Reedy said:
I believe coercion is universally applicable and should pass on every
machine. I am not sure about the other two.
Actually, it probably is not a problem. If you run the regression
tests with the -v option, I expect you will see that the differences
are that cc_r is using signed zeroes.
I think there was a bug report. It doesn't seem to be open (at least
I can't find it), so it was probably closed as a Wont Fix. Basically
I was too lazy to fix it when I had access to AIX and I no longer have
access.
test_mmap is likely a known problem:
http://python.org/sf/678250
that is probably just a bad test.
test_pty is notoriously bad:
http://python.org/sf/713169
probably also a test problem.
It would be nice to clean up the tests and have them pass, but we need
people to spend the effort. It would also be nice to verify the
results on a couple of machines. Unfortunately, there hasn't been
much support in the past.
Neal