<rtyler>
which is one of those nice fixes waiting in 1.7.2 :/
<rtyler>
.20*
bbrowning has quit [Ping timeout: 244 seconds]
camlow325 has quit []
havenwood has quit []
zorak8 has joined #jruby
zorak8 has quit [Max SendQ exceeded]
zorak8 has joined #jruby
zorak8 has quit [Max SendQ exceeded]
zorak8 has joined #jruby
zorak8 has quit [Max SendQ exceeded]
pjammer has joined #jruby
bbrowning_ is now known as bbrowning
subbu has quit [Ping timeout: 264 seconds]
pjammer has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
mitchellhenke has joined #jruby
subbu has joined #jruby
<bbrowning>
headius: regarding the memory leak with 1.7.20-SNAPSHOT, I'm seeing lots of runtimes being kept around by some ThreadLocal RubyClass instances. All the RubyClass instances have reifiedClass fields containing various java package names
<bbrowning>
I know that doesn't help much, but it's where I'm at so far
<bbrowning>
so something in the java integration changes that landed in 1.7.20, I think
fivebats has joined #jruby
subbu has quit [Ping timeout: 252 seconds]
<rtyler>
bbrowning: are 20-snapshot builds published anywhere?
<bbrowning>
rtyler: not that I know of - I'm just building from source
<rtyler>
rats
<rtyler>
I could really use some snapshot builds tonight :(
yfeldblum has quit [Remote host closed the connection]
skade has quit [Quit: Computer has gone to sleep.]
skade has joined #jruby
marr has joined #jruby
skade has quit [Quit: Computer has gone to sleep.]
havenwood has quit [Remote host closed the connection]
skade has joined #jruby
elia has joined #jruby
vtunka has joined #jruby
mister_solo has joined #jruby
vtunka has quit [Client Quit]
benlovell has joined #jruby
mister_solo has quit [Ping timeout: 255 seconds]
mister_solo has joined #jruby
skade has quit [Quit: Computer has gone to sleep.]
pitr-ch has joined #jruby
yfeldblum has joined #jruby
benlovell has quit [Ping timeout: 276 seconds]
yfeldblum has quit [Read error: Connection reset by peer]
yfeldblum has joined #jruby
benlovell has joined #jruby
dinfuehr has quit [Ping timeout: 256 seconds]
dinfuehr has joined #jruby
shellac has joined #jruby
vtunka has joined #jruby
vtunka has quit [Ping timeout: 256 seconds]
colinsurprenant has joined #jruby
<kares>
chrisseaton: is there a need for all truffle dev to happen on the same branch (master) as 9K (it's quite hard to track stuff) ?
vtunka has joined #jruby
vtunka has quit [Ping timeout: 250 seconds]
colinsurprenant has quit [Quit: colinsurprenant]
pjammer has joined #jruby
chamila has joined #jruby
dinfuehr_ has joined #jruby
dinfuehr has quit [Read error: Connection reset by peer]
anaeem1 has joined #jruby
anaeem1 has quit [Client Quit]
<chamila>
Hello all, I am trying to learn how to implement a jruby extension in java with modular level methods (eg - Oj.dump(), etc), can u suggest me a existing gem source where I can have a look and get a idea how can I do that?
<chamila>
Hi kares, thanks, is it enough to add a similar file to ext folder as in https://github.com/cdwijayarathna/GemTest, or is there any dipendencies to load the module?
<kares>
chamila: it's not that simple - you need to build the .jar and load it with your ruby code - while it's possible to compile .java on demand during gem build it's not common as with MRI's C exts
<chamila>
I did the loading parts with some classes defined on module in the above mentioned repo, now I need to add a module level methods to the same gem
a5i has joined #jruby
drbobbeaty has joined #jruby
<kares>
in that case yes ... you seem to compile everything under ext
shellac has quit [Remote host closed the connection]
<kares>
this is something you can try easily - thought you're asking on how to do module methods not all the details :)
<chamila>
kares: yeah, until now I'm loading classes using BasicLoadService, do I have to include this there as well
<JRubyGithub>
[jruby] chrisseaton pushed 6 new commits to master: http://git.io/vfghT
<JRubyGithub>
jruby/master a82a006 Chris Seaton: [Truffle] fcntl module used by Rubinius.
<JRubyGithub>
jruby/master d715d31 Chris Seaton: [Truffle] Simple server test.
<JRubyGithub>
jruby/master ce4af59 Chris Seaton: [Truffle] Rubinius::FFI::Library shim.
JRubyGithub has left #jruby [#jruby]
<chrisseaton>
kares: we originally started on a separate truffle branch - headius decided it was best to merge it
<chrisseaton>
kares: we've been experimenting with some feature branches for the bigger work recently, but some of the merges between them are horrific
<chrisseaton>
kares: I guess there will be a release branch at some point soon, which will achieve the same thing
yfeldblum has quit [Ping timeout: 245 seconds]
vtunka has joined #jruby
<kares>
chrisseaton: i see ... it's simply hard to follow progress (been trying that for a while now) since at this point I'm mostly interested in 9K commits but than most commits I tend to find are trufle specific
<kares>
but if headius suggested a separate branch for all trufle commits with periodical merges of master is a no-no than he probably knows why
djellemah has quit [Quit: Leaving]
<kares>
... oh, I realize the ugliness of a separate truffle branch since eventually it needs to be merged both ways
pjammer has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
pitr-ch has quit [Quit: My Mac Mini has gone to sleep. ZZZzzz…]
djellemah has joined #jruby
rsim has quit [Quit: Leaving.]
<chrisseaton>
kares: we prefix all Truffle work with [Truffle] so if you're reading a log you should be able to filter out
<chrisseaton>
kares: I'll be honest though and say we're probably going to get even noisier - we have a new hire starting next Friday making it four of us working full time
<kares>
chrisseaton: thanks ... using GitHub mostly :)
<kares>
chrisseaton: woow - I thought graal/truffle should make it less of an effort to implement a language :) !
<kares>
just trolling
<chrisseaton>
kares: we're just compressing it into a short amount of time
<chrisseaton>
plus eregon and I are also writing PhDs as we go
<kares>
chrisseaton: is that another language :) ?
<chrisseaton>
kares: try writing a truffle language yourself - lucasallan did Lua and got it running fibbonaci not far off luajit with just a few hours of work
<kares>
yeah I know ... Ruby is quite hard compared to some
pierreatkillbill has joined #jruby
bbrowning_ is now known as bbrowning
<kares>
I'm actually looking forward for a truffle Ruby ... it's just that there's seems to be a lot of work on master that is of 2 kinds
<chrisseaton>
yeah I know - and we really don't want the impression we are swamping or taking over - 9k is at the tweaking phase, and we're at the green field phase
<chrisseaton>
if the community wants us to use a different workflow we will - but merges are the worst
<headius>
kares: imagine merges with the amount of activity happening
<headius>
I'm still amazed merges from 1.7 usually work
<kares>
headius: yeah I know ... but there might be some examples of even more activity on separate branches e.g. linux?
<headius>
sure...they may have more patience for merges :-)
<headius>
chrisseaton: it may make sense to branch your work as we approach 9k final, since we'll move more into maintenance at that point
<headius>
rather than branching just for 9k maintenance
<chrisseaton>
chrisseaton: we can work with that - just let us know when you want it to happen
<headius>
kares: if we'd had time it would have been nice to make them separate modules atogether, but the rest of JRuby wasn't really set up for that yet
<headius>
they're separate now, so it might be possible to do that
<kares>
headius: have not thought of that ... soundz good!
<headius>
kares: oh, one big thing you could do is exclude truffle/ from logs
benlovell has joined #jruby
<headius>
jruby proper depends on nothing from there and can run without any of it present
nateberkopec has joined #jruby
benlovell has quit [Ping timeout: 240 seconds]
dinfuehr_ has quit [Remote host closed the connection]
<chrisseaton>
headius: do you know jnr-unixsocket or was that Wayne Meissner? Trying to figure out why there is no getaddrinfo call
mister_solo has quit [Ping timeout: 264 seconds]
vtunka has quit [Quit: Leaving]
nirvdrum has joined #jruby
gazay has quit [Ping timeout: 252 seconds]
<headius>
chrisseaton: well I know it, but I assume he just never thought it was needed
gazay has joined #jruby
<headius>
addrinfo comes back as a struct anyway, right? mapping structs is a pain
<chrisseaton>
yeah I think there might be some higher level way I'm supposed to do - as it's the one obvious missing call and it's pretty essential
<chrisseaton>
do you know where he hangs out or what his handle is?
skade has joined #jruby
zorak8 has joined #jruby
<headius>
he quit the internet
<headius>
according to my gtalk client, I haven't seen him online in 11 weeks
<headius>
we're on our own :-)
zorak8 has quit [Ping timeout: 272 seconds]
dinfuehr has joined #jruby
dinfuehr has quit [Read error: Connection reset by peer]
ahadding1 has quit [Ping timeout: 245 seconds]
benlovell has joined #jruby
pjammer has joined #jruby
zorak8 has joined #jruby
<chrisseaton>
maybe I can reimplement it using the Java APIs anyway
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton pushed 1 new commit to truffle-head: http://git.io/vf24b
<JRubyGithub>
jruby/truffle-head 8f883f5 Chris Seaton: Merge branch 'master' into truffle-head...
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton pushed 2 new commits to truffle-head: http://git.io/vf2BZ
<chrisseaton>
Is there a flag to tell JRuby to use lib from somewhere other than the complete jar?
codefinger has joined #jruby
pjammer has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<chrisseaton>
ah just JAVA_HOME
pjammer has joined #jruby
<bbrowning>
headius: not sure if you saw, but the memory leak culprit in 1.7.20 builds appears to be the new ClassInitializer and InterfaceInitializer stuff
<bbrowning>
they end up setting RubyModule instances as Java ThreadLocals which never get cleared when the runtime is torn down
<bbrowning>
so runtimes never get GC'd
mrmargolis has joined #jruby
<bbrowning>
the name seems to imply that the work here isn't finished? "unfinishedProxyClassCache"
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] lbianc opened issue #2873: Class not found exception on ppc64le http://git.io/vf2NU
JRubyGithub has left #jruby [#jruby]
mitchellhenke has joined #jruby
pjammer has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<headius>
I have no concern about doing it for 1.7 line as well
<enebo>
dump:poms has random extra lines
<headius>
if someone is upset that we dropped it we can pull the security card
<headius>
enebo: the installer can prompt to go get java right?
<enebo>
yeah I agree. It is not a horrible idea to have jvm-enabled installer if you are brand new to JRuby and a Ruby windows developer but that is splitting the hair
<enebo>
headius: prompt? like a single message saying, go get it?
<enebo>
headius: or provide a link?
<headius>
well it's an installer for Java apps, I assume there's some sort of warning or bailout or something
<enebo>
headius: it is in fact written in Java
<enebo>
headius: that is partially tangential :)
<headius>
to early for meta
<enebo>
headius: but yeah I don’t remember. I think so
<headius>
too too
<headius>
ok
<headius>
well, let's just not publish it for this next round and see if anyone cares
skade has quit [Quit: Computer has gone to sleep.]
<headius>
the jre version
<headius>
I don't really get the windows user obsession with installers, but whatever
<headius>
I guess it's because Windows is so damn hard to navigate
<chrisseaton>
if it's an MSI installer than I think it interacts well with the huge policy and admin systems large Windows IT departments have
<enebo>
chrisseaton: although install4j is not an msi
<headius>
there's that, indeed
<chrisseaton>
I used Windows for the first time in ages to try to build Graal - and it was genuinely a sad experience
<headius>
enebo: does it have an option for it?
<headius>
chrisseaton: did you actually succeed?
<chrisseaton>
not yet!
<headius>
I've never attempted to build openjdk on Windows but it sounds horrific
<enebo>
headius: perhaps? but our licensed copy is actually quite old (from codehaus days)
<headius>
enebo: yeah that's pretty old
<enebo>
I think it is install4j 4.x
<enebo>
ah yeah they are on 6.x
<headius>
looks like 6 is "very customizable" when it comes to detecting JDK, error handling, etc
<headius>
I don't see anything about msi though
<enebo>
700$ for single dev license
<enebo>
with no upgrades
<headius>
at least one forum post says they have no plans to publish an MSI because the two are "different concepts"
<enebo>
hmm
<headius>
it must be weird in the world of installers if those two are considered "different concepts"
<enebo>
ok well I say we limp this along and look at free msi installers at some point
<headius>
anyway, every IT department I know either handles non-MSI stuff or wouldn't be installing Java, much less JRuby
<chrisseaton>
When I was a teenager I remember desperately wanting something like InstallShield to make my apps look more professional :)
<chrisseaton>
That blue gradient background during install meant you were a proper developer back then
<headius>
chrisseaton: I published a piece of shareware for some years in the 90s using a free installer
<headius>
if I had a few beers I might remember what the installer was called
<headius>
the software was Hack-It...as far as I know it still works and I still get people trying to register it
<headius>
enebo: ok, so back to biz
subbu has joined #jruby
<headius>
I'm only afraid enough of the crypto hack to get a thrill, like riding a roller coaster
<headius>
removing jre from installers mitigates the last concern I think
<headius>
so there it is...moving on
<headius>
all builds have been green at some point during the past few days :-)
<headius>
they probably aren't right now
<headius>
failure on 1.7 last night looked like a JDK6 bug and I haven't done anything about that yet
<headius>
some thread-safety issue deep inside jsse
havenwood has joined #jruby
<headius>
bbrowning found a memory leak in 1.7.20 related to recent JI fixes and work, so I'll be working on that and greening up build today
<enebo>
headius: ok but pre2 first then 1.7.20 since both will not happen in 1 day
<headius>
some person called gazay has been around on and off reducing rails failures, so you may see a few of those bugs reported
<enebo>
headius: and I have to pick Valerie up at 1:30 today
<headius>
ok
<enebo>
headius: so I plan on building bits and having it ready
<enebo>
headius: I just hope nothing is found :)
<headius>
the JI work is the same in 1.7.20 and 9k so I will do it in 1.7 and merge
<headius>
but I'll make sure 9k is green first
<enebo>
headius: ok. I will then probably mostly just make sure dist bits are consistent and windows more or less works
<headius>
yeah cool
<headius>
I'll keep you posted and try to do the minimal fix for bbrowning issue
<headius>
bbrowning: how fast can you turn around confirmation once I get fix in?
<bbrowning>
headius: 1.5 hours or so, probably? have to build jruby, build TB, run TB3 integs
<bbrowning>
it takes about 30 minutes into the integs for the OOM to start showing up
<headius>
ok
<headius>
I'll work quickly, like a MASH doctor
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] enebo pushed 1 new commit to master: http://git.io/vfalC
<JRubyGithub>
jruby/master d77456d Thomas E. Enebo: Remove JRE installer media entries from windows installer
JRubyGithub has left #jruby [#jruby]
<bbrowning>
you don't have to rush on my behalf
<headius>
well, I'm rushing for pre2
<headius>
and I'm not rushin' cuz I'm not Russian
baroquebobcat has joined #jruby
<enebo>
bbrowning: it’s always about you
benlovell has quit [Ping timeout: 248 seconds]
<bbrowning>
enebo: hey you asked to to test against 1.7.20 :)
<enebo>
bbrowning: :)
<enebo>
bbrowning: you can’t spell bbrowning without I
camlow325 has joined #jruby
<Aethenelle>
gonna be in san jose next week if anyone wants to meetup...
mkristian has joined #jruby
e_dub has joined #jruby
rcvalle has joined #jruby
<nirvdrum>
enebo lives!
<mkristian>
enebo, your psych problem is a new one but it is due to fact that gem install uses the jruby from the build which has no psych installed but needs psych to install the psych gem
<enebo>
mkristian: yeah i noticed it uses itself
<enebo>
mkristian: we talked about an ENV var for working around this (which I guess would only be used for dist’ing
<mkristian>
enebo, this is a local branch of yours ? do not see 9.0.0.0.pre2 anywhere
<enebo>
mkristian: I only changed VERSIONS and a dump:poms so other than a few random line changes from that tool it is only a version update commit
<enebo>
mkristian: I do not push that commit because I am only testing that disting works
kares has joined #jruby
<mkristian>
enebo, give me an hour to find a way - otherwise we use jruby-1.7.12 for installing psych
<mkristian>
enebo, I never got to this ENV or build.properties settings. will do it now
<enebo>
mkristian: ok
<mkristian>
enebo, ok - will change me VERSION and go from there
<enebo>
mkristian: yeah. I think you will see a couple of extra lines written out to poms here and there
<enebo>
mkristian: seems like minor difference between pom.rb and generated poms
skade has joined #jruby
mrmargolis has quit [Ping timeout: 256 seconds]
<headius>
Aethenelle: ahh nice...I wish I were going to be out there too :-)
<nirvdrum>
I suspect if there were ruby-build/RVM/ruby-install on Windows, you'd see less of a desire for a JRuby installer. I think there's one project kinda trying to do it, but I'm not sure where it's at.
<Aethenelle>
headius: the week after is near philadelphia...
<mkristian>
mkristian, I do see version changes and a one or two extra <jruby.home> lines
mrmargolis has joined #jruby
<enebo>
nirvdrum: there is one but I cannot remember its name
donValentin has joined #jruby
donV has quit [Ping timeout: 265 seconds]
mitchellhenke has quit [Read error: Connection reset by peer]
mitchellhenke has joined #jruby
<chrisseaton>
I think the last truffle-core failure might have been a fluke - I'll see how it goes again
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] kares pushed 1 new commit to jruby-1_7: http://git.io/vfa5U
<JRubyGithub>
jruby/jruby-1_7 f7b8a8a kares: deprecation warning when using the old `__jsend!` internal method (no longer in 9K)
JRubyGithub has left #jruby [#jruby]
<rtyler>
mkristian: YO
<rtyler>
mkristian: while you were out yesterday Dirk got pwnt by that "+" in the jar filename issue again; what's left on your plate for 1.7.20?
lanceball is now known as lance|afk
subbu is now known as subbu|afk
pietr0 has joined #jruby
mrmargolis has quit [Read error: Connection reset by peer]
mrmargolis has joined #jruby
<headius>
kares: got a moment to review this change?
<kares>
headius: yy
<headius>
pushing in a moment
<headius>
sorry, I just realized a problem with it
<codefinger>
it occurs on a handful of specs that use jrubyc (from within warbler)
lanceball is now known as lance|afk
<codefinger>
i'd have to play around somemore, but it seems i can reproduce it by running `rvm install jruby-9k` with JDK7, and then running `jruby ...` with JDK8. Is that possibly a cause?
<codefinger>
maybe that's wrong, i just got the same on jdk8. too many variables...
<bbrowning>
codefinger: perhaps StandardASMCompiler doesn't exist in 9k?
<bbrowning>
I don't see it
yipdw_ is now known as yipdw
yipdw has joined #jruby
yipdw has quit [Changing host]
<bbrowning>
codefinger: so I'd expect this to fail any time it's run under jruby 9k, regardless of JVM version
<codefinger>
warbler has a little jar in it. i think i may need to recompile that against jruby-9k
<headius>
yeah that class is gone in 9k
<bbrowning>
so warbler provides StandardASMCompiler in 9k?
<headius>
ahh pre1
<headius>
I fixed that post-pre1
<headius>
codefinger: I have not done a great deal of testing of the new AOT impl...looks like you're hitting that here
<headius>
if you run with a snapshot you should at least get past that
<headius>
man, if we could use java 8 we could use lambdas
<headius>
that would be so slick
<headius>
and it would totally kill merging from 1.7 because I can see using lambdas everywhere
<nirvdrum>
Keep in mind the Ruboto problem :-/
<headius>
there is that :-\
<nirvdrum>
But I sure would love to use default methods.
<headius>
we just need a substrate vm build for android
<headius>
android could adopt java 8 syntax if they convert the lambda bootstraps into lazy static init of inner classes
<nirvdrum>
I've advocated in the past for keeping a long-term 1.7 branch around for Android, but I suspect limiting to 1.9 support would be problematic for people as the ecosystem embraces 2.0+ features.
<headius>
nirvdrum: yeah, 1.9 is a little far behind now
<nirvdrum>
I think they use it at the very least. Probably saved them a lot more than it cost.
<lopex>
well, otoh, I'm surprised that's the case :)
<lopex>
hardware wise ?
<nirvdrum>
Yeah. I mean, I'm just speculating.
<lopex>
but it seems to be a very narrow portion of code the benefit from
<nirvdrum>
But at their scale, a 10% increase in performance has to be huge.
<lopex>
it's almost like it would be worth rewriting in another lnaguage
<lopex>
yeah
<nirvdrum>
I think what they're doing makes sense. Rewriting would not be trivial. And my guess is they're able to hire a bit cheaper by drawing from the PHP pool.
<nirvdrum>
Just as I think it makes more sense to make JRuby faster than to rewrite in Go :-)
<lopex>
but the effort doesnt seem to be centralized
<lopex>
lots of tools, approaches etc
<nirvdrum>
Yeah.
<lopex>
let's wait for rust hype
x1337807x has joined #jruby
<donValentin>
Hi all!
<donValentin>
headius: Using 9k on Ruboto would great!
yfeldblum has quit [Remote host closed the connection]
yfeldblum has joined #jruby
<tarcieri>
headius: ohai
<headius>
tarcieri: yo
<tarcieri>
so I tried all of those changes... and if anything it seems slower o_O
<headius>
donValentin: what if that weren't possible until Android supports Java 8 features? :-)
<headius>
really?
<tarcieri>
worse, VisualVM is crashing and YourKit seems to make it even worse
<donValentin>
headius: I have not heard of any plan for Android to ever support Java 8 features.
pipework is now known as |work
<tarcieri>
headius: I wish Concurrent::ThreadPoolExecutor had a batch API for submitting work o_O
* tarcieri
tries YourKit again...
|work is now known as pipework
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] bjfish pushed 1 new commit to master: http://git.io/vfr4V
<JRubyGithub>
jruby/master 9f60295 Brandon Fish: [Truffle] Add Kernel#test to kernel.rb.
JRubyGithub has left #jruby [#jruby]
<tarcieri>
headius: maybe I should try to completely eliminate any usage of blocks
<donValentin>
headius: maybe even Java 8 features.
<headius>
tarcieri: don't see why they couldn't add a batch API
subbu has quit [Ping timeout: 265 seconds]
<headius>
donValentin: yeah, I think some things work, but not enough for us to switch to 8 *and* start using language features
<tarcieri>
headius: I'll keep poking at it and see if I can figure out what the bottleneck is
<headius>
ok, keep me posted
<headius>
I can try it later too, but bugs right now
<tarcieri>
headius: but so far, just reading data out of those arrays and putting it into the pool, I can't seem to get any faster than ~30,000 messages / s
<tarcieri>
which was like... a few orders of magnitude slower than I was hoping
<tarcieri>
perhaps I should go with my original plan of just reimplementing the whole thing as a Java extension to JRuby, heh
mrmargol_ has quit [Remote host closed the connection]
<lopex>
tarcieri: not using blocks is enough of an excuse
<headius>
tarcieri: I'm sure we can improve it
<donValentin>
headius: We can use JRuby 1.7.x for many months as long as it is supported, but at some time we need to switch to 9k.