_whitelogger has joined #jruby
ur5us has quit [Ping timeout: 260 seconds]
ur5us has joined #jruby
_whitelogger has joined #jruby
_whitelogger has joined #jruby
ur5us has quit [Ping timeout: 240 seconds]
_whitelogger has joined #jruby
ur5us has joined #jruby
sagax has joined #jruby
nirvdrum has joined #jruby
ur5us has quit [Ping timeout: 260 seconds]
rdubya[m] has joined #jruby
<kares[m]> hey, so I was looking into this sporadic auto-loading related failure and was wondering whether to patch RGs as a work-around for 9.2
<kares[m]> might do some more digging - was a bit hard to reproduce and the reproducer is still not great - if time permits tomorrow :
<headius[m]> kares: it might be a decent workaround but of course would get wiped out immediately when someone upgrades
<headius[m]> unless we patch it in defaults/jruby.rb or something
subbu is now known as subbu|lunch
subbu|lunch is now known as subbu
NightMonkey has quit [Ping timeout: 260 seconds]
NightMonkey has joined #jruby
<kares[m]> yeah, true
<headius[m]> I want to get that load service PR finished but there's so much of everything to do
ur5us has joined #jruby
<kares[m]> yy - np. I'll see if there's anything small we could do here ...
<headius[m]> I think this week I'm going to get the major test suites passing on Java 11 so we can bless that for 9.3
<headius[m]> it's nearly there so I don't expect to see a lot of issues
<headius[m]> FWIW if anyone's interested, I'm going to start moving jobs over to 11 that are already passing. Support for 11 will largely be a subset of support for 8, so if 11 passes I doubt we'll ever see issues with 8
<headius[m]> of course we'll keep a few key suites running on 8 forever, probably (MRI tests, rubyspecs)
<headius[m]> that disappearing bin/rake is a pain in the ass
<headius[m]> I'm going to see if I can fix the default gem install to be smarter about reinstalls... right now if there's a specification file, it won't reinstall, which means it won't install the rake binscript on master when switching from 9.2
<headius[m]> kares: we need to talk about jossl too... I can deal with access warnings in JRuby master but without being able to run tests on jossl I'm not comfortable changing and releasing fixes there
<headius[m]> at the very least we need to tweak all reflective access to java.base to use the addOpens trick, which should eliminate all of those warnings once a for all
<headius[m]> long term we should stop using reflective access and try to use "proper" SPI access to BC, or just use standard JSSE whereever possible (like, RSA can't *really* produce different results on JSSE versus BC, right?)
<headius[m]> I hate crypto
<headius[m]> the remaining work on 9.3 should largely be around stabilization at this point
nirvdrum has quit [Ping timeout: 258 seconds]
thegeekinside has joined #jruby
<headius[m]> hmmm
<headius[m]> I hate this old Java integration stuff hanging around
<headius[m]> JavaAccessibleObject.setAccessible just uses AccessibleObject.setAccessible, so it fails a test that tries to do this to a core JDK class
<headius[m]> I'm torn on whether it should be more aggressive or not
<headius[m]> personally I wish we could remove these classes but they've been there for 15 years
<headius[m]> meh I think we leave it the way it is and skip this test on 9+
<headius[m]> I tried to make it use backport9 accessibility logic but since it's called from a module (jruby.dist) we don't use addOpens to open it up
<headius[m]> FWIW, patch I won't be pushing: https://gist.github.com/headius/1771afaab62d2bae31767edb4489500f
<chrisseaton[m]> Does the JRuby team still recommend either Drip or Nailgun today?
<chrisseaton[m]> GraalVM is considering dropping the kind-of-similar functionality in Native Image, as it just gets polluted and causes frustration.
<headius[m]> Drip was better, since the instances can only be as old as the last launched command, but we aren't working to support either anymore
<headius[m]> none of these options really help booting a large app from the filesystem, though, so they weren't very useful overall
<headius[m]> our policy at the moment is that userland preforking tools that work on JRuby are the best option
thegeekinside has quit [Quit: Leaving]