Fix send periodic

[Tbruno25]

Closes #1710

The proposed fix is to check whether duration is exceeded before the transmit of the next message.

[grant-allan-ctct]

I'm delighted to see that you've added a new test case. 🏆

[zariiii9003]

msg_due_time_ns and msg_index should only be incremented after the message was sent.

[hardbyte]

Thanks both @grant-allan-ctct and @Tbruno25 for the bug report, fix and test.

[zariiii9003]

Now the CI is failing 😅
@Tbruno25 Could you try to improve the robustness of test_send_periodic_duration?

[Tbruno25]

@zariiii9003 absolutely! Although it appears to only be failing on the very specific windows/pypy combinations. Would it be wrong if I simply skipped this test for this environment?

[grant-allan-ctct]

It's a interesting thing. I wonder what it's telling us.

Perhaps it has found a bug for which there is no other test coverage yet existing. For example, perhaps the message timestamp values are coming back as integers on this platform. Or perhaps one the calls to self.bus1.recv is choking on one of the arguments that the testcase is passing it.

Disclaimer: I just had to Google what pypy is, so please disregard if you think this comment could well be nonsense.

[zariiii9003]

@zariiii9003 absolutely! Although it appears to only be failing on the very specific windows/pypy combinations. Would it be wrong if I simply skipped this test for this environment?

Yes that's okay. I think we disable most timing sensitive tests in CI.