<JRubyGithub>
jruby/master 0b6787b Thomas E. Enebo: Exclude out symbol encoding mismatch for the moment to re-green ci
JRubyGithub has left #jruby [#jruby]
e_dub has quit [Quit: e_dub]
<headius>
enebo: at work a bit early today?
anaeem1 has joined #jruby
<enebo>
headius: no not really
<headius>
I usually wait until 9:01 to message you
<headius>
on the days I'm up early, that is :-)
<enebo>
headius: ah. I do usually read email and crap before I start committing
anaeem1 has quit [Read error: Connection reset by peer]
<enebo>
headius: this morning the first thing I did was look at fallout from the symbol fix
<headius>
ahh yes
<headius>
do you know what's wrong?
<enebo>
headius: well somewhat. These symbols are created during boot from various encoded files since they are operators and they have random encodings
<enebo>
headius: So there is some logic which needs to say, hey guys you are ascii
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] mkristian pushed 7 new commits to master: http://git.io/AmhkGQ
<JRubyGithub>
jruby/master da96af9 Christian Meier: refactor FileResource.inputStream to be easier to use...
<JRubyGithub>
jruby/master fb0281b Christian Meier: redefine uri:classloader: meaning to be the parent classloader of runtime.getJRubyClassLoader...
<JRubyGithub>
jruby/master 40e4c8f Christian Meier: add testcase for the case where jruby.home is inside a jar but not on the context-classloader
JRubyGithub has left #jruby [#jruby]
<nirvdrum>
lopex: Why do encoding and encoding entries have different indices?
<enebo>
That must have been an exciting merge
<enebo>
going to update jnr projects now
<enebo>
I think they are all done
<mkristian>
yes it was exciting indeed
<enebo>
mkristian: Does ‘jruby -S rake maven:dump_poms’ not work when run with master?
anaeem1 has joined #jruby
dfas has joined #jruby
<mkristian>
enebo, just checking . .
<enebo>
mkristian: A second enhancement also would be to generate stuff stored in hashes in a sorted order to not noise up generation diffs:
<JRubyGithub>
[jruby] headius pushed 1 new commit to jruby-1_7: http://git.io/dBx-RQ
<JRubyGithub>
jruby/jruby-1_7 0576fd5 Charles Oliver Nutter: Handle NCDFE for odd envs like GAE that throw it for Unsafe....
JRubyGithub has left #jruby [#jruby]
<enebo>
I will commit my new pom.rb and the pom.xml it generated (which I think is the only thing which would change from new artifact versions
<mkristian>
just remove the require 'rake' in maven/jruby-jars/jruby-jars.gemspec and rerun the rake task
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] headius closed issue #2304: Unable to use JRuby with google app engine due to use of sun.misc.Unsafe http://git.io/7IVNcg
JRubyGithub has left #jruby [#jruby]
anaeem1 has quit [Ping timeout: 240 seconds]
<enebo>
mkristian: I will try that in a sec
anaeem1_ has quit [Remote host closed the connection]
<enebo>
mkristian: Changing that I get:
<enebo>
[ERROR] Failed to execute goal on project jruby-lib: Could not resolve dependencies for project org.jruby:jruby-lib:pom:9.0.0.0-SNAPSHOT: Could not find artifact org.jruby:jruby-core:jar:9.0.0.0-SNAPSHOT in rubygems-releases (http://rubygem
<mkristian>
OK - I fix the task for those cases that there a no local snapshots.
<enebo>
mkristian: oh but I did not have your last merge commit
<enebo>
mkristian: Could that last merge possibly have helped?
mitchellhenke has quit [Quit: Computer has gone to sleep.]
<mkristian>
the last error is usually because maven can not find the SNAPSHOT from core/ - this happens for example when you change the VERSION and then run the rake tasks.
elia has quit [Quit: (IRC Client: textualapp.com)]
brettporter has joined #jruby
<enebo>
headius: once you see that I would like to respin so this will end up green
<headius>
yeah I know this failure
<headius>
but I thought I mitigated it
<headius>
hmm
<enebo>
headius: I thought semaphore fix fixed it
<headius>
this is something else
<enebo>
oh
<enebo>
restarted
<headius>
ThreadFiber uses SizedQueue and normal Ruby thread eventing, but for this test case sometimes the interrupt comes in before it is ready to handle it properly
x1337807x has joined #jruby
phrinx has joined #jruby
<headius>
oops, I didn't catch you in time
brettporter has quit [Ping timeout: 245 seconds]
<headius>
pretty sure I know the issue but I didn't copy the exception in time
<headius>
if you see it again file an issue for me
mister_s_ has joined #jruby
phrinx has quit [Ping timeout: 260 seconds]
e_dub has joined #jruby
<headius>
enebo: did you see capture3 error on last 1.7 build?
<headius>
my fault I guess...probably a 1.9-only API that I added a spec for
<JRubyGithub>
jruby/master badca70 Thomas E. Enebo: Passing now (probably from recent symbol fixes)
JRubyGithub has left #jruby [#jruby]
<headius>
some days I just want to sit and watch builds turn green
brasten has joined #jruby
<rtyler>
I want to watch builds turn green every day :(
<nirvdrum>
I started caring a lot more about green builds after I learned that Travis won't promote a JRuby dev build that was red.
<headius>
indeed
<headius>
nirvdrum: looking at the evalScope thing now
<headius>
the code is still used but I'm not sure it should be
<enebo>
gah well 43839718 should flip green after two resta rts
<nirvdrum>
Cool. That was one I didn't feel comfortable just changing myself.
<headius>
enebo: two?
<nirvdrum>
Implementing stuff in Truffle that JRuby has had for years gives me the opportunity to revisit some stuff that otherwise wouldn't be looked at.
<headius>
nirvdrum: indeed...and hopefullly more improvements will fall out of that process
<nirvdrum>
Yeah. I'm lazy, so I don't want to reimplement stuff that I don't need to.
<nirvdrum>
After 15 years, I do find myself finally wanting friend classes though.
<nirvdrum>
rtyler: So, you get to work on JRuby full-time now?
<rtyler>
hah,no
<rtyler>
nirvdrum: why do you ask?
<headius>
enebo: oh, was it the one where you don't see *any* build output?
<headius>
like, it goes into the maven build line and sits there
<mkristian>
eregon, yes and no. I honestly do not remember the details. it had something to do with eclipse incremental compile or so. if it bothers you just remove it
<mkristian>
my ec
<nirvdrum>
rtyler: I guess I misunderstood a tweet headius made about sponsorship.
<rtyler>
nirvdrum: mkristian works with me on Core Systems at Lookout
<enebo>
headius: It might have made it further but some tests grab those app servers and what not
<rtyler>
I'm not smart enough to contribute to JRuby
<rtyler>
working up to it xD
<mkristian>
my eclipse setup is not really compatible with all those files coming from github
<headius>
enebo: yeah, there's a lot o fnetwork traffic that could fail on travis
<nirvdrum>
rtyler: Ahh. Well, thanks for the clarification.
<nirvdrum>
You seem plenty smart though.
<headius>
mkristian, nirvdrum: I thought it was something to do with getting Eclipse to recognize our build-time-generated code
<headius>
oops
<headius>
mkristian and eregon, not nirvdrum
<headius>
I have not used Eclipse in years though
<eregon>
mmh, does anyone use Eclipse to develop JRuby atm?
calavera has joined #jruby
<rtyler>
nirvdrum: my time is spent largely elsewhere nowadays for better or worse, but that does give me the ability to hire awesome hackers like mkristian to help projects like JRuby
<eregon>
mostly we'd like to have it as we have some good IDE support there for the Truffle anno processor, and ideally also incremental build, which could make the feedback cycle much faster
<rtyler>
one of these days I'd love to be able to contribute mnore to jruby though, jruby-gradle is as close as I can get
<rtyler>
(right now)
<nirvdrum>
rtyler: Cool. Well thanks. The efforts are appreciated.
<mkristian>
eregon, what I do with eclipse I set the eclipse classes/build directory to target/class-eclipse so eclipse does not interfere with my maven build
<rtyler>
nirvdrum: I'm really excited that we can do this, I've been trying to get funding for initiatives like this for years
<mkristian>
that works really OK
rsim1 has quit [Quit: Leaving.]
<eregon>
mkristian: OK, do you think having both working on the same generated dir is never going to work?
<eregon>
mkristian: the same class dir*
jumex has joined #jruby
<mkristian>
eregon, not with me. eclipse start to recompile things when maven runs and then they both compile the same classes at the same time :(
<mkristian>
eregon, that cause more pain then gain with me
<eregon>
mkristian: mmh, right. So you actually develop under Eclipse most of the time?
<mkristian>
I would give something with those core/.classpath core/.settings and core/.project would not be on github
<mkristian>
yes I am using eclipse for jruby
<eregon>
yeah, I thought no one was using Eclipse anymore
<mkristian>
old habit - never used anything else ;)
<mkristian>
beside emacs
<eregon>
I am not sure what's the good way to share or not the config files, to still have easy import for new users
<eregon>
but you use the Maven plugin, right? So builds are not really incremental?
<mkristian>
convert the project into a maven project is what I do
<mkristian>
yes, I convert the project to a maven project. it looks very incremental to me.
<eregon>
mkristian: does it still generate the big jruby jar at every compilation?
<mkristian>
eregon, no this I do with maven.
<eregon>
mkristian: sounds like we want sth very similar :) What would be the easiest way for me to try your setup?
<mkristian>
eregon, delete core/.classpath core/.settings and core/.project, create a project from existing project (or how they call it), convert to maven
<eregon>
OK, which maven plugin do you install?
<mkristian>
maven support comes with eclipse since a few versions now
<mkristian>
I am using Eclipse Luna right now
<mkristian>
then I did set the output folder to target/classes-eclipse - but just saw it is not set right now. so it might work OK even without it.
<headius>
enebo: didn't you say crypt is platform-specific? I don't think this spec can be made to pass everywhere unless it uses a known mechanism for generating the crypt data to verify against
<headius>
e.g. a shell-out to some command or something
<enebo>
headius: though crypt specs should be tagged out and really they should just be removed
<enebo>
headius: I don’t know why they are running though
<enebo>
headius: I thought I tagged them
<enebo>
oh merged from 1.7
<enebo>
headius: I will tag these out on master. fwiw, I don’t know why this behavior is being spec’d n the first place. It is just verifying how crypt works itself and does not have much to do with Ruby. Plus it is clearly something you should never do based on how inconsistent it behaves
<headius>
agreed
<headius>
enebo: evalScope stuff may expose a bug in parsing
<enebo>
NOOOO
<enebo>
:)
<headius>
the logic that does the parse immediately before eval uses DynamicScope.getEvalScope to get a DynamicScope to hand off to the parser
<headius>
but when executed, we create a new scope
<headius>
this "bug" is that if I do eval "a = 1"; eval "b = 1", the second eval's "b" is at index 1 instead of index 0
<headius>
I'm still seeing "b" at index 1 in jruby -Xir.compiler.debug=true -X-C -e "def foo; eval 'a = 1'; eval 'b = 2'; end; foo"
<enebo>
I just realized Block and LocalStaticScope was cooalesced to a single class
<headius>
yeah
<headius>
there's only one true StaticScope now
<enebo>
yeah I liked the three class design more but it is ok :)
<headius>
I'm confused why this isn't given b a 0 offset
<enebo>
headius: does Binding come into play at all. There is logic there looking for a cache too
<headius>
it should be a new scope for every parse
<headius>
hmm
<headius>
this is the no-binding path though
<headius>
Interpreter.evalSimple
brettporter has joined #jruby
<enebo>
yeah so in this case that logic should just plain be removed from DynamicScope
<enebo>
just put your factory call in evalSimple
<headius>
I'll try that
<enebo>
Although it will make no difference but since there is no special relationship any more it makes more sense
<eregon>
mkristian: tried a couple things but still seem to not work so well, maybe you could send me your eclipse .prject/.classpath/.settings to see what I did wrong?
<headius>
enebo: small change I'm falling into binding path too
<headius>
which was still using cache logic
<enebo>
aha
<eregon>
mkristian: seems only truffle factories are not generated, did you need any special treatment for them?
<headius>
bleh...didn't fix it, still putting b at 0:1
<headius>
I don't want to step through this :-(
brettporter has quit [Ping timeout: 272 seconds]
<dfr|work>
mornign all
triple_b has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<headius>
dfr|work: hello there!
<dfr|work>
headius, for the protobufs stuff: besides just adding supoport for jruby, I want to see whether I can migrate my current project to use it ;)
<dfr|work>
'cause I feel that they don't really have any ruby customers at all yet, so not sure whether their API makes sense ;)
<dfr|work>
but the fact that protbufs are finally being opensourced is pretty exciting
<dfr|work>
:D
<dfr|work>
from the simple look at it, seems like the java part is exactly what I use within Google, so it's pretty close
nateberkopec has quit [Quit: Leaving...]
<enebo>
headius: oh but we do store %block at 0
<enebo>
headius: I think we do now anyways but hmm that is weird
<enebo>
headius: you are doing eval a =1 ; eval b = 1;
<enebo>
<c(0:2)> = Fixnum:1
<enebo>
hmm
<enebo>
of course I have not changed anything so this is correct with no changes
<enebo>
headius: You sure you are recompiling? :P
nateberkopec has joined #jruby
<dfr|work>
headius, there's even RPC stuff :D
nateberkopec has quit [Client Quit]
<enebo>
dfr|work: does it use byte[] or char[]?
<dfr|work>
enebo, ByteString, I think
<dfr|work>
lemme link
phrinx has joined #jruby
multibot_ has quit [Read error: Connection reset by peer]
<dfr|work>
I think the only thing is missing is the RPC server implementation that google uses, but that's probably okay, because it's pretty Google specific =/
<headius>
enebo: pretty sure
<enebo>
headius: ok now I am looking at this in a debugger
<headius>
I'm going to confirm normal eval goes through these paths...something's fishy
colinsurprenant has joined #jruby
<enebo>
headius: neight evalSimple not evalWithBinding are ever called
<enebo>
in Interpreter :)
<headius>
yeah I noticed that
phrinx has quit [Ping timeout: 244 seconds]
<enebo>
My guess is this somehow must go through a block dispatch path
<headius>
well wait
<headius>
evalWithBinding appears to be used from evalCommon
<enebo>
HAHA…This is baffling
<mkristian>
eregon, yes, I can send you my project files. but not sure when those truffle factories are missing. cd core;mvn package should do everything right
<headius>
I'm thinking we should make travis just run openjdk8 so we don't have so many backed-up builds
<headius>
we can have the rest run in 7 on 'bees
<headius>
travis needs to be faster turnaround to be useful for us
<nirvdrum>
Does it build on every commit or every push?
drbobbeaty_ has joined #jruby
drbobbeaty has quit [Read error: Connection reset by peer]
<headius>
every push
<headius>
but I think most of us have a tendency to push every commit
<headius>
jenkins, on the other hand, polls...so it will only run a maximum of N builds in a given day that each aggregate a few commits
<headius>
I love travis but it's not useful when we wait over an hour to get results
<headius>
plus it gets massively backed up
<headius>
gonna give it a try
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] headius pushed 1 new commit to master: http://git.io/7iz9Hg
<JRubyGithub>
jruby/master 8ddeaaf Charles Oliver Nutter: Try speeding up travis by only running one JDK, one platform.
JRubyGithub has left #jruby [#jruby]
tenderlove has quit [Ping timeout: 250 seconds]
<headius>
mkristian: do you have any objections to moving the other -P test runs to 'bees and not running on travis?
<headius>
-Posgi, -Pcomplete, etc
tenderlove has joined #jruby
<headius>
I really want to get travis runs down to a reasonable length
<nirvdrum>
headius: Okay. I knew it was push triggered, but I wasn't sure if N commits in a single push meant N builds.
<headius>
ahh, no, just one
colinsurprenant has joined #jruby
<Aethenelle>
i'd leave the other jdks and platforms just make them optional
<headius>
that's not apparent in travis's UI
<headius>
Aethenelle: optional?
<Aethenelle>
allowed to fail
<headius>
oh
<headius>
that wouldn't keep us from backing up on them though
<nirvdrum>
That'd still back the builds up though.
<headius>
we need total travis time to be much lower than it is now
<Aethenelle>
hrm... true...
e_dub has quit [Quit: e_dub]
colinsurprenant has quit [Client Quit]
<headius>
it doesn't help that the travis instances are dog slow, of course
<headius>
vm within a vm
<Aethenelle>
not that it takes long but ffi can move
<nirvdrum>
Of course, if it only contains a subset of total builds, it might be worth considering if it's worth doing at all.
<Aethenelle>
-Pmain probably too (dups -Pmain,test -Dinvoker.test=extended ?)
mkristian has quit [Ping timeout: 264 seconds]
<Aethenelle>
i'd prbably also move the tests that are allowed to fail, -Pcomplete, spec:jrubyc, and spec:profiler
colinsurprenant has joined #jruby
<Aethenelle>
maybe fast finish the rest and dup them on 'bees without the fast fail?
<headius>
nirvdrum: well, I don't think that's true...do you run all these suites locally every commit?
<headius>
there's value in having a fast CI even if it doesn't run the extensive suite
<nirvdrum>
headius: No, but that's the point of CI.
<headius>
right, we'd still do them in a cloudbees
<Aethenelle>
nirvdrum: that's why we also have cloudbees
<headius>
but people look at travis to see build status, and we're pretty much always behind by like 5-10 builds
<Aethenelle>
do you know at all which test suites might be good canaries?
<headius>
if it ran faster or we got more concurrent jobs, I'd be fine running everything
<nirvdrum>
Aethenelle: I guess my point is if there's one source of total information and one of partial, looking at the total maybe makes sense. But CloudBees is pretty slow too, so maybe that's a moot point.
<headius>
we used to only run the comprehensive suite nightly
<headius>
Aethenelle: at least MRI and RubySpec
<Aethenelle>
nirvdrum: the idea is to speed up the CI non-jruby devs will look at
<Aethenelle>
headius: regression and jruby too
<headius>
yeah
<headius>
those four would catch the vast majority of failures
<headius>
maybe spec:ji
<headius>
nirvdrum: the other issue is that it takes so long for a build to fail, the person who committed it has gone home for the day :-)
<headius>
so someone else has to mop up, or we sit all night with a broken build
<headius>
that's counterproductive
<nirvdrum>
Clearly I'm on the losing end of this discussion. I don't care a whole lot.
<headius>
there's no losing end...I won't remove jobs from travis unless we add them to 'bees
<headius>
we just need one or the other to give results fast
<headius>
either that or go back to running our own CI :-)
<nirvdrum>
Fair enough. We just seem to keep making concessions to get faster Travis builds.
<Aethenelle>
headius: i think it might be a good idea to include all tests on cloudbees not just the transfered ones...
<headius>
nirvdrum: travis's fault, really
<nirvdrum>
Is there a way to buy a faster machine and maybe get a sponsorship?
<headius>
unfortunately not
<headius>
everyone gets the same, even paid users
<headius>
another reason bees could be more flexible
<nirvdrum>
A $5 DigitalOcean droplet would probably cut the time in half :-/
<Aethenelle>
perhaps something could be arranged with travis... they rely on jruby for testing, so it makes sense they might be willing to up the power or something...
<headius>
Aethenelle: yeah maybe
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] headius pushed 1 new commit to jruby-1_7: http://git.io/azJ_Uw
<JRubyGithub>
jruby/jruby-1_7 1724893 Charles Oliver Nutter: Reduce travis build times by only running on one JDK version.
JRubyGithub has left #jruby [#jruby]
<headius>
asarih: is there any way to get faster instances?
<headius>
$$$$
<asarih>
unfortunately not.
<asarih>
it's a one-size-fits-most for the time being.
<asarih>
I *think* you can get dedicated queue, though, if RHT is willing to pay.
<headius>
too bad :-\ we're having to make concessions to get travis runs to short enough
<headius>
asarih: hmmm that's interesting...would that help overall job time?
<asarih>
headius: hang on. let me try something…
<headius>
I'm wondering if we should reinstate the aggregate build of all test: targets...we split them out so we wouldn't hold an instance for so long, but now we have N builds all taking >10m
<headius>
trying to figure out how to best utilize travis
mkristian has joined #jruby
<headius>
asarih: earlier today I mused that some days I just sit and watch builds turn green...I could revise that and say some days I sit and cancel builds the whole day so that stale commits aren't slowing down CI status
<headius>
our builds are just too big to run on every push during a busy day
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] BanzaiMan created ha-feature-sudo-false (+1 new commit): http://git.io/8IrKrQ
<JRubyGithub>
jruby/ha-feature-sudo-false 51f9786 Hiro Asari: Use conatiner-based infrastructure
JRubyGithub has left #jruby [#jruby]
<asarih>
understood.
<Aethenelle>
it'd probably also help if more use was made of the skip ci magic comment
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] BanzaiMan pushed 1 new commit to ha-feature-sudo-false: http://git.io/yX3Kfg
<JRubyGithub>
jruby/ha-feature-sudo-false e15f77e Hiro Asari: Build this branch!
<asarih>
by caching ~/.m2, I think we can cut down on time more.
<headius>
switching JDK requires sudo because of update_alternatives I am guessin?
<headius>
if they're already installed we could just ENV it too
<headius>
caching .m2 would be big
<asarih>
headius: yes.
<asarih>
the jobs calling for Oracle JDK 8 in that build would fail
<enebo>
Personally I would rather test on 7 than 8
<enebo>
At least for fast turnaround to point out someone is usnig an 8 API
<headius>
yeah
<headius>
test for lowest supported version
<headius>
only things that change in higher JDKs are going to be one-time failures because of some new API conflict
<headius>
rare
<asarih>
wheeee… for .1, the build time went from 18+ minutes to 11+.
<headius>
that's much better
<asarih>
~12.
<enebo>
I sort of like nirvdrum worry about splitting ci across two systems because inevitably we will start ignoring cloudbees … but I guess with some discipline we can make that work
<enebo>
sounds like a big improvement to me
<headius>
enebo: I think having fast feedback is more important, so long as we're not leaving huge testing gaps
<nirvdrum>
enebo: I got five words into that and was really excited. Then deflated.
anaeem1 has joined #jruby
<enebo>
I have also wondered if we could split some test targets across process-launching/resource ones and !proceass-...
anaeem1 has quit [Remote host closed the connection]
<nirvdrum>
enebo: I wouldn't underestimate the cost of file system access in an environment like this.
<enebo>
nirvdrum: It would be cool if testing software could emit time taken stats as it goes too
<nirvdrum>
Aethenelle: I don't think they do the badge thing. But our builds haven't been consistent for a while, making the badges markedly less useful.
<enebo>
which I guess we could do if we timestamped output and made each test produce output
<Aethenelle>
nirvdrum: I was thinking the added badge would help prevent ignoring cloudbees
<nirvdrum>
Ahh. I was thinking you meant the PR auto-test notification thingie.
<JRubyGithub>
[jruby] kares opened pull request #2311: changes that address tests with current jruby-openssl master (jruby-1_7...update-openssl-1_7) http://git.io/CvOJlg
JRubyGithub has left #jruby [#jruby]
skade has quit [Quit: Computer has gone to sleep.]
yfeldblu_ has quit [Remote host closed the connection]
mkristian has quit [Ping timeout: 260 seconds]
jgdavey has joined #jruby
e_dub has joined #jruby
rsim1 has joined #jruby
<jgdavey>
What's the best way to embed a Ruby web server in a java process? I'd like to spin up a middleman server programmatically for local development in an existing JVM-based project.
rsim has quit [Ping timeout: 265 seconds]
<dfr|work>
jgdavey, ruby web server?
<dfr|work>
jgdavey, you mean so that ruby handles all the low level socket stuff?
<mjc_>
it sounds like he wants to just boot webrick or something inside an existing jvm
<dfr|work>
mjc_, yea, that sounds scary. :)
triple_b has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<dfr|work>
although I'm sure you could do: Ruby.newTime().evalScriptlet("require 'webrick'; WEBrick.start 80"); or whatever way you start webrick >.<
<dfr|work>
*Ruby.newruntime
mkristian has joined #jruby
<jgdavey>
My exact use case: I'd like to spin up a middleman server, which uses webrick
<jgdavey>
I tried Ruby.getDefaultInstance().evalScriptlet("require 'middleman'; require 'middleman-core/preview_server'; Middleman::PreviewServer.start")
<jgdavey>
It does bind the port, but nothing comes out
<jgdavey>
I also tried doing the evalScriptlet within a Thread
<dfr|work>
jgdavey, what does "nothing comes out" mean?
<dfr|work>
jgdavey, the server doesn't respond on the default port?
<jgdavey>
TCP port is bound and I can connect to the port, but HTTP requests just hang
<jgdavey>
I'm sure there's a more idiomatic way to do this, but was confused by the hanging.
<dfr|work>
jgdavey, maybe your server is messed up?
<jgdavey>
I'll try it from irb instead.
<jgdavey>
If the same "scriptlet" hangs there, then it's definitely server setup stuff
<dfr|work>
jgdavey, if you do: ruby -e "require 'middleman'; require 'middleman-core/preview_server'; Middleman::PreviewServer.start") does it work fine?
<jgdavey>
Trying now
<jgdavey>
Yep, that works.
<dfr|work>
jgdavey, I'd maybe try Ruby.newRuntime rather than default, so that you get a brand new runtime. But otherwise not sure.
<dfr|work>
if there's some initialization going on, maybe add puts or whatever statements to see what's going on.
<jgdavey>
It's working now!
<jgdavey>
It seems like an error was being swallowed.
<jgdavey>
When I did evalScriptlet, I had a missing require, which raised undefined constant.
triple_b has joined #jruby
<jgdavey>
Also, using newInstance rather than getDefaultInstance is working much better
<jgdavey>
dfr|work, thanks for the help!
<dfr|work>
jgdavey, yw. But for the record, spawning stuff that listens to sockets from Java still seems very ugly. :)
anaeem1_ has quit [Remote host closed the connection]
anaeem1_ has joined #jruby
<asarih>
the downside of this setup is that we can't spin up a debug instance yet.
<nirvdrum>
lopex: I'm just trying to figure out how encodingIndex in EncodingService isn't a problem.
<nirvdrum>
Multiple Encodings can map back to different RubyEncodings.
<nirvdrum>
Sorry. The same Encoding can map to multiple RubyEncodings.
<byteit101>
currently using mkristian's jbundler and mvn tools on a jruby project, I get "jbundler support needs jruby to create a local config: jruby -S jbundle install"
<byteit101>
I've not used maven much, and am unsure what the error means/ how to fix it
<lopex>
nirvdrum: afair, those are just for faster lookups
<byteit101>
that line is thrown in the rakefile when I do `wt = Warbler::Task.new`
<byteit101>
anyone know what i'm missing?
<lopex>
nirvdrum: I'd have to recall the details
<nirvdrum>
lopex: It seems to be used to implement Encoding.compatible? I'm just not sure how they're not colliding.
<mkristian>
byteit101, did you run $jbundle install ?
<byteit101>
this is from a polyglot-mvn'ed pom file through normal maven
<byteit101>
using pmvn works on systems with jruby installed
rcvalle has quit [Quit: rcvalle]
<byteit101>
using normal mvn on systems with jruby does not work
<mkristian>
byteit101, on first glance it all looks ok. let me see more . .
<byteit101>
mkristian: also is there actual documentation that is up to date with mavenfiles? I could only find old information and had to figure it out from existing mavnfiles
<mkristian>
can you run the rake tasks without maven ?
<byteit101>
yep
<lopex>
nirvdrum: wow, how can bugs live for so long
<byteit101>
(assuming jruby is installed)
kedibasi has quit [Remote host closed the connection]
<nirvdrum>
lopex: This is probably an atypical case. But I ran into it because I was trying to use EncodingService#getEncodingFromString, which returns an Encoding. Then I was trying to use that index. It caused one of the rubyspecs to fail.
<byteit101>
mkristian: it builds fine using rake and mvn (calling rake) on systems with jruby installed, but fails when its only using jruby inside the mvn call
<mkristian>
byteit101, that is strange. could gist me the output of mvn -Djruby.verbose
<mkristian>
is there a .jbundler/classpath.rb in your directory ?
<byteit101>
no
<byteit101>
should I commit that file from my dev system?
<mkristian>
so warbler needs this file to warble the war ;)
subbu has joined #jruby
<byteit101>
ok. is that safe to commit or should I add some step to building to generate that?
<mkristian>
I think it would be better to get jbundler to build it.
<byteit101>
ok
<mkristian>
it is not safe to commit since it contains absolute pathes
<byteit101>
mine donesn't, but ok. how can I have jbunder build it? should I do that in the mavenfile or rakefile?
drbobbeaty_ has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<mkristian>
yeh - I tried not to, so it might be safe for you.
<mkristian>
if you run proper maven you should to it in the Rakefile. jbundler-0.7.0 has JBundler.install
purplefox has quit [Ping timeout: 264 seconds]
rsim1 has quit [Quit: Leaving.]
<byteit101>
mkristian: is it safe to do this to avoid rake always re-bundling? `JBundler.install unless File.exist? ".jbundler/classpath.rb"`
<byteit101>
(aside from when people update that file, of course)
<mkristian>
that is fine.
<mkristian>
I think that should better go into warbler itself instead of this warning
<byteit101>
yea, that would benice
<mkristian>
need to go now - hope that all fixes your problem
<byteit101>
mkristian: ok, seems like it
<byteit101>
thank you!
<mkristian>
welcome . . .
mkristian has quit [Quit: bye]
blaines has joined #jruby
diegoviola has quit [Ping timeout: 244 seconds]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton pushed 1 new commit to master: http://git.io/ZeiiZg
<JRubyGithub>
jruby/master 2df1533 Chris Seaton: [Truffle] Last of the def-2.0 specs.
JRubyGithub has left #jruby [#jruby]
diegoviola has joined #jruby
lanceball is now known as lance|afk
e_dub has quit [Quit: e_dub]
brasten has quit [Quit: brasten]
subbu has quit [Ping timeout: 250 seconds]
triple_b has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton pushed 1 new commit to master: http://git.io/HquhFA
<JRubyGithub>
jruby/master 9668c24 Chris Seaton: [Truffle] Fix regexp errors.
JRubyGithub has left #jruby [#jruby]
bbrowning_away is now known as bbrowning
rsim has joined #jruby
<chrisseaton>
Lots of Truffle 'last of the ... specs' commits going by in the last week!
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] BanzaiMan opened issue #2313: Consider dropping OracleJDK 8 for now, and move builds to container-based Travis CI http://git.io/JNz5hQ
JRubyGithub has left #jruby [#jruby]
bbrowning is now known as bbrowning_away
enebo has quit [Quit: enebo]
brettporter has joined #jruby
brettporter has joined #jruby
brettporter has quit [Ping timeout: 260 seconds]
mitchellhenke has quit [Quit: Computer has gone to sleep.]
skade has quit [Quit: Computer has gone to sleep.]
<lopex>
chrisseaton: this must be something new in onigmo, the a modifier, for example
<chrisseaton>
lopex: do you want a bug open for this? at the moment it's a failing spec, but unless you went to look there you wouldn't think to work on it
<chrisseaton>
egp: what exactly do you want to do? evaluate little bits of Ruby code, use classes written in Ruby, or something higher level than that?
<egp>
chrisseaton: I want to use this gem: https://github.com/staqapp/staq-msbin, and really be able to call MSBIN::Record.DecodeStream(is) where is is an inputstream or whatever I need to do
<egp>
in such a way I can just do a rake build/copy that gem somewhere, and have my ScriptExecution or whatever code setup in a nice way
<egp>
and just have the gem in my classpath or some such
mitchellhenke has quit [Quit: Computer has gone to sleep.]
<chrisseaton>
egp: this code looks relatively well self-contained - I would recommend trying the JRuby AOT compiler, which produces normal Java classes - https://github.com/jruby/jruby/wiki/JRubyCompiler
<egp>
that could be neat
<egp>
can I actually do the library
<egp>
or do I effectively need to create a script
diegovio1 is now known as diegoviola
<chrisseaton>
I'm really not sure beyond pointing you at that wiki page, sorry
<chrisseaton>
hang around on here and someone else might chime in