<headius[m]> Oh yes, I remember a couple issues they filed
_whitelogger has joined #jruby
ur5us has joined #jruby
ur5us has quit [Ping timeout: 260 seconds]
_whitelogger has joined #jruby
nirvdrum has joined #jruby
ur5us has joined #jruby
reidmorrison[m] has quit [Quit: Idle for 30+ days]
kalenp[m] has quit [Quit: Idle for 30+ days]
_whitelogger has joined #jruby
_whitelogger has joined #jruby
<kares[m]> csharpsteen: Hey, have seen this in the past ... at a big Rails app.
<kares[m]> did not investigate fully at what's going on -> was an consulting and had no time to look at what's being thrown away from code-cache
<kares[m]> but I recall seeing a bug on Java 8 around code-cache sweeping ... not sure if it's still a concern on 11
<kares[m]> the conclusion I've reached is to stop the JRuby JIT after some time as it was stil jitting several hours after start ...
<kares[m]> you can try setting it lower if you still see JRuby jitting after several hours and see if it helps the code-cache issue
<kares[m]> I did not hear from the app whether they got around the issue with jit.max support in 9.2.9.0
<kares[m]> byteit101: just wanted to say there's great stuff in https://github.com/jruby/jruby/pull/6141 :)
_whitelogger has joined #jruby
ruurd has quit [Ping timeout: 256 seconds]
ruurd has joined #jruby
enebo has quit [Ping timeout: 256 seconds]
<headius[m]> kares: no doubt, great to see these two areas start to unify
<headius[m]> regarding code cache... I would like to know what's continuing to compile forever
<headius[m]> one-off generated methods shouldn't JIT, and if they're used heavily they probably should... so I am somewhat reluctant to encourage people to throttle the JIT without knowing why things are jitting
<headius[m]> enebo: I think we should just go with jnrproject.org
<enebo[m]> headius sure makes sense and it is not remotely confusing
<headius[m]> module and package will (eventually) be org.jnrproject
<enebo[m]> funny how many people use names like jillbilly13 because some others used none-12 :) OSS projects need to go that route
<enebo[m]> (made up the name jillbilly on the spot...but I bet there is a username like that somewhere)
<headius[m]> I guarantee there's someone named Jill from Appalachia with that username
<headius[m]> chrisseaton: so you have used name.com and are happy with it?
<kares[m]> headius: should have stressed jit-ing after hours of up time (heavy-load). this goes back to Rob's Rails app that we have discusions almost a year back ... would also like to understand why JRuby did not stop JIT-ing
<headius[m]> one of the tests I do while benchmarking webapps is to turn on JIT logging and make sure it eventually stops
<headius[m]> there's also JMX endpoints to track it
<headius[m]> hopefully they're still working, I don't think we have tests
<headius[m]> $60 for five years of jnrproject.org
<headius[m]> DM me your credit card number enebo
<enebo[m]> lol
<headius[m]> jnrproject.org lives
<csharpsteen[m]> kares: Thanks for the info! This is JRuby 9.2.11.1. The interesting bit is that _something_ is able to eventually free nearly 50% of the `nonprofiled` methods usage. Since this is Java 11, I'm looking into whether the Java Flight Recorder could shed some light. Looks like it may have some event probes down here in the code cache sweeper.
<headius[m]> and google pages is up
<headius[m]> next question is what to do with the dashes in the JNR project names
<headius[m]> they can't be used as module names
<headius[m]> org.jnrproject.ffi for jnr-ffi?
<headius[m]> of course oss.sonatype goes down for maintenance right now
<headius[m]> these javadoc errors are getting pretty specific
<headius[m]> [ERROR] /home/runner/work/jnr-ffi/jnr-ffi/src/main/java/jnr/ffi/Pointer.java:29: error: tag not supported in the generated HTML version: tt
<headius[m]> [ERROR] * This class provides operations on a native memory address. Most <tt>Pointer</tt>instances will represent direct
<headius[m]> can't say I even knew about that tag
<headius[m]> hmm this is an interesting one
<headius[m]> because jnr-constants has a Platform class in the same location as a platform package the strict javadoc kicks it out
<headius[m]> generally folks shouldn't be poking around in the platform package (that's where the generated constant tables are) so I guess we could rename it?
<headius[m]> "platforms"
<headius[m]> why is github's search so fuzzy
<headius[m]> search for jnr.constants.platform also brings up jnr/constants/platforms
<headius[m]> -s
<headius[m]> JRuby itself doesn't reference anything under jnr.constants.platform so I think rename is pretty safe
<headius[m]> enebo: any objection?
<headius[m]> I want this release round to also build and test green on Java 11+ without tweaking
<enebo[m]> yeah I think it is ok. I mean something needs to change so it might as well be with major number change
<headius[m]> the error is largrely right... on Windows and MacOS you can't have a package platform.html and a class Platform.html
<headius[m]> I am doing jnr-constants 0.10.0 from 0.9.14 rather than 1.0 ... I think the major flip for each of these should probably update packages to reflect module name
<enebo[m]> still a siginificant change and based on most peoples general opinions on numbers < 1.0 means developing the final aPI
<headius[m]> oh drat... actually I just realized EVERYTHING is in this platform dir
<headius[m]> all the baseline constant enums are there, and then in linux/windows etc subpackages are the platform specific stuff
<headius[m]> seems like the wrong layout to me... you'd expect all the constant tables to be in jnr.constants
<headius[m]> I mean all the public enums
<headius[m]> hmph
<headius[m]> so this is a much more visible change then
<headius[m]> so maybe renaming the class is better then
<headius[m]> we don't appear to reference this class anywhere
<headius[m]> in JRuby
<enebo[m]> but we do reference another class Platform right?
<headius[m]> this largely mirrors the Platform class we have in JRuby, jnr-ffi, and other places
<enebo[m]> I sort of remember there are two of them
<headius[m]> yeah there's like a half-dozen of these
<enebo[m]> hahah yeah ok more than two then
<headius[m]> they're all basically doing the same thing
<enebo[m]> It is our 'Context'
<headius[m]> hah yes
subbu is now known as subbu|lunch
subbu|lunch is now known as subbu
<headius[m]> what a quagmire
<headius[m]> stupid Buffer methods
ur5us has joined #jruby
<headius[m]> ok, I think they're all set
<headius[m]> going to push a branch of JRuby that runs CI against the snapshots
sagax has quit [Remote host closed the connection]
sagax has joined #jruby
<headius[m]> that maven cache on travis is sometimes more trouble than it's worth
ruurd has quit [Read error: Connection reset by peer]
ruurd has joined #jruby
<headius[m]> wots all this then
nirvdrum has quit [Ping timeout: 256 seconds]