<subbu>
headius, what is the best way to add these save/restore tests ..?
<headius>
hmm
<subbu>
regression specs? rubyspec? mri tests?
<headius>
I'd add to compiler
<headius>
spec/compiler
<headius>
that will test both interp and jit
<subbu>
these were core IR generation bugs and probably also expectations about how $! save/restore should behave .. so spec:compiler seems the wrong place.
<headius>
well...IR compiler :-)
<headius>
adding to rubyspec would be good too though
<headius>
under language there's spec for $! and rescue/ensure
<subbu>
ok .. will take a look at exists there currently.
<subbu>
thanks.
<subbu>
so, i suppose i should clone rubyspec for that?
<headius>
no, you can just commit to our local copy
<headius>
eregon has magic juju to merge from there back to ruby/rubyspec
<subbu>
ok, even better. :)
lance|afk has quit [Ping timeout: 264 seconds]
lanceball has joined #jruby
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] headius pushed 1 new commit to jruby-1_7: http://git.io/vstgK
<JRubyGithub>
jruby/jruby-1_7 6992f8e Charles Oliver Nutter: Don't use deprecated Thread.stop anymore. Fixes #3232.
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] headius pushed 1 new commit to master: http://git.io/vstgX
<JRubyGithub>
jruby/master caa5904 Charles Oliver Nutter: Merge remote-tracking branch 'origin/jruby-1_7'
JRubyGithub has left #jruby [#jruby]
cpuguy83 has quit [Ping timeout: 256 seconds]
mdedetrich has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<JRubyGithub>
[jruby] benlovell opened pull request #3265: WIP: Updates for RSpec 3 (master...rspec-3) http://git.io/vsqFv
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] pinya opened issue #3266: "NotImplementedError: waitpid unsupported or native support failed to load" with 9.0.0.0 during install jruby-launcher gem http://git.io/vsqFn
JRubyGithub has left #jruby [#jruby]
<benlovell>
hey folks - looking for a little guidance on #3265 before I steam in
digitalextremist has quit [Read error: Connection reset by peer]
rsim has quit [Quit: Leaving.]
samphippen has quit [Read error: Connection reset by peer]
<mkristian__>
benlovell, yes, pom.rb is thing to modify
<benlovell>
mkristian__: thanks, I gave it a go and there is a build pending
<benlovell>
i wonder if this is going to bust the specs elsewhere though
<mkristian__>
after you ran the build once the pom.xml will change as well. there are just there to be used by IDEs since they do not understand the pom.rb format (yet or never - who knows)
<benlovell>
cool
cristianrasch has joined #jruby
<benlovell>
ok that seemed to work, lots of failing specs but that's the point :)
mdedetrich has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
mkristian__ has quit [Quit: This computer has gone to sleep]
DomKM has quit [Quit: Connection closed for inactivity]
samphippen has joined #jruby
mkristian__ has joined #jruby
mkristian__ has quit [Quit: This computer has gone to sleep]
bbrowning_away is now known as bbrowning
mdedetrich has joined #jruby
nirvdrum has joined #jruby
<olleolleolle>
benlovell: did you ever use transpec?
<kares>
headius: subbu: but eregon hopes to have the changes under there in separate commits
<enebo>
kares: thinking of release 1.7.22 today anything worry you about that?
benlovell has joined #jruby
<benlovell>
olleolleolle: indeed, once I realised the magnitude of my task :D
<benlovell>
almost done now
<enebo>
benlovell: thanks a million for doing this….Every time I install rspec 3 for a project and then blow up locally because I am running rspec 3 on our specs ...
<benlovell>
enebo: my pleasure!
<enebo>
benlovell: well I am glad someone gets pleasure from modifying spec syntax :)
<kares>
enebo: go for it ... my most wanted fixes should be in :)
<enebo>
kares: great
<benlovell>
:)
<kares>
there's one issue thought with jruby-openssl :(
<enebo>
kares: not a regression though right? just something you wish was fixed?
<kares>
there's a new release but tests seem to block on jruby-1_7 ... works fine on 9K
<kares>
but due that its not updated
<kares>
enebo: yeah I wished someone would look into that and fix it :)
<headius>
kares: where is it getting stuck?
<kares>
barely runs any tests/spec
<kares>
let me get you the link
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] eregon pushed 3 new commits to master: http://git.io/vsYoa
<kares>
wasn't gonna release if I knew its that bad - the problem is with the pre-release almost everything failed thus I went on with a release
<kares>
headius: not sure
<kares>
its been a few days :)
<headius>
I'll give it a try here
subbu has joined #jruby
<enebo>
so does this mean if people see a new openssl release and gem update their 1.7.x install will be hosed?
<kares>
yes - it might mean that ... sorry
<kares>
should we yank the release on rubygems?
<kares>
the thing is 0.9.7 had a regression or two as well
<kares>
this was supposed to be the fix ... but new issues appeared in the last moment
<enebo>
well if we can figure this out and put out another version then we would yank. If we cannot figure it out maybe we also yank if 0.9.7 works better than hanging
<kares>
spent almost a week getting this out ;( than I got too frustrated
<kares>
this is 0.9.8 we're talking about
yorickpeterse has joined #jruby
<yorickpeterse>
headius: this is probably easier than Twitter
<kares>
yanking 0.9.7 as well ... it's part of several releases already ?
<kares>
jruby releases
<enebo>
kares: so you are saying both 0.9.8 and 0.9.7 have serious issues with 1.7.x?
<yorickpeterse>
headius: prior to my patches that benchmark absolutely destroyed MRI and Rbx to a certain extend
<headius>
yeah that's what I was playing with last night... jruby was like 8x MRI and 5x rbx
<headius>
pretty sure it was after your commit
<headius>
I had to bump up the inner loop for anything to get hot but then it ran great
<headius>
2 vs 4 threads, only a slight drop (this machine has 4 real cores)
<yorickpeterse>
I added a config option for that now
<yorickpeterse>
since I had to bump it up as well for rbx after my patch
donV has joined #jruby
<headius>
above 4 things drop off but that's what I'd expect
<yorickpeterse>
I still haven't been able to prove it accurately, but I suspected the GC was hammered so badly it would basically pause everything all the time
<yorickpeterse>
Though I'm still confused how Array#include? calls can cause that (= basically what I removed)
<yorickpeterse>
GC pressure is the only thing I can think of that would increase as threads/workloads go up
<headius>
it does do a lot of allocation... -J-XX:+PrintGCDetails showed a few full GCs and several young GCs
<kares>
enebo: no serious not - a cert NPE parsing regression in 0.9.7 and I'm not sure what's the deal with 0.9.8 (seem to work well with master)
<headius>
I used concurrent collector to remove some pause variability
<yorickpeterse>
the default collector is not concurrent?
<enebo>
kares: then I think perhaps we can look at 0.9.8 and if we can figure it out then we will make a 0.9.9 but worst case we could yank 0.9.8 and make a 0.9.9 with only the cert NPE fix too?
<headius>
the default collector is parallel STW, uses as many threads as you have cores to pause for a shorter time
<yorickpeterse>
Ah
<headius>
concurrent collector has some throughput reduction so it's not default
<headius>
young gen is always parallel STW
<headius>
-J-XX:+UseConcMarkSweepGC
<headius>
"CMS" collector
<kares>
enebo: sounds right, how about 0.9.7.1 ? but let me check the changelog
<yorickpeterse>
I suspect that things will improve when I upgrade to Java 8 and invokedynamic, but doing that on Amazon Linux is going to be fun
<yorickpeterse>
though I think it by now should have at least java 8
<headius>
I would hope so... Java 7 is officially EOL as a supported release
<enebo>
hmmm Array#include calls == and I wonder if you are hitting coerce in cases
<enebo>
I was just examining this for the 1 / 1.0 case
<enebo>
which constructs an array [1.0, 1.0]
<headius>
CMS would be good to try if you have a big old gen heap
<headius>
yorickpeterse: if it's memory bottleneck we should be able to see that by watching GC evacuation rates
<enebo>
if 2-4x the garbage fills young gen and you end up performing full gcs it may feel that way
<kares>
although that might be causing the jruby-1_7 failures for whatever reason
donV has quit [Quit: donV]
<enebo>
kares: so this is 2 of 9 commits
<kares>
headius: great, my machine is also kind of dying on me so I'm not really able to do decent work :(
dinfuehr has joined #jruby
<enebo>
kares: Seems like we can maybe bisect and figure out the hang
<yorickpeterse>
headius: Yeah, it didn't occur to me until late last night it would be fairly easy to prove/disprove it
<yorickpeterse>
by just tracking GC runs
<yorickpeterse>
but I'll try that this evening probably
<enebo>
kares: but I am ok delaying a day to see if we can figure this out
<headius>
yeah that PrintGCDetails should help... it will show <before> => <after> for each part of the heap
<enebo>
yorickpeterse: if you run with —profile on jruby I would be interested if you see any #coerce happening
iloveitaly has joined #jruby
<kares>
enebo: ok as noted I have some sort of issues with my laptop going slow - so I'm not sure I'll be helpful
<headius>
for alloc profile you can use visualvm or command line -J-Xrunhprof:depth=20
<headius>
the latter is a full allocation profile with backtraces of depth 20, so it will be very slow...but it doesn't have to run long, just long enough to see what your counts are
jeremyevans has joined #jruby
<subbu>
headius, not sure if you saw my msg y'day night, but I get: fatal: repository 'https://github.com/jruby/rubyspec.git/' not found .. when i tried to push from jruby/rubyspec/rubyspec
<headius>
subbu: you don't need to push anywhere special, just edit directly in spec/ruby and commit/push as normal
<headius>
it's set up as a subtree so eregon can pull commits out for ruby/rubyspec
skade has joined #jruby
dinfuehr has quit [Ping timeout: 260 seconds]
<subbu>
oh, i edited in rubyspec/rubyspec .. not in spec/.. let me look.
<headius>
yeah, just spec/ruby in our repo and normal commit
<yorickpeterse>
enebo: the script has an option for that :P
slash_nick is now known as rthbound
rthbound is now known as slash_nick
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] subbuss pushed 1 new commit to master: http://git.io/vsYQf
<JRubyGithub>
jruby/master 51af4f8 Subramanya Sastry: rubyspec: Added a few new specs for save/restore of $!
<headius>
yeah appears to get stuck on both 7 and 8
<kares>
key thanks, trying to get it to run locally as well ... seems I tried and got stuck
<headius>
test_pair.rb:23
<headius>
oops test_pair.rb:232
<headius>
not surprising, it's a select call
<headius>
do these tests still exist in MRI? I thought we figured out that select doesn't work reliably with SSLSocket
<headius>
buffers get drained early for decrypt, etc
<headius>
yeah, I guess they're still there
<headius>
kares: the thread that tests starts is bombing out with EOFError
<kares>
not sure - probably not since 9K is fine
<headius>
so it never tries to connect
<headius>
or rather it never accepts the connection
<headius>
that's the root cause
<headius>
oh wait...that's from the ensure block...bleah
<kares>
more like it!
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] subbuss pushed 1 new commit to master: http://git.io/vsYbc
JRubyGithub has left #jruby [#jruby]
<JRubyGithub>
jruby/master bbc88e5 Subramanya Sastry: s/RESTORE_PERLY_EXC/RESTORE_EXCEPTION_VAR/g and friends
<subbu>
headius, enebo ^
<headius>
#<OpenSSL::SSL::SSLError: (OpenSSL::SSL::SSLErrorWaitReadable) read would block>
<headius>
kares: that's what happens in the thread
<headius>
something screwy with the WaitReadable include into that exception?
<headius>
it should have been rescued
shellac has joined #jruby
bbrowning is now known as bbrowning_away
<enebo>
subbu: thanks! :)
<kares>
headius: I'll check but I do not recall any changes there
<kares>
locally I'm still getting a JVM SIGSEGV ;(
<headius>
I just added th.abort_on_exception = true to get it to bubble out, then watched for it in the ensure
<headius>
still annoys me that threads silently die by default
<headius>
kares: it looks like WaitReadable is not being included into that error anymore
<headius>
or it's getting double-wrapped somewhere
skade has quit [Ping timeout: 255 seconds]
digitalextremist has joined #jruby
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] pinya closed issue #3266: "NotImplementedError: waitpid unsupported or native support failed to load" with 9.0.0.0 during install jruby-launcher gem http://git.io/vsqFn
<eregon>
enebo, headius, maybe it's time to run them on Travis again? or tag only the really slow specs?
<subbu>
afk for a while .. but yes, i agree with eregon
vtunka_ has quit [Client Quit]
<enebo>
eregon: we onlty moved them there because these did not run in <50 minutes
<enebo>
eregon: which will just abort as failed on travis
<subbu>
oh .. hmm.
subbu is now known as subbu|afk
<enebo>
eregon: but yeah if we can tag enough spawning specs perhaps we can tune it under 50 minutes
<eregon>
enebo: yeah in truffle we just split between language/core and library and that seems to work out, even with our startup time
skade has quit [Ping timeout: 245 seconds]
<enebo>
eregon: yeah I think we also took some out just to make a individual travis run not drag out too
<enebo>
eregon: but I am not against getting more onto travis. we just need a happy medium…I think we moved these off before asarih helped up speed up runs too
<eregon>
subbu: it's an explicit exclude: '^' + SPEC_DIR + '/language/predefined_spec.rb', in spec/jruby.2.2.mspec
<eregon>
so seems you were just unlucky
<enebo>
eregon: we were having the stacking problem falling 20+ commits behind at the beginning of the year
<eregon>
enebo: I know a few exclude for which I removed the slow ruby_exe subprocesses so I'll clean some of them tomorrow or so
<eregon>
I could also copy the auto-tagger for slow specs from the truffle config, that's rather practical :)
vtunka has quit [Quit: Leaving]
<eregon>
but in any case doing a full run would be a good thing to do, at least time to time
mkristian__ has quit [Quit: This computer has gone to sleep]
<enebo>
yeah and I think the fact that we have the full run on a different system makes it easier to forget about it
<enebo>
but I think our failures are probably small over there…I only see it for integration when it first stopped being green
<enebo>
if I had to guess on rest it might be new specs not being tagged
<headius>
kares, enebo: master looks ok, checking 1.7 now
<mkristian>
enebo, I think I recently delete the test/pom.xml since it is not needed anywhere - at least that is what I thought. I will at the property which will dump a pom.xml when it runs this "test" module. could you trigger the build after this change ?
shellac has joined #jruby
camlow325 has quit [Remote host closed the connection]
samphipp_ has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<enebo>
sorry need to afk a bit bbiab
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] mkristian pushed 1 new commit to master: http://git.io/vsOQL
<JRubyGithub>
jruby/master 36a80fc Christian Meier: [build] dump pom.xml in ./test
JRubyGithub has left #jruby [#jruby]
camlow325 has joined #jruby
pitr-ch has quit [Ping timeout: 256 seconds]
camlow325 has quit [Read error: Connection reset by peer]
camlow325 has joined #jruby
samphippen has joined #jruby
<mkristian>
kares, headius are you releasing jruby-openssl-0.9.9 ? would be welcome since it can fix weird warbler classpath issue we ran into yesterday
<headius>
I got one failure on 1.7
<headius>
test_cipher_strings(TestIntegration):
<headius>
OpenSSL::SSL::SSLError: Received fatal alert: handshake_failure
<headius>
java 8u60
<headius>
I can try it again and see if it's intermittent
<headius>
ok it fails with 0.9.7 too, might be my JDK install
<headius>
I don't think I've tweaked it to have unlimited strength
<headius>
kares, enebo: SHIP IT
<headius>
looks ok on MRI tests on master and 1.7
<rtyler>
mkristian: the new jruby-openssl fixes that issue you and dave were working on?
<mkristian>
rtyler, not directly but we can release a new fast-rss-engine which will use the new ruby-openssl and this combination will fix that issue until we pick a ruby which comes with the new josh embedded :)
<mkristian>
so there were more than one issue yesterday
<kares>
headius: enebo releasing 0.9.9
_djbkd has joined #jruby
nirvdrum has joined #jruby
subbu|afk is now known as subbu
pitr-ch has joined #jruby
shellac has quit [Quit: Computer has gone to sleep.]
<headius>
kares: woohoo!
<headius>
enebo stepped out but if you update 1.7 to use it I'll re-run tests locally
<headius>
doesn't hurt to triple check
<kares>
enebo: its on rubygems
<headius>
neeeato
<headius>
I think I just need to update lib/pom.xml
<kares>
nexus is released as well ... but it will take a while
<kares>
headius: not sure if it will work
<kares>
until it bubbles up into central ... right mkristian ?
<kares>
(jruby-openssl)
<kares>
headius: local cd jruby-openssl & mvn install should be fine though
<headius>
ah maybe
<headius>
ok
yfeldblum has joined #jruby
skade has joined #jruby
<headius>
looks good, running tests
shellac has joined #jruby
<kares>
headius: thx again ... owe you a czech lager!
<enebo>
mkristian: so another data point is it happens on my machines on macos
<enebo>
machine
<mkristian>
I understand the linux bit. let me have a look at MacOS as well
baroquebobcat has joined #jruby
benlovell has quit [Quit: Lost terminal]
lance|afk is now known as lanceball
<bbrowning>
mkristian: I wonder if symlinks are tripping it up?
<bbrowning>
the FileInputStream it's creating in Lookup#loadDefaultJavaCACertsFile is to a symlink
<bbrowning>
in my case /etc/pki/tls/certs/ca-bundle.crt
<mkristian>
ok
benlovell has joined #jruby
skade has joined #jruby
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] bbelleville opened pull request #3268: [Truffle] Create layout as a static variable (master...static-layout) http://git.io/vs32P
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton closed pull request #3268: [Truffle] Create layout as a static variable (master...static-layout) http://git.io/vs32P
JRubyGithub has left #jruby [#jruby]
dinfuehr has joined #jruby
<enebo>
bbrowning: you think FileInputStream is not following a softlink?
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] chrisseaton pushed 2 new commits to master: http://git.io/vs3VB
<JRubyGithub>
jruby/master 29ca7db Chris Seaton: Merge branch 'master' of github.com:jruby/jruby
<JRubyGithub>
jruby/master c8b55cf Chris Seaton: [Truffle] Remove old require debug code.
JRubyGithub has left #jruby [#jruby]
<bbrowning>
enebo: I googled around a bit and couldn't find a definitive answer
<bbrowning>
but I wonder
<enebo>
bbrowning: with how crappy Java’s support for FS-related things I would be surprised if it didn’t follow
<enebo>
bbrowning: although I guess I can script it :)
<mkristian>
I just tried a symlink and it works. but the problem is that the code expect a JavaCACertsFile as default - and /etc/pki/tls/certs/ca-bundle.crt is not java keystore
<bbrowning>
ahh
<bbrowning>
mkristian: well the code tries this pki cert file first
hobodave has quit [Quit: Computer has gone to sleep.]
skade has quit [Ping timeout: 245 seconds]
hobodave has joined #jruby
djbkd has joined #jruby
<headius>
back
<mkristian>
bbrowning, before I start patching. is the /etc/pki/tls/certs/ca-bundle.crt a pem file. the cart file I found on my system is a PEM file. but I want to be sure
<bbrowning>
mkristian: yes, it is
<bbrowning>
it's symlinked to /etc/pki/ca-trust/extracted/pem/tls-ca-bundle.pem
<headius>
wazzup
<bbrowning>
we're just working through an issue in the latest jruby-openssl
badfish129 has quit [Quit: badfish129]
benlovell has joined #jruby
baroquebobcat has quit [Quit: baroquebobcat]
<headius>
I was able to install a gem with 1.7 + 0.9.9 locally
skade has joined #jruby
<bbrowning>
headius: so you must not have /etc/pki/tls/certs/ca-bundle.crt ?
<headius>
I'll check
<headius>
I have no /etc/pki at all
<bbrowning>
yeah
<headius>
OS X 10.10
<mkristian>
headius, this is redhat thingy
<mkristian>
or linux
<headius>
ahh I see
<bbrowning>
I'm on Fedora
<headius>
ok
<bbrowning>
F21
<headius>
we have a fix in progress then?
<bbrowning>
mkristian: well enebo hit it on mac
<enebo>
bbrowning: I do not either on Mavericks
<mkristian>
so, I have a patch but how can I test it.
<mkristian>
bbrowning, but on different file location
<bbrowning>
ahh right
<bbrowning>
mkristian: does jruby-openssl have CI?
<bbrowning>
run on travis or something? that would trigger it I'd think
<mkristian>
it did not for the 0.9.9 release today :(
<enebo>
I also do not hve /etc/pki so this is a different location
skade has quit [Quit: Computer has gone to sleep.]
<enebo>
on my machine which is dying
<bbrowning>
enebo: yeah you must be hitting the 2nd location it tries
<bbrowning>
enebo: /System/Library/OpenSSL/cert.pem I think
camlow325 has quit [Read error: Connection reset by peer]
<enebo>
oh /System/Lirbary/OpenSSL
<bbrowning>
mkristian: is there a test that would exercise this code path?
camlow325 has joined #jruby
<enebo>
file /System/Library/OpenSSL/cert.pem
<mkristian>
enebo, do you have a /System/Library/OpenSSL/cert.pem ?
<enebo>
mkristian: I do
<enebo>
it is ASCII text
<enebo>
-----BEGIN CERTIFICATE-----
<enebo>
...
yfeldblum has quit [Ping timeout: 246 seconds]
<bbrowning>
mkristian: and I'm happy to test locally if you can help me figure out how to get my local build of jruby-openssl inserted into a local build of jruby-1_7
<mkristian>
enebo, that is not a java keystore either.
<bbrowning>
so that's why enebo fails
<enebo>
nope but I would never think openssl would have one
<bbrowning>
headius must not have either of those
<enebo>
is java keystore format specific to java?
<mkristian>
bbrowning, I do not have such a cert.pem on my MacOS
<enebo>
mkristian: well you could just mae a bogus file
<enebo>
mkristian: then it would choke and you can see it fail and whether you patch fixes it
* enebo
is a lifehacker
<mkristian>
enebo, ok, I can manually test it. make the patch and see if I can setup jruby-1_7 build to use the staging gem
<bbrowning>
enebo: a bogus, but valid, cert.pem
shellac has quit [Quit: Computer has gone to sleep.]
<enebo>
mkristian: yeah or just try installing a gem from the command-line
<bbrowning>
enebo: because he's going to try parsing it as a pem afaik
<headius>
I do not have that pem file either
<headius>
workaround: use OS X 10.10
<enebo>
bbrowning: but I think he is choking because it is not in java keystore format right?
<bbrowning>
just add in the release notes "sudo rm /etc/.../cert.pem"
<enebo>
bbrowning: which a valid or invalid pem file is not
<headius>
hah
<bbrowning>
enebo: yes to that
<bbrowning>
enebo: but if we actually prefer these files first and want to try loading them as pem files, then better test with a valid pem file :)
<bbrowning>
I assume they're tried first because they're more desirable/correct/something
<enebo>
bbrowning: ah so it tries to reload those later as a pem format
<enebo>
bbrowning: vs first as a pem?
<enebo>
bbrowning: :)
<bbrowning>
right now it doesn't do anything with them as a pem I don't think
<mkristian>
bbrowning, the idea was to stay more in line with MRI openssl and fall back the java in case we have no PEM. i.e. on windows
<bbrowning>
mkristian: yeah
<enebo>
yeah I would say it should not be trying to hit it at all until it knows how to load a .pem and only then try to load it as a .pem and never use this java keystore loading on it
<enebo>
so this is transitional code not meant to really work but appear to be closer to how MRI is doing it?
<enebo>
a later version will load in this order but also be able to load the pem
havenwood has quit [Ping timeout: 244 seconds]
benlovell has quit [Ping timeout: 246 seconds]
hobodave has quit [Quit: Computer has gone to sleep.]
benlovell has joined #jruby
shellac has joined #jruby
badfish129 has joined #jruby
hobodave has joined #jruby
dabradley has joined #jruby
benlovell has quit [Ping timeout: 255 seconds]
badfish129 has quit [Ping timeout: 244 seconds]
dinfuehr has joined #jruby
dinfuehr has quit [Ping timeout: 244 seconds]
<headius>
looks like 0.9.9 CI was a mixed bag
<headius>
we should get those results green and piped into here...I know the noise sucks but we might have seen some of these regressions
<headius>
looks like the failures were issues with torquebox proxy?
<rtyler>
rubygems.lasagna.io FTW
<rtyler>
that torquebox proxy seems to be slowly getting worse and worse
<headius>
we need to get an official proxy somewhere :-\
<rtyler>
duh and or hello
<headius>
mkristian: you know some folks at Sonatype don't you?
<headius>
rtyler: ideally we'd get rubygems.org or sonatype to host it, but that's a third option
<rtyler>
why sonatype?
mje113 has joined #jruby
<mkristian>
headius, yes, I am still in contact with sonatype
<rtyler>
headius: the problem IMO with somebody else hosting it is the exact same problem we have right now with the torquebox proxy
<rtyler>
unless we have the keys to the castle, we will run into this problem again
<headius>
that's true, but we don't have problems like this with maven central or rubygems.org
<headius>
the problem with hosting it ourselves is that "ourselves" changes year to year
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/jruby-1_7 6b85ea3 Christian Meier: [build] get jossl-0.9.10 up for testing
JRubyGithub has left #jruby [#jruby]
<JRubyGithub>
[jruby] mkristian pushed 1 new commit to jruby-1_7: http://git.io/vssmw
<rtyler>
there's been a few constant folks in jruby headius :P
samphippen has joined #jruby
<headius>
heh, biggest problem with free software is that the rest of the world isn't free too
<headius>
admn you world!
<headius>
damn too
<rtyler>
so rubygems.lasagna.io is just an elastic beanstalk deployment of rubygems-servlets.war
<rtyler>
if anybody we know is willing to donate an IAM acocunt, then that covers it
<headius>
notice for everyone...I'm tweaking jruby commit messages to be shorter and message channel without joining
<headius>
that may open channel up to spam, we'll see...if it does I'll just go back to join mechanism
<rtyler>
or if redhat would be willing pay for some openshift bandwidth then, that's also super easy (scp'd the war there too)
<rtyler>
all the real work is 302'd to rubygems.org already
<GitHub134>
[jruby-openssl] headius pushed 1 new commit to master: http://git.io/vssY4
<GitHub134>
jruby-openssl/master 1b5dbbe Charles Oliver Nutter: Notify #jruby of CI results.
<headius>
rtyler: maybe we could just sneak it onto jruby.org... EY sponsors that still
<rtyler>
if they take .wars easily, then sure
<headius>
it's just an EC2 instance
<rtyler>
a single ec2 instance?
<headius>
let me check
<rtyler>
that'd be bad :P
<enebo>
I do not think we should use that box
<headius>
what is there at this point? just the URL redirects?
<enebo>
nothing I hope
<headius>
we've been phasing it out
<rtyler>
it has to download and cache gems for pom generation
<headius>
bugs.jruby.org, wiki.jruby.org etc are redirects on nginx there
<rtyler>
it doesn't serve up the raw gem
<headius>
rtyler: ahh, hrm
<enebo>
headius: oh I thought you set up something for redirects
<headius>
enebo: not I said the pig
<rtyler>
rubygems.lasagna.io is auto-scaled and doing ~2-5k requests a day
<headius>
that's still nsieger's stuff
<rtyler>
it's still under a $100 a month for me
<rtyler>
that's not bad
<enebo>
headius: oh well good thing I did not tell them we are no longer using it then
<headius>
there's a small amount of secret JRuby funds available for this but I don't know how much we're talking
<rtyler>
surely redhat would want some PR fluff for openshift amirite?
<headius>
do we know why the torquebox instance is having problems?
<rtyler>
no, it's running a very old version of the software though
<rtyler>
mkristian fixed a number of bugs for lookout in the tool
<headius>
jruby.org is a single c1.medium
<rtyler>
I wish I didn't know that
<headius>
the actual jruby.org pages are github pages though
<headius>
all that's on that server at this point are some url redirects and other miscellany
<enebo>
no other miscellany I know of?
<headius>
I don't know :-)
<enebo>
Although I thought we moved redirects off
<rtyler>
known unknowns
<mkristian>
the torqubox code is really old. the rubygems-verlets and nexus ruby plugin use the same library for converting a gem into a gem-artifact. well, tested by now.
<enebo>
well there is not much that could be on there
<headius>
enebo: I don't think so, because they rewrite URLs and redirect to github
<enebo>
it was our normative git repo (now gone)
<mkristian>
we never could upgrade since the torque box server needs jdk6 code
<enebo>
it was out website (now gone)
<headius>
bugs.jruby.org/#### goes to that issue on github
<headius>
etc
<enebo>
and the redirects are left
<bbrowning>
headius: mkristian: rtyler: fwiw, the torquebox.org proxy was moved to a new server a few weeks ago
<bbrowning>
I don't know all the server details, but I can find out. I do think it has a newer jdk though
<mkristian>
bbrowning, that is interesting news
<enebo>
but if that feature went away at the cost of maintaining a machine I am all for it :)
<headius>
bbrowning, mkristian: if we could just update that server and fix it I think this is solved then?
<bbrowning>
mkristian: new server has java 7 on it
<headius>
enebo: don't we have a google apps domain for jruby.org? I think that can do some limited URL redirects
<bbrowning>
it's now on a server hosted in-house at RHT vs being outsourced
<enebo>
headius: we do
<nirvdrum>
I know the launchers are being replaced, but why is there a bin/jruby and bin/jruby.bash that look mostly identical but aren't guaranteed to be so?
<headius>
nirvdrum: bin/jruby should be a copy of bin/jruby.bash unless replaced by native executable
<enebo>
headius: I also thought I have all cnames to point at gh pages at some point which is another reason I thought we fixed redirects
<headius>
bin/jruby should not be versioned, I think
<mkristian>
bbrowning, so could we see and get the rubygems-servlets running there instead of the old gem-proxy ?
<nirvdrum>
headius: I guess it isn't. I take it one of the maven tasks generate it then?
<headius>
nirvdrum: yeah, either normal build or bootstrap build
<bbrowning>
mkristian: I think that would be doable - I'd need to rope in tcrawley-away to find out some more details since he handled our server move. But I think we could replace what's running.
<headius>
bootstrap probably installs the native
<bbrowning>
if the server has become less stable in the past few weeks then perhaps we need to figure out why
<bbrowning>
something may be up with the new server
<nirvdrum>
headius: Weird. I tried bootstrap and it didn't regenerate.
<rtyler>
mkristian: isn't the new proxy mdoe URL incompatible with the old one?
<rtyler>
mkristian: instead of /releases there's /proxy/
<headius>
bbrowning: it was more than the last few weeks
<mkristian>
rtyler, there is still this legacy mode. need to look. yes, it is just different url mapping
<bbrowning>
headius: well, it may have been a month or so now even
<rtyler>
I've had issues for months with that proxy
<rtyler>
i'm just lucky I guess :P
<bbrowning>
definitely not months on the new server
<bbrowning>
we don't actually use the proxy anymore except for our TorqueBox 3 builds which don't happen that often
<bbrowning>
so we're usually not the first to find issues with it
<rtyler>
bbrowning: the proxy is an integral part of making jruby/gradle work, so I've definitely exercised the functionality quite a bit :P
<headius>
if that server is in no danger of going away, getting it stable would be the best option
<nirvdrum>
Bah. It looks like "mvn -Pbootstrap" won't regenerate the file. It'll only create it if it doesn't exist.
<headius>
it's already up and we use it right now
<headius>
nirvdrum: ah, yes...I remember someone else complaining about that
<headius>
I thought it was you
<enebo>
yeah and technically it is sponsored by RHT already
<headius>
right
<bbrowning>
headius: it's the same server that hosts torquebox.org, immutant.org, etc
<bbrowning>
quite a few jboss sites
<headius>
bbrowning: ok sure
<headius>
seems like updating it is the first line of attack then
<headius>
bbrowning: can you put mkristian in contact with tcrawley-away and copy enebo and I?
<nirvdrum>
headius: It could've been. But I usually remember very well when I've been scorned.
<bbrowning>
headius: I think I can actually do the update. I just need to get a few details from tcrawley-away.
<headius>
nirvdrum: maven is a harsh mistress
<headius>
bbrowning: ok
<bbrowning>
But that way I don't have to throw the task on him :)
<mkristian>
nirvdrum, the file only gets copied over on fresh github checkout
<headius>
yeah whatever works
<enebo>
nirvdrum: headius: I think it was eregon
<nirvdrum>
mkristian: Is there any reason to not just have it always generate on a bootstrap?
<enebo>
nirvdrum: we never scorn you…my friend
<nirvdrum>
eregon probably got burned by the same thing I just did then :-P
<headius>
I can't think of any
<bbrowning>
mkristian: the updated code is still a .war, right?
<headius>
I can't remember how to log into this EC2 instance
<headius>
I must have a pubkey somewhere here?
<enebo>
headius: I sure hope so since you push maven artifacts to sonatype :)
<mkristian>
bbrowning, yes
<headius>
I mean jruby.org instance
<enebo>
headius: to log in?
<headius>
oh, I guess I can just try my key
<headius>
I thought there was an EY blessed key or something
<enebo>
headius: also you can create ssh connection through EY dashboard
<headius>
ok
<nirvdrum>
enebo: Software scorns me all the time :-)
<headius>
I'm not sure I've ever logged into this machine
<mkristian>
bbrowning, but you need to build the -Plegacy profile - should still work. the main part is used by rubygems.lasagna.io
<headius>
dashboard says it is not associated with a PKI pair
baroquebobcat has joined #jruby
<headius>
so I guess just the awsm login
damnski has joined #jruby
<bbrowning>
mkristian: ok - I'll have to wait to hear back if I can just update this myself or if I have to submit a ticket through the managed hosting provider to do it.
<bbrowning>
old machine I could just log in and do whatever for the most part and I do have a login on the new one so we'll see :)
<rtyler>
headius, mkristian: is the torquebox proxy just used for testing, or as part of the build too?
<mkristian>
rtyler, it is used for testing and for installing the default gems in ./lib/ruby/
tcrawley-away is now known as tcrawley
<mkristian>
also build
<rtyler>
using ruby-maven?
baroquebobcat has quit [Quit: baroquebobcat]
<bbrowning>
rtyler: how does your openshift proxy work out? seem to do ok?
<rtyler>
it works just fine, the only problem with it that I've seen is that it's using the openshift free tier, so if it doesn't get a work out every now and again I think openshift pauses the instance
<bbrowning>
yes, that's true
tikue has joined #jruby
skade has joined #jruby
baroquebobcat has joined #jruby
<bbrowning>
mkristian: do you have any details on the new rubygems-servlet code resource-usage wise? does it use more/less disk space for caching, more/less memory, etc?
<tikue>
hey all, question about using a jruby gem with maven -- I'm using torquebox's rubygems proxy for maven, along with the gem-maven-plugin, to manage a ruby dependency. I'm able to load the dep with "require 'rubygems'; require 'dep'" but none of the dependency's methods can be found. Is this something that's commonly diagnosed?
<rtyler>
it does embed a bitcoin miner, I'm not sure if that's a problem... :)
<bbrowning>
heh
<rtyler>
the infrastructure that pays for itself!
baroquebobcat_ has joined #jruby
baroquebobcat has quit [Ping timeout: 244 seconds]
baroquebobcat_ is now known as baroquebobcat
<mkristian>
bbrowning, the disk usage should be about the same as the old code. not sure about the memory. maybe rtyler can sy more how the lasagne server is doing memory wise. but I ran it with out of the box jetty locally
<mkristian>
bbrowning, I did setup jruby-1_7 branch with new joss-0.9.10.gem which did work locally via the gem-maven-plugin
dinfuehr has joined #jruby
rsim has joined #jruby
<mkristian>
tikue, how do you execute jruby ? via the gem:exec goal (bad naming I know) ?
<rtyler>
bbrowning: rubygems.lasagna.io runs on an auto-scale group of m3.mediums (it's only requiring one machine right now) and it averages about 5% CPU utilization, it's running in tomcat so its hard to gauge how much the servlet's memory usage is
<rtyler>
bbrowning: but the java process was started with a max heap of 256m, so it's not that big :P
<bbrowning>
rtyler: cool
<mkristian>
I see. the ScriptingContainer does not know where the gems are found. you need to set environment GEM_PATH to target/rubygems which is the location the maven plugin installs the gems
<mkristian>
tikue, -^
<tikue>
mkristian: ah interesting -- what's confusing to me is that it didn't choke on the require lines, if it can't find the gems?
dinfuehr has quit [Ping timeout: 246 seconds]
<mkristian>
let me look at the exception first
<tikue>
it's (NoMethodError) undefined method `render` for GitHub::Markup::Module
<tikue>
(as a sanity check, I installed the gem regularly and ran the script in plain jruby, and that worked fine)
<mkristian>
just print out Gem.loaded_specs before this render line as see
<tikue>
it's like it's getting corrupted during installation
camlow325 has joined #jruby
tcrawley is now known as tcrawley-away
camlow325 has quit [Ping timeout: 244 seconds]
skade has quit [Ping timeout: 272 seconds]
camlow325 has joined #jruby
<headius>
twitter busted for anyone else?
<headius>
I'm getting invalid cert error in desktop client
<GitHub110>
[jruby-openssl] mkristian created better-default-cert-loading (+1 new commit): http://git.io/vssyg
<GitHub110>
jruby-openssl/better-default-cert-loading 1d3ba21 Christian Meier: the default cert could be PEM or a java keystore. it should load both whatver is used as default.
camlow325 has quit [Ping timeout: 246 seconds]
<nirvdrum>
headius: The web site seems to be working.
<bbrowning>
mkristian: my first take at replacing our old proxy with rubygems-servlet results in the proxying throwing 503 errors so I've rolled back
<bbrowning>
trying to dig into logs now to see what the actual errors were
<bbrowning>
I did end up having to edit web.xml and point to correct storage place
<mkristian>
bbrowning, if this works for you then it is ok
<bbrowning>
yeah I just added it as part of my deploy instructions
<mkristian>
enebo, it also first load the defatted cert as java keystore and if this fails tries it as PEM
<bbrowning>
a simple sed on it
<headius>
bbiab
<bbrowning>
so newer code and I gave it some more memory so hopefully things will run smoother
havenwood has joined #jruby
<mkristian>
bbrowning, will remove the defaults from the web.xml otherwise I can not configure those via java for embedded jetty.. let's see how it works
<mkristian>
bbrowning, now - the second "demand" - could you see jruby-1_7 with my jossl patch ?
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] nirvdrum pushed 2 new commits to truffle-head: http://git.io/vssFY
<JRubyGithub>
jruby/truffle-head dd8a80f Kevin Menard: Bump Truffle version.
<JRubyGithub>
jruby/truffle-head 7fcefac Kevin Menard: Updated jt to work with new JVMCI Graal builds.
JRubyGithub has left #jruby [#jruby]
<bbrowning>
mkristian: building now, although it may take a while because I wiped out lots of stuff in $M2_REPO/rubygems to test the new proxy :/
<mkristian>
:)
dinfuehr has joined #jruby
camlow325 has joined #jruby
enebo has quit [Quit: enebo]
camlow325 has quit [Read error: Connection reset by peer]
dinfuehr has quit [Ping timeout: 260 seconds]
colinsurprenant has quit [Quit: colinsurprenant]
camlow325 has joined #jruby
camlow325 has quit [Read error: Connection reset by peer]
<mkristian>
travis builds 9999 right now :)
drbobbeaty has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<GitHub149>
[jruby-openssl] mkristian merged better-default-cert-loading into master: http://git.io/vssN7
camlow325 has joined #jruby
lanceball is now known as lance|afk
camlow325 has quit [Ping timeout: 255 seconds]
yfeldblum has quit [Quit: Leaving...]
camlow325 has joined #jruby
Aethenelle has joined #jruby
camlow32_ has joined #jruby
camlow325 has quit [Read error: Connection reset by peer]
yfeldblum has joined #jruby
pitr-ch has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
mkristian has quit [Quit: This computer has gone to sleep]
shellac has quit [Quit: Computer has gone to sleep.]
mje113 has quit [Quit: Connection closed for inactivity]
donV has joined #jruby
<donV>
Hi all!
hobodave has quit [Quit: Computer has gone to sleep.]
djbkd has quit [Remote host closed the connection]