<GitHub61>
jruby/master 0269680 Pavel Rosický: fix quoted paths
e_dub has quit [Read error: Connection reset by peer]
e_dub has joined #jruby
momomomomo has joined #jruby
camlow325 has joined #jruby
camlow32_ has joined #jruby
camlow325 has quit [Read error: Connection reset by peer]
nirvdrum_ has joined #jruby
Aethenelle has quit [Ping timeout: 240 seconds]
<dfr>
headius, Hey, I've added some more info about the crash. You're right, I really need to verify with latest 1.7 branch to make sure it actually happens.
nirvdrum_ has quit [Ping timeout: 244 seconds]
<headius>
dfr: yeah, generally a JVM crash isn't our fault (unless it's in FFI or something obviously) so we need to know latest stuff fails and need a way to reproduce
<headius>
otherwise JVM folks won't look at it
<dfr>
headius, I think it's interplay between garbage collector and JRuby's JIT
<dfr>
I have a feeling that GC tends to remove JITted methods or something like that, 'causing crashes when it tries then to evaluate them.
<dfr>
but I'm a pretty noob with so deep in JVM stuff, so just a hunch :)
<headius>
that would be pretty weird but still nothing we're doing wrong
<headius>
we're not injecting native code into JVM or anything...everything should be above board
<dfr>
headius, yep, could be a bug in CMS =/
<headius>
that's certainly a possibility :-)
<dfr>
headius, I just wish I could grab some JRuby issues, so that they're attached to me rather than hanging unassigned :)
<dfr>
and props for doing spring cleaning on the issues queue :)
<headius>
google 1.8 eh?
<dfr>
headius, what surprises you about it? :)
<dfr>
I don't think google's JRE deviates a lot from OpenJDK, but the issue did start with a google JRE build upgrade, so it is possible that it's a Google bug. That said, folks did say that that update was mostly pulling from upstream.
<dfr>
The unfortunate part is that my repro case is pretty bulky [including a barebones rails app].
bbrowning_away has quit [Ping timeout: 244 seconds]
<headius>
dfr: turning off method flushing still fixes it gtoo?
<headius>
yeah I don't figure it's very different, just never seen that build in the wild
<headius>
gtoo=too
<dfr>
headius, It mostly fixes it, but not reliable, I believe. We still run in prod and automated tests with method flushing disabled and it seems to work, but I _think_ that with extreme JIT settings and method flushing disabled, I still saw some failures.
<headius>
well if you've got cycles to get some scenarios for us, that's great...but perhaps you won't be able to release a reproduction for us
<dfr>
headius, what I really need to do is to reduce the repro to a jar, so that it can be tested with custom JDK java and whatever JRuby jar.
<headius>
yeah
<headius>
that would be helpful
<headius>
but a multi-step process that's basically "unpack this and run that" is ok too
<dfr>
headius, I don't think there's _that_ much custom code in repro, so I think I'd be able to clean it up.
bbrowning has joined #jruby
<headius>
dfr: I posted more info as well...there's a similar-looking bug with concurrent GC that's apparently fixed in 9
<headius>
I'm not sure if it's related to our bug but it shows they may have some concurrency issues in this flushing
<dfr>
headius, hmm. I think my repro is serial, but the GC is concurrentsweep, so maybe?
<headius>
I know Vladimir so if we can get something to reproduce he will probably help us
<dfr>
either way, that's good info, thanks. :)
<dfr>
headius, I also have Google java folks in the loop, although they cannot really do much until I get them a clean repro as well. So all hinges on me. :)
<headius>
ok, keep us posted then
skade has quit [Quit: Computer has gone to sleep.]
<reto_>
illegal access on 'getStart': Class org.jruby.javasupport.JavaMethod can not access a member of class COMPANY.PROJECT.Graph with modifiers "public"
<reto_>
fuck!
<reto_>
the class itself is not public
<headius>
reto_: hi there
<headius>
how are you trying to call this?
<headius>
JavaMethod and that stuff is pretty much deprecated...I'd recommend going a different route
<reto_>
class Foo {
<reto_>
public static void method() {}
<reto_>
}
<reto_>
note the missing public on the class
<reto_>
Foo.method will not work
<reto_>
that was confusing..
<reto_>
same with instance methods, non public class is super confusing
momomomomo has joined #jruby
<reto_>
then I tried java_send which gave me the 'illegal access' error which made me realize that I forgot to make the class itself public
<headius>
hmm
<headius>
well #method is also a method in Ruby
<headius>
I'm don't think we wipe it out for static methods in Java
<reto_>
'no access to method' would have been helpful
<headius>
did you just get NoMethodError I assume?
<reto_>
yap
pitr-ch has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<headius>
we have thought about binding private methods anyway with Ruby private visibility, but people would expect visibility-changing and send to be able to call them
<headius>
which would only work if we could modify the Java methods too
<headius>
it's worth exploring though
<reto_>
or warn when calling anything on a private class?
<headius>
ahh private class...yeah perhaps
<reto_>
I looked everywhere, but not private class :)
<reto_>
anywho, i'm in the weeknd
<reto_>
see ya! thanks for the help
pilhuhn has quit [Quit: Leaving...]
<headius>
ta ya!
<headius>
er, ta ta!
pitr-ch has joined #jruby
momomomomo has quit [Quit: momomomomo]
pawnbox has joined #jruby
shellac has quit [Quit: Leaving]
pawnbox has quit [Ping timeout: 276 seconds]
donV has quit [Quit: donV]
lacce has joined #jruby
deathy has quit [Remote host closed the connection]
bruceadams has quit [Remote host closed the connection]
Scorchin has quit [Remote host closed the connection]
fidothe has quit [Remote host closed the connection]
aemadrid has quit [Write error: Broken pipe]
guilleiguaran__ has quit [Write error: Broken pipe]
deepak has quit [Write error: Broken pipe]
mccraig has quit [Write error: Broken pipe]
knowtheory has quit [Remote host closed the connection]
flavorjones has quit [Remote host closed the connection]
pitr-ch has quit [Ping timeout: 246 seconds]
<headius>
hmm
<headius>
ugh
pitr-ch has joined #jruby
deepak has joined #jruby
mccraig has joined #jruby
fidothe has joined #jruby
Scorchin has joined #jruby
deathy has joined #jruby
aemadrid has joined #jruby
<headius>
those failures on travis don't happen locally...but it's just an array spec