14:19 <kares> jmalves: please do, did not realize that - kind of makes sense. thanks

14:20 <kares> jmalves: so the only problem is time sync than, right?

14:20 <kares> wonder how much of an issue that is using curTimeMillis for elapsed time with thread sleep/joins

14:22 <jmalves> kares: Not sure if it is the only problem, but the fact that currTimeMillis is not guaranteed to be monotonically increasing make it not very suitable for time deltas.

14:23 <jmalves> But I am not very much into the root of the problem you are trying to solve, might be that it is indeed fine

14:24 <kares> the problem is that I'd like to be able to have accurate time (running ntp) on a system with a long running JVM process

14:25 <kares> so the idea was to have the sleep/join primitives in JRuby's threading impl right

14:29 <jmalves> Is this related to `[refactor] should not use REALTIME for elapsed time #5546`? Its quite the quirky problem!

15:04 <kares> yes, that would be the first part - simply replace curTimeMillis

15:05 <kares> do some tuning later to avoid java.lang.Thread's join (does curTimeMillis) since JRuby does its own time pausing anyway

15:06 <kares> would like to be sure whether its worth pursuing ...

15:06 <kares> > people would pause a virtualized OS and restart and then get weird timing errors

15:08 <kares> compromise would be to allow switching with an --option, thought its not worth having curTimeMillis around

15:15 <travis-ci> jruby/jruby (master:77d1af4 by Uwe Kubosch): The build has errored. https://travis-ci.org/jruby/jruby/builds/474835613 [184 min 13 sec]

22:16 <headius> hey folks...back to it a bit today, officially back int the office Monday

22:27 <lopex> at LAST

22:29 <lopex> headius: I had flu too though, hope you;re ok now

22:29 <headius> yeah I'm mostly fine but everyone in the family is sick

22:29 <headius> I'm trying to stay well and not let a bit of sniffles and throat get worse

22:32 <lopex> headius: look at how hacky it is https://chao-tic.github.io/blog/2018/12/25/tls

22:32 <lopex> I think it might be interesting to you

22:33 <lopex> havent read thought all that though

22:34 <lopex> my main question was actually how java does that, since TL lookup is just a hash lookup ?

22:34 <lopex> so seems to go pthreads way

22:37 <headius> lopex: interesting

22:37 <headius> we moved away from thread-local access for context way back when because of that cost

22:38 <lopex> and how astonishing is that those GS / FS usage is so accidental

22:40 <lopex> also, recently read some rants why segmentation was so good, forgot where though

22:41 <lopex> apart from mem savings of course

22:54 <travis-ci> jruby/jruby (master:562e1f9 by Charles Oliver Nutter): The build has errored. https://travis-ci.org/jruby/jruby/builds/475524921 [185 min 43 sec]

23:03 <lopex> headius: have you seen https://rubyreferences.github.io/rubychanges/2.6.html ?

23:17 <headius> lopex: I have not

23:17 <headius> have not started to look at 2.6 work but I suspect that will be the big time sink the next couple months

23:17 <lopex> mostly gist from mri log

23:17 <headius> yeah

23:18 <lopex> no that many changes

23:18 <lopex> *not

23:18 <lopex> the endless range smells with a lot of bugs

23:19 <headius> yeah

23:19 <lopex> I tried to get on that \X thin but got sick too

23:20 <lopex> after that my priority will be match data, since we waste a lot if time for nothing there

23:23 <headius> yeah