lib/timeutils: Fix problem with dates prior to Mar 1, 2000

[dhylands]

The easiest way to see this is to do:

>>> import time
>>> time.localtime(0)
(2000, 1, 1, 0, 0, 0, 5, 1)
>>> time.localtime(1)
(2000, 1, 2, 233, 197, 197, 6, 2)

With this patch, we instead get:

>>> import time
>>> time.localtime(1)
(2000, 1, 1, 0, 0, 1, 5, 1)

Doh - In C % is NOT a modulo operator, it's a remainder operator.

[coveralls]

Coverage remained the same at 93.81% when pulling feee431 on dhylands:fix-timeutils into 97ce883 on micropython:master.

[pfalcon]

Can there be a testcase perhaps? Also, as a dumb sanity check style question - you don't need to check days for <0, no?

[dhylands]

Sure - I'll fix the test code (in tests/pyb/modtime,py).

It currently passes the tests when seconds % 86400 == 0 (which it happens to in the test).

We should check the negative range (the test code currently tests only 2000..2034) which is the +ve range of 31 bit int. It can test the -ve range as well, and also test some offsets.

I think I'll just add some spot tests for the negative and positive side (the positive side exhaustively tests every day.

[pfalcon]

Yes, spot tests (within each distinct range/subset) is what we need, otherwise testsuite runs too long and people avoid running it ;-).

[dhylands]

I added some spot-tests. Along the way, I discovered a bug in mktime. I'll deal with that in a separate PR.

[coveralls]

Coverage increased (+0.0%) to 93.81% when pulling 14c28a3 on dhylands:fix-timeutils into 97ce883 on micropython:master.

[pfalcon]

Merged, thanks.