08:33 <kares[m]> <headius[m] "kares: this seems to have broken"> interesting, another 9.2 regression - should be able to fix that one it's just that I am too busy to get anything JRuby done atm

08:34 <kares[m]> btw. Windows CI 🟢 with test:jruby and -Ptest https://github.com/jruby/jruby/pull/6340

14:21 <kares[m]> hmm, Java 15 shipped Hidden Classes ... another interesting perk for JRuby

15:22 <headius[m]> kares: nice about Windows... I need to get my spec fixes landed too

15:23 <headius[m]> I have been hearing about hidden classes, definitely a lot of things in Java 9-15 we should start looking into

15:54 <headius[m]> c941c41e54f988ce1ab6b023e910c02ab6583417

15:54 <headius[m]> oops

15:54 <headius[m]> https://github.com/jruby/jruby/issues/6379#issuecomment-694328219

15:55 <headius[m]> I figured out when we fixed the teardown issue, and it might be a safe fix to backport for 9.2.14... enebo kares maybe you can have a look and think about impact

16:52 <enebo[m]> headius: ok

16:52 <enebo[m]> kares: nice on windows ci

18:20 <kares[m]> +1 from me on the exit! fix - seems like a proper bug resolution

20:53 <headius[m]> kares: ok

21:26 <headius[m]> enebo: apparently there's now a large number of specs for win32ole

21:26 <headius[m]> I don't remember how to set it up... is it just in the dist or a gem?

21:26 <enebo[m]> It is its own project

21:27 <headius[m]> I will probably ultimately exclude these specs from e.g. CI but I wanted to give it a shot

21:28 <headius[m]> ah we do publish a gem

21:28 <enebo[m]> https://github.com/enebo/jruby-win32ole and https://github.com/enebo/racob

21:29 <enebo[m]> If you build from source then both may need considerable updates depending on how you try and compile

21:29 <headius[m]> hmm pretty much all fails with the same error

21:29 <headius[m]> with the latest released gem

21:29 <enebo[m]> The original difficulty was that win32ole headers were not provided for free on windows and you needed a particular visual studio

21:29 <enebo[m]> I believe that has changed in recent years though

21:30 <headius[m]> I wasn't planning to hack on it today but it might be an interesting weekend project to update it and try specs

21:30 <enebo[m]> well that is pretty weird

21:30 <headius[m]> yeah getting something nil when it should not in registry access

21:30 <enebo[m]> I would guess win32ole is not broken due to anything with win32ole

21:30 <headius[m]> doesn't look like it from that error, no

21:30 <enebo[m]> hehe err registry

21:31 <enebo[m]> win32/registry most likely is broken for some other reason

21:31 <enebo[m]> like fiddle or dl or something?

21:31 <enebo[m]> but without being able to call registry it probably prevents doing actual win32ole stuff

21:32 <headius[m]> yeah that would seem to be a prerequisite

21:32 <enebo[m]> OTOH people are reporting problems after they have an OLE thing in hand

21:32 <enebo[m]> So I don't know

21:32 <headius[m]> yeah it needs some updating for sure, I just don't know what and that's not on my list today

21:33 <enebo[m]> Perhaps some things do not look up clsid or stuff like that via registry if you provide proper value

21:33 <headius[m]> yeah pretty tough though if registry access doesn't work, there's a lot of stuff that needs that access

21:33 <enebo[m]> I know a bunch of people provide those uuids directly in their programmers to prevent lokoing them up

21:33 <enebo[m]> yeah I would say that is a significant issue for sure

21:34 <enebo[m]> and possibly bigger than individual win32ole issues since you use registry for a lot more than that on windows

21:35 <headius[m]> the good news is that fast specs run to completion now

21:35 <headius[m]> there's maybe 50-60 failures on Windows that I will tag for now

21:35 <headius[m]> unsurprisingly a bunch of them involve subprocesses

21:38 <headius[m]> I'm having some trouble getting rake + ant to work, maybe kares has solved that on his branch

21:38 <headius[m]> we can probably merge that PR in now since it's green

21:40 <headius[m]> hmm I should probably wait on these tags, I think some of kares fixes may improve this

21:40 <headius[m]> kares: I think you should go ahead and merge and I'll sync up with your changes before proceeding to tag off remaining spec failures

21:40 <headius[m]> they'll be tagged with "windows" so we'll know what we need to look at if we run a tag report

21:44 <headius[m]> I have merged my branch with master and pushed

21:48 <headius[m]> fast specs are successfully executing about 98.8% of the expectations on Windows that they do on Linux...not too shabby