00:00 <headius[m]> I mean we can probably also eliminate frame stack but it's used in more places

00:01 <headius[m]> and any methods that need to read these out of band still need a way to find them, but e.g. a block that forces a dynscope could all escape if nothing else needs to get the scope from context

00:01 <headius[m]> deep thoughts

01:44 <travis-ci> jruby/jruby (master:3143288 by Charles Oliver Nutter): The build was broken. https://travis-ci.org/jruby/jruby/builds/589721007 [201 min 50 sec]

01:49 <travis-ci> jruby/jruby (master:3143288 by Charles Oliver Nutter): The build was broken. https://travis-ci.org/jruby/jruby/builds/589721007 [218 min 19 sec]

06:29 <kares[m]> headius: hmm, those time series sound interesting, did you find out where from?

06:31 <kares[m]> there will be lots of nanoTime() on enebo's counter nuke - marked that work WiP

06:31 <kares[m]> for now, not sure I will be able to gather real-world (Rails app) numbers in the end

06:32 <kares[m]> subjectively I feel like the jit.time.delta default should be higher, but nothing to back this

06:33 <kares[m]> or we could ship jit.time.delta as an experimental feature, off by default?

06:35 <kares[m]> shared class-loader would be great to get into 9.2.9

06:35 <kares[m]> but I need someone to review - esp. if naming sounds okay e.g. `-Xjit.loader.mode=SHARED_SOURCE`

06:36 <kares[m]> * shared class-loader would be great to get into 9.2.9

06:36 <kares[m]> but I need someone to review - esp. if naming sounds okay

06:36 <kares[m]> e.g. `-Xjit.loader.mode=SHARED_SOURCE`

06:36 <kares[m]> https://github.com/jruby/jruby/pull/5880

06:37 <kares[m]> also class-loader mode by default is as before -> `loader.mode=UNIQUE` loader per class

06:37 <kares[m]> * class-loader mode stays as before -> `loader.mode=UNIQUE` loader per class

06:39 <kares[m]> shared-source looks very promising and like it makes sense to be a default ...

06:39 <kares[m]> just wasn't sure if I am overly optimistic about any problems it might cause

12:07 <headius[m]> I'll review today

12:08 <headius[m]> Haven't looked into the millions of timing calls but my environment is back up and working now

13:48 <enebo[m]> kares: there was 0 tuning for that original value and it definitely affects some benchmarks

13:50 <enebo[m]> I have been thinking about the one nanoTime per 50 calls and that it will happen every 50 calls and this feels like something missing in the logic. Like if it fails on first 50 don't do it again for a while (what a while means is not clear to me yet)

13:52 <enebo[m]> kares: the other thing I want to figure out is how to mark time in method using the same counter

13:52 <enebo[m]> like threadpolls increment the counter as well

13:54 <enebo[m]> That is more complicated though since a method call can happen once do a million thread polls then 20s later get called again. counter is 1_000_000 > 50 but is 20s later so is not compilable

13:55 <enebo[m]> So I think this second dimension is detecting work for methods which are not called with regularity but some heuristic exists to promote it to JIT

14:21 <travis-ci> jruby/jruby (master:a67e86b by Karol Bucek): The build was fixed. https://travis-ci.org/jruby/jruby/builds/589968048 [225 min 0 sec]

14:34 <travis-ci> jruby/jruby (jit-exclude-block-working:e3ed02b by kares): The build was canceled. https://travis-ci.org/jruby/jruby/builds/589990581 [42 min 24 sec]

14:58 <travis-ci> jruby/jruby (jit-exclude-block-working:2bd4f41 by kares): The build passed. https://travis-ci.org/jruby/jruby/builds/589991983 [210 min 48 sec]

18:29 <travis-ci> jruby/jruby (master:117c75f by Thomas E. Enebo): The build was broken. https://travis-ci.org/jruby/jruby/builds/590079692 [203 min 33 sec]

18:38 <enebo[m]> blerg...something I should not have added +1 to I guess

18:38 <enebo[m]> looking into it

18:40 <travis-ci> jruby/jruby (master:117c75f by Thomas E. Enebo): The build was broken. https://travis-ci.org/jruby/jruby/builds/590079692 [221 min 28 sec]

18:43 <headius[m]> kares: reviewing now finally

18:46 <headius[m]> aargh so my broken build earlier was more Inflater closed stuff...I do not understand that one

18:49 <headius[m]> hmmm

18:49 <headius[m]> https://bugs.java.com/bugdatabase/view_bug.do?bug_id=7003462

19:10 <headius[m]> seems like we're closing all the relevant streams but I'll tweak the code a bit to try to make sure

20:16 <headius[m]> ok I pushed some code that flips a bit of load logic to use try-with-resources for closing those URL streams...maybe will help but it looked like we were cleaning them up before

20:21 <headius[m]> I will continue review after I change location

20:21 <headius[m]> bbiab

21:22 <headius[m]> ok

22:04 <headius[m]> kares: 👍 on the share CL PR

22:05 <headius[m]> I added my thoughts about eval

22:19 <travis-ci> jruby/jruby (master:0cde42b by Thomas E. Enebo): The build was broken. https://travis-ci.org/jruby/jruby/builds/590165132 [200 min 31 sec]

22:35 <travis-ci> jruby/jruby (master:0cde42b by Thomas E. Enebo): The build was canceled. https://travis-ci.org/jruby/jruby/builds/590165132 [217 min 9 sec]