Fix tick.core/micros for negative durations (#208)

[yuhan0]

Closes #208.

(Side note: I had an unrelated test failure here

tick/test/tick/api_test.cljc

Lines 37 to 38 in bd2eefd

	(is (= (t/date "2018-01-11")
	(t/date (t/instant 1515691416624))))

which seems to be a faullty test? The t/date conversion depends on ambient clock / timezone, which in my case happened to return #time/date "2018-01-12")

[henryw374]

couple of failing tests there.

... and yes, that other test could have a t/with-clock around it to make it work

[yuhan0]

oops, I didn't think to run cljs tests locally, and attempting to run make test-cljs-shadow failed with

#error {
 :cause Exec failed, error: 2 (No such file or directory)
 :via
 [{:type java.io.IOException
   :message Cannot run program "/Applications/Google Chrome.app/Contents/MacOS/Google Chrome": Exec failed, error: 2 (No such file or directory)

(I use ungoogled-chromium 😬)

But in any case looking at the CI test failure it seems to be an underlying inconsistency between js-joda and java-time? The failing tests are

    (is (= -1 (t/millis  (t/of-nanos  -1999999))))
    (is (= -1 (t/millis  (t/of-micros -1001))))

where clj returns -1, cljs returns -2.

I believe that means the following also returns different results depending on platform? (I don't have a cljs dev setup at hand to test with)

(cljc.java-time.duration/to-millis (cljc.java-time.duration/of-nanos -1000001))

Would you rather I fix it by wrapping in a reader conditional #?(:clj -1 :cljs -2) or remove the block of tests entirely (punting semantic quirks to upstream libs)

[henryw374]

hmm. tricky. fix it by wrapping in a reader conditional #?(:clj -1 :cljs -2) I think.

please just add a comment that this is discrepancy from underlying lib

[yuhan0]

ok pushed along with the with-clock fix, let me know if you prefer the commits squashed

[henryw374]

looks good. thanks