00:50 <lopex> headius[m]: maybe there's a bug in our case folding or some tables changed in 2.7

00:50 <lopex> which test is that ?

00:51 <lopex> headius[m]: we're not using jdk for case mapping btw

00:55 <headius[m]> I did not try using jcoding case logic

00:55 <headius[m]> the test is this:

00:55 <headius[m]> and test_nonascii_constant is the same

02:23 <headius[m]> pecking away at MRI failures now

02:48 <travis-ci> jruby/jruby (ruby-2.6:0e13efb by Charles Oliver Nutter): The build is still failing. https://travis-ci.org/jruby/jruby/builds/668249296 [189 min 59 sec]

02:50 <headius[m]> @kare

02:50 <headius[m]> kares: I think you added this test... only thing failing in JRuby suites on ruby-2.6 branch:

02:50 <headius[m]> pkg count is 6... seems like the extra ones are "Java", "JavaIo", and "JavaLangReflect"

02:51 <headius[m]> I found a reference to java.lang.reflect in our ruby kernel but it's been there for a while: https://github.com/jruby/jruby/blob/b104023283c65568ae4bc7c95815e1b7c249eb30/core/src/main/ruby/jruby/java/java_ext/java.lang.rb#L9

02:51 <headius[m]> and there's lots of java::lang access in that same file

02:52 <headius[m]> so I'm not sure which of these packages shouldn't be showing up... perhaps you can take a look?

03:49 <travis-ci> jruby/jruby (ruby-2.6:80db909 by Charles Oliver Nutter): The build is still failing. https://travis-ci.org/jruby/jruby/builds/668257781 [186 min 44 sec]

05:29 <travis-ci> jruby/jruby (ruby-2.6:1a199c5 by Charles Oliver Nutter): The build is still failing. https://travis-ci.org/jruby/jruby/builds/668273890 [188 min 34 sec]

05:44 <kares[m]> headius: interesting, will take a look what changed in terms of JI loading

06:33 <headius[m]> jeremyevans: sequel seems to be testing some Time behaviors that aren't in specs or MRI test suite: https://travis-ci.org/github/jruby/jruby/jobs/668283511

06:34 <headius[m]> I mention that as a good thing, but if I can figure out what these cases are I'll try to add specs

06:34 <headius[m]> kares: thanks, I'm trying to wrap up 2.6 stuff and it's so close

06:49 <travis-ci> jruby/jruby (ruby-2.6:b501008 by Charles Oliver Nutter): The build has errored. https://travis-ci.org/jruby/jruby/builds/668283501 [202 min 11 sec]

11:38 <kares[m]> think the new packages being defined on every boot are due the new .rb parts of Range

11:41 <kares[m]> loading `java.lang.Double` into JI triggers `java.io` package to be initialized ...

11:44 <kares[m]> oh right, `java.lang.Number` implements `java.io.Serializable` that's why

11:46 <kares[m]> and since there's `Double::Type` the `java.lang.Class` gets into JI causing `java.lang.reflect` a few classes from `java.lang.reflect` to get into JI

11:46 <kares[m]> * and since there's `Double::Type` the `java.lang.Class` gets into JI causing `java.lang.reflect` a few classes from `java.lang.reflect` to init

11:47 <kares[m]> guess its the usual dilema - load less code at startup or now, a trivial fix would be to not `java_import` those in range.rb

11:48 <headius[m]> Ahh I see

11:48 <kares[m]> * guess its the usual dilema - load less code at startup or not, a trivial fix would be to not `java_import` those in range.rb

11:52 <headius[m]> There's just a couple uses...could be bound as JRuby::Util methods

11:52 <headius[m]> Or we just port this into Java

11:53 <headius[m]> The backports library uses pack and unpack to get the float bits

11:53 <headius[m]> Slow but works I suppose

12:28 <kares[m]> or we could have a java_import 'java.lang.Double', load: false

12:29 <kares[m]> probably not worth the hustle ... seems it would complicate things a bit

12:30 <kares[m]> either way (also just doing `java.lang.Double.doubleToLongBits`) is fine by me ...

12:31 <kares[m]> `Math` is not needed since we could just `double.abs`

12:57 <headius[m]> I tried using Float#abs and it didn't work the same

12:58 <headius[m]> I think it had different behavior around Infinity

13:09 <headius[m]> kares: I'm just going to add those as utility methods on BSearch that are implemented in Java

13:09 <headius[m]> no imports or JI needed at all then

13:16 <kares[m]> binary search on an infinite range? did not realize its supported 😉

13:18 <headius[m]> you had the same reaction I did

13:20 <headius[m]> https://gist.github.com/headius/17ddfb54a61452870f3024b16f34edaa

13:20 <headius[m]> there's the basics... just cleaning up now

13:22 <headius[m]> I'll just go ahead with this and revert my test change... any further improvements are welcome

13:26 <headius[m]> kares: thanks, seems like that was it

14:52 <jeremyevans> headius[m]: I think all of those cases were existing tests that originally used DateTime, and after ruby 2.6 was release I added support for Time using the new timezone API in 2.6, then figured out they could actually be supported without the timezone API

14:52 <jeremyevans> headius[m]: Anyway, it's good that Sequel is helping you find issues prior to release :)

15:00 <headius[m]> yeah I'll have to dig into these and see what we're not doing right

15:00 <headius[m]> all the new Time tests and specs are passing so it's edges that didn't get tests

15:01 <headius[m]> I did tag a couple things I thought were obscure or minor differences but I'll revisit those

15:57 <travis-ci> jruby/jruby (ruby-2.6:329b45c by Charles Oliver Nutter): The build is still failing. https://travis-ci.org/jruby/jruby/builds/668362515 [185 min 6 sec]

23:39 <headius[m]> hello again

23:44 <headius[m]> excellent, sequel timezone specs run without db