jrafanie has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
lopex has quit [Quit: Connection closed for inactivity]
sidx64 has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
sidx64 has quit [Client Quit]
sidx64 has joined #jruby
sidx64 has quit [Client Quit]
sidx64 has joined #jruby
sidx64 has quit [Read error: Connection reset by peer]
sidx64 has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
sidx64_ has joined #jruby
sidx64 has quit [Read error: Connection reset by peer]
sidx64 has joined #jruby
sidx64_ has quit [Ping timeout: 252 seconds]
sidx64 has quit [Client Quit]
sidx64 has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
<GitHub25> [jruby] kares pushed 2 new commits to master: https://git.io/vpju7
<GitHub25> jruby/master db8f5e3 kares: [ji] support (Java) String proxy unwrapping on a few places
<GitHub25> jruby/master 23dd24f kares: minor: remove unused private method
Specialist has joined #jruby
<GitHub45> [jruby] kares opened pull request #5180: [ji] RubyString implements CharSequence (master...rubystring-impl-charsequence) https://git.io/vpjgQ
lopex has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
Specialist has quit [Ping timeout: 245 seconds]
<kares> headius: have extracted bits from your (likely defunct by now) report_on_exception PR https://github.com/jruby/jruby/pull/3937
<kares> would be great if you have some thoughts on this https://github.com/jruby/jruby/pull/5180
<kares> ... since we only get a chance to do 'small' breaking changes in major releases
chasm has joined #jruby
<chasm> re: rvm , is jruby-head jruby-9.1.17.0 ? thanks
chasm has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
<travis-ci> jruby/jruby (master:db8f5e3 by kares): The build was fixed. (https://travis-ci.org/jruby/jruby/builds/381566210)
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
sidx64 has quit [Read error: Connection reset by peer]
sidx64 has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
kamal__ has joined #jruby
chasm has quit [Client Quit]
<kamal__> Hi, I am facing one issue with activerecord-jdbc-adapter on jruby 1.7.4. Here is the question on stackoverflow.
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
jrafanie has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
chasm has quit [Client Quit]
chasm has joined #jruby
Puffball has joined #jruby
chasm has quit [Client Quit]
sidx64 has joined #jruby
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
jrafanie has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
CT-Rahul has joined #jruby
<CT-Rahul> hi all
chasm has quit [Client Quit]
chasm has joined #jruby
<CT-Rahul> OpenSSL::SSL::SSLError: handshake alert, i am facing this error in jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
CT-Rahul has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
chasm has joined #jruby
chasm has quit [Client Quit]
drbobbeaty has quit [Ping timeout: 276 seconds]
drbobbeaty has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
drbobbeaty has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
jsvd has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
sidx64 has quit [Client Quit]
<GitHub84> [jruby] kares created deprecate-subclasses (+4 new commits): https://git.io/vpj7h
<GitHub84> jruby/deprecate-subclasses 83e4e76 kares: [refactor] avoid using Function1 -> replaced with Java 8 Consumer
<GitHub84> jruby/deprecate-subclasses e310f43 kares: deprecate Function1 and replace its only use with a Java 8 Consumer
<GitHub84> jruby/deprecate-subclasses 91ddfd8 kares: [fix] keyword extract -> don't use Hash#keySet since converts toJava
<GitHub186> [jruby] kares opened pull request #5181: deprecate jruby/core_ext Class#subclasses (master...deprecate-subclasses) https://git.io/vpj5q
<kares> kamal__: that might be non fixable depending on AR version you're using
<kares> it usually works when the AR pool is avoided -> never got to the bottom of this
<kares> anyways reconnect: true is really an anti-pattern on MySQL
<GitHub191> [jruby] kares pushed 1 new commit to master: https://git.io/vpjFv
<GitHub191> jruby/master bae5d69 kares: [fix] keyword extract -> don't use Hash#keySet since converts toJava
<GitHub28> [jruby] kares force-pushed deprecate-subclasses from 313e5e6 to 003b9ee: https://git.io/vpjFT
<GitHub28> jruby/deprecate-subclasses 3475604 kares: deprecate Function1 and replace its only use with a Java 8 Consumer
<GitHub28> jruby/deprecate-subclasses 3fb8039 kares: [refactor] avoid using Function1 -> replaced with Java 8 Consumer
<GitHub28> jruby/deprecate-subclasses 003b9ee kares: deprecate Class#subclasses from jruby/core_ext ... move it to JRuby...
jsvd has left #jruby [#jruby]
sidx64 has joined #jruby
sidx64 has quit [Client Quit]
sidx64 has joined #jruby
jrafanie has joined #jruby
kamal__ has quit [Ping timeout: 260 seconds]
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
shellac has joined #jruby
xardion has quit [Remote host closed the connection]
xardion has joined #jruby
shellac has quit [Ping timeout: 245 seconds]
subbu is now known as subbu|away
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
sidx64 has quit [Client Quit]
ahorek has joined #jruby
akp has joined #jruby
akp has quit [Client Quit]
ahorek has quit [Client Quit]
drbobbeaty has joined #jruby
subbu|away is now known as subbu
Specialist has joined #jruby
jrafanie has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
ahorek has joined #jruby
ahorek has quit [Client Quit]
jrafanie has joined #jruby
ahorek has joined #jruby
<GitHub35> [jruby] enebo pushed 1 new commit to jruby-9.1: https://git.io/vhe93
<GitHub35> jruby/jruby-9.1 9d6f877 Thomas E. Enebo: Fixes #5179. On eval of Time#strftime read from file: Java::JavaLang::Error: Unknown special char: Y...
<headius> heh
<headius> enebo: the problem is that the array gets accessed without going through Formt
<headius> Format
<headius> so the Format enum never inits
<headius> so the array never populates
<enebo> yes
<headius> there's a previous hack in there that kares commented out, I just commented on the issue
<headius> same thing you did
<headius> I have a fix that avoids needing a hack
<enebo> I was hoping I had commented enough to explain that
<GitHub163> [jruby] enebo pushed 1 new commit to master: https://git.io/vhe9u
<GitHub163> jruby/master 3b228b2 Thomas E. Enebo: Merge branch 'jruby-9.1'
<GitHub199> [jruby] enebo closed issue #5179: On eval of Time#strftime read from file: Java::JavaLang::Error: Unknown special char: Y https://git.io/vphSF
<enebo> headius: on one hand this static side-effect is a bit clever but on the other it is sort of surprising javac cannot see that constructor writing a static field. I guess I wonder what enum semantics actually are?
<headius> it never accesses the enum
<enebo> I mean are enums defined as lazy loaded on demand
<headius> if ASCIIEncoding
<enebo> yeah I explained all of that I thought
<enebo> in the commit message
<headius> ok what are you confused about then?
<headius> I must not be following
<enebo> why Java works this way I guess
<enebo> not that it does
<headius> well the enum is never initialized...so the enum values are never constructed
<headius> the array accesses elsewhere don't trigger that initialization
<enebo> so enums being defined is a halting problem if they have side-effects
<headius> the array is a static field on an unrelated class so accessing it doesn't trigger the Format constructors
<kares> staticInitializerHack :)
<headius> well, it's not a halting problem
<headius> but you have to actually access it
<enebo> yeah the access part I think was confusing part for me
<headius> all classes in Java are lazily initialized on first access
<enebo> I would expect them to be like types or I have thought of them at that level
<headius> static inits don't run until you access the class the first time though
<enebo> I guess I have not see this with types themselves
<headius> which includes constructing all the enum values
<enebo> but that does make more sense
<headius> the bug is that accessing the array does not trigger the initialization to fill it
<enebo> if I have a static field in RubyString but never reference it I do not expect that field to have been hit yet
<headius> the fix I was working on locally forced all accesses to go through a Format method
<enebo> yeah I totally understand why it did not work I just had mismatched expectations on how enums come into being
<headius> yeah it was confusing until I saw it was empty
<headius> at that point it had to be a class init issue
<enebo> It is doubly weird in a way since the method where we have that code path will reference the enum in a few ways ... but nothing constructs until the first actual reference
<kares> couldn't the current fix blow if say an agresive compiler removed dead code
<enebo> kares: yes
<kares> e.g. on graal - assuming in treats enums as side effect free
<enebo> kares: C1 runs first though
<enebo> kares: I would think once it is live it is ok to stay alive even if that init is DCEd
ahorek has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
<kares> okay maybe that wasn't a good example
<enebo> kares: but generically you are right it may not work somewhere
<kares> maybe if it was compiled into a native image ...
<enebo> kares: for all I know j9 kills it before it makes bytecode
<kares> najs - good to know
<enebo> kares: excelsior may kill it as an AOT perhaps
Specialist has quit [Ping timeout: 256 seconds]
<enebo> I did add a FIXME: too. I just did not think this was worth a lot of time this week
<kares> yy np just asking :)
<enebo> kares: yeah np I think it is good to bring up questions
<headius> I'll just finish my fix for your fixmu
<enebo> headius: yeah cool
<lopex> fixµ
<headius> 👍
<enebo> I thought about values() for looping and doing it outside then went bleh
<headius> yeah I'm just cheating
<headius> moved the static array inside format and used static util methods to populate it
<enebo> headius: so it will just static init with values and take it out of the constructor
<headius> so now you can't get it without going through format
<enebo> yeah I should have just did that. I was lazy :)
<headius> entertaining bug though
<headius> java puzzler quality
<headius> enebo: so I was never able to get stack overflow btw
<headius> I messed with it on friday but never got anywhere so went back to 9.2 finalization
<enebo> headius: well I should rerun again but you have your result numbers?
<enebo> headius: my merge was bad :P
<enebo> I will assume you will fix as you land your fix
<headius> enebo: yeah I'll fix you up
<enebo> headius: and rerun sqlite3 since I want to see how we differ
<headius> give me your exact command line again
<enebo> headius: I did run and notice invidual files will run clean so something in testing is messing something up
<enebo> RAILS=`pwd`/../rails jrake rails:test_sqlite3 --trace 2>&1 > aaa
<enebo> I am only redirecting because it creates a couple of hundred thousand lines of backtraces
<headius> no --dev in env or anytthing
<enebo> headius: no although --dev will cause other issues
<lopex> 9.2 soon ?
<headius> ok
<headius> lopex: if we figure out this error, maybe :-)
<enebo> lopex: Also perhaps an AS issue
<lopex> as/400 ? :P
<enebo> active support 400
<lopex> enebo: I biased so forgive me
<enebo> lopex: forgiven
<GitHub152> [jruby] headius pushed 1 new commit to jruby-9.1: https://git.io/vhe5L
<GitHub152> jruby/jruby-9.1 6b7d821 Charles Oliver Nutter: Try to avoid class init hacks by forcing accesses through Format....
<headius> hah
<headius> I feel like I've done this enum pattern before without these problems
<enebo> headius: yeah me too
<headius> I guess the main difference is that the accesses of the array happen directly within this class
<enebo> 5226 runs, 14463 assertions, 10 failures, 92 errors, 8 skips
<headius> normally they'd happen from outside and force some inits
<enebo> headius: no doubt we will access something right away so we never noticed when it was stood up
<GitHub52> [jruby] headius pushed 1 new commit to master: https://git.io/vhe5a
<GitHub52> jruby/master 44c6157 Charles Oliver Nutter: Merge branch 'jruby-9.1'
<enebo> only accessing an enum indirectly from an array or other structure is probably rare without at least some check like, "is it this enum value"
<headius> that fix isn't great but it avoids having a poison pill lying around
<headius> enebo: I guess I mark that as 9.1.18?
<enebo> headius: yeah
<enebo> lowest version should always be the fix until github allows us to mark multiples
<headius> ok I'm trying to run sqlite3 now
<enebo> so running independent tests with these problems do not trigger
<enebo> I SMELL PREPEND IN TESTS
<headius> you and your jrake
<enebo> I only have like 2 aliases too
<enebo> jrake and jb for mvn -pl core
<headius> so first off this is way slower than I remember it, not sure if I have some env issue
<enebo> you said that last week too...how long do you remember these tests taking?
<headius> and now it gets to this point and seems to just hang
drbobbeaty has quit [Ping timeout: 252 seconds]
<headius> maybe I'm just not noticing the hang soon enough
<headius> but it never completes
<enebo> def test_connection_pool_starts_reaper
<enebo> I will exclude this
<enebo> headius: you hitting this again?
<headius> I thought I nuked that
<headius> ugh, and ctrl+\ doesn't work?
<headius> oh I see, goes to redirect
<enebo> ah yeah
<headius> I added unless false to that test method def
<headius> I'll just delete it though
sidx64 has joined #jruby
<enebo> I am doing exclude...I will commit once I verify I put the exclude in right place
<headius> running with -v now and I'll see where it's hanging
<headius> test is completely deleted from my local copy now
<enebo> I would hope you do not hang since I don't without that test
<headius> oh dummy...unless false
<headius> I feel like a n00b
<headius> so yeah running running running now
<headius> doesn't appear to be belching stack traces yet
<travis-ci> jruby/jruby (jruby-9.1:9d6f877 by Thomas E. Enebo): The build was fixed. (https://travis-ci.org/jruby/jruby/builds/381844722)
<headius> 5226 runs, 14447 assertions, 12 failures, 89 errors, 8 skips
<headius> pretty close
<enebo> yeah so not backtraces in aaa?
<headius> I stopped appending but I see a stack issue in add_column
<enebo> yeah add_column and extract_backtrace I think
<headius> also translate_exception
<enebo> yeah I meant that
<enebo> ok same errors then
<enebo> I ran two files which had these independently and they pass
<enebo> so something is mucking up our inheritance during testing
<enebo> exclude pushed
<headius> going to try to get ancestors at that point
<enebo> I could add some loop state in the methods and dump ancestors once the count is highish
<enebo> My last look at ancestors did not show anything but I had a lot of them
<headius> well I'm hoping it will print out immediately before the overflow so I'll be able to find it
<headius> I want to confirm it's the issue my branch is supposed to fix before I spend time finishing the branch
<headius> yeah it works for a while doesn't it
jrafanie has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<headius> and eventually kaboom
<enebo> your prepend branch definitely has issues
<enebo> I could not run AS tests at all with it as one thing I remember
<headius> super fixes branch I mean
<headius> yeah included modules aren't overriding the search logic right
<enebo> so I have two ancestors and I see no difference from just the list of names
<headius> yeah I have some output
<enebo> you see any difference?
<headius> hmm
<headius> well unsure what before and after would be
<headius> I just have the output immediately before a failure
<headius> it's not different from the output immediately before that though
<headius> I mean, there will be as many of these calls as there are stack overflowing super calls
<enebo> headius: I did a short run and captured in one which did not fail
<enebo> headius: then a full run where I outputted only when I saw it call add_column 10 times (put a counter in it)
<headius> ok
<enebo> There are two unnamed modules in both otherwise all modules/classes are exactly the same
<headius> there is one def translate_exception in tests
<enebo> yeah
<headius> actionpack/test/abstract_unit.rb
<enebo> so the other big data point for me is that I can run some of these failures as individual test and they work without error
<enebo> so some test is doing something to us at some point during the run
<headius> yeah or triggering something to load that messes stuff up
<headius> like some additional prepends somewhere
<headius> so the weird thing is that these should be standard supers
<headius> all the definitions of this method I can find are within classes
<headius> i.e. not unresolved
<headius> but something might be triggering IR to make them unresolved
<enebo> I wonder if somehow these definitions are being used in another type but it is remembering the impl class or something like that
<enebo> yeah at this point I have no real clue here past something is changing it during testing
<enebo> Perhaps I can more simply make a brute force tester
<enebo> like add one failing test and all tests before it as n runs until it fails
<headius> %v_30 := instance_super(%current_module, %t_exception_31, %t_message_32, %v_0, callType: SUPER, name: translate_exception, potentiallyRefined: false)
<enebo> look legit
<enebo> unless current_module as input somehow leads to current_module as parent but something is goofy
<headius> hmm
<enebo> I will try this brute force idea
<enebo> I know first error in list and the files before it
<enebo> so I think I can maybe even bisect half the files and just keep eliminating
<enebo> funny thing is I feel like we will be surprised at whatever is wrong
<enebo> like it won't be prepend at all somehow
<headius> yeah huh
<enebo> stripping them away
<enebo> less files testing and some exclusions end up having no failures
<enebo> getting closer
<headius> they are unresolved in the test run
<headius> oh
<headius> it's a module in our code
<headius> but a class in theirs
<enebo> true
<enebo> ours is a little weirder but the class at bottom of our adapter includes that as a module
<enebo> but it is definitely a difference
<headius> yeah
<headius> well that at least explains how this can happen
<headius> heh yeah
<enebo> with luck it is just a bug from that difference on our side
<travis-ci> jruby/jruby (jruby-9.1:6b7d821 by Charles Oliver Nutter): The build was fixed. (https://travis-ci.org/jruby/jruby/builds/381857355)
<enebo> Ok I have it down to one file breaking another files tests: "test/cases/finder_test.rb" "test/cases/ar_schema_test.rb"
<enebo> so finder_test is doing something
<headius> hmm
<enebo> happens to be translate_excetion as first error
<headius> I don't see anything obvious yet
<enebo> it pulls in a lot of code too
<enebo> just deleting tests until it stops
<headius> yeah
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<enebo> somehow I noticed as reducing this that assert_raises tests seem important in seeing this
<enebo> I removed different chunks of the file and it woudl go away but then as I remove I can generally keep the problem if I leave those tests in
<enebo> which is not super instructive since the problem is translate_exception
<headius> yeah I went through most of the calls that test makes but none seem to do anything weird with hierarchies
<enebo> I am down to like 5 tests
<enebo> but weirdly I have seen removing different tests will eliminate the issue
<enebo> so both of those test must happen or the overflow in the NEXT FILE does not happen
<enebo> this is super duper weird still
<headius> yeah I can't figure out if that helps or not :-D
<enebo> interesting difference in ARJDBC vs AR in that a massive id in arjdbc raises an error for us but Ar seems to ignore it
<enebo> so that first test probably generates an exception that perhaps we catch again somewhere (or in AR) but it is different in how it bubbles out
<enebo> 'require "mocha/setup" # FIXME: stop using mocha'
<enebo> heh
<enebo> I do not see mocha being used but no doubt it uses some metaprogramming
<enebo> oh wait perhaps I did not save after I removed all those models
<headius> mocking could definitely muck something up
<headius> all bets are off
<enebo> oh ffs emacs is hanging
<enebo> well I had it down to those two tests with all fixtures required and loaded at top of file
<headius> how are you running just those tests
<headius> oh I see
<enebo> so check out updated finder test
<enebo> both of those private def methods seem to also be needed
<enebo> It is getting creepyt
<headius> weird
<enebo> aha
<enebo> It passed once
<enebo> yeah this is occasionally passing
<enebo> mostly not though
<enebo> headius: Can you run that and see same thing
<headius> enebo: so replace their finder_test with this
<enebo> yeah
<enebo> headius: I whittled it down to as small as I thought it could go
<enebo> but now I wonder if perhaps I run the single ar_schema_test 100 times will it fail on its own
<enebo> ok I cannot actually get it on its own to fail...that small finder_test needs to be there too
<enebo> I was able to remove most of the requires and leave the fixture to topic and companies
<enebo> b
<headius> enebo: yeah
<headius> hmmm
<enebo> class TestCase
<enebo> end
<enebo> parallelize_me!
<enebo> end
<enebo> if RUBY_ENGINE == "ruby" && PROCESS_COUNT > 0
<headius> enebo: it works without JIT
<enebo> it does?
<headius> oh wait no
<headius> it worked ok with jit with your modified test
<headius> no error
<enebo> well test/cases/finder_test.rb load error: cases/helper -- java.lang.StackOverflowError: null
<headius> not for me
<enebo> I cannot bootstrap the tests with JRUBY_OPTS=-X-C
<headius> works ok for me
<headius> oh wait...-X-C works but --dev errors?
<headius> nevermind, they both error
<headius> but jit threshold=-1 is ok
<headius> threshold=0 fails the same way
<headius> hmmm
<headius> something wrong with full build?
<headius> and by extension jit
<headius> so that is probably not related though
<enebo> headius: it is weirdly also a stackoverflow for me too
<enebo> just different
<headius> I still don't get stack error with your modified finder_test
<enebo> so I commented out minitest parallel stuff which I do not think is for these tests just in case
<enebo> headius: I am thinking it must be a race
<enebo> headius: and you added the larger snippet once I edited the gist?
<headius> aha
<headius> I just got it running a second time
<headius> so yeah some background thread
<enebo> So something comes in and modifies something in a bad way but not always
<enebo> once it is in a bad way though it never seems to resolve after that point
<enebo> since I get pretty much the same results on full runs
<enebo> headius: dinner for me
<headius> ok
<enebo> headius: are you on IM?
<headius> I have it handy
<enebo> I have been messaging you so I was not sure if perhaps I am not actually on hangouts
<lopex> what ims are you using ?
<lopex> enebo: btw have you seen that ebcdic video ?
<enebo> lopex: no
<enebo> lopex: google hangouts
<enebo> lopex: going to eat dinner
<lopex> enebo: especially for a dinner
projectodd-ci has quit [Ping timeout: 256 seconds]
<headius> enebo: last message I have from you in hangouts is from Thursday
<lopex> so unreliable
<lopex> enebo: busted
<lopex> headius: hey, have you thought about running windows ci on that windows subsystem for linux ?
<lopex> I know it's dev only
<headius> lopex: you mean the linux for windows thing?
<lopex> yeah
<lopex> headius: mri builds on it easily
<lopex> but I had problems with jruby
<headius> huh, do you remember what?
<lopex> headius: and hats down for them actually
<lopex> headius: I'll have to reproduce
<headius> I suppose this is a use case we should be looking to support
<lopex> headius: jvm and maven worked
<lopex> and all through that subsystem calls via glibc
<lopex> I think is quite astonishing
<lopex> er, like glibc calls through their impls
<lopex> headius: I can check that if you want
<headius> yeah that would be good, I don't have a windows handy right now
<lopex> headius: the only thing they lack is cgroups so the cannot run any docker thingies
<lopex> otherwise seems very complete
<lopex> headius: but jvm on linux is just a glibc compliant process right ?
<headius> lopex: well in theory, but that's a pretty big surface area
<headius> I remember seeing shipilev complaining about people trying to run Linux JDK on that Linux Bash environment
<lopex> hmm
<lopex> and there's this https://github.com/google/gvisor thing
<lopex> well, not windows topic of course
<lopex> so many layers
<lopex> headius: so wrt that compliancy it's more a glibc fault or what ?
<headius> I don't know what issues he was running into
<headius> I can't guess what they were either :-) in theory if you have the right libs and kernel functions it should work
<lopex> headius: but sementic incomatibility for a single function concurency issue can ruin the whole thing wrt compat
<lopex> not just apis
<lopex> and how do they deal witth it
<lopex> I dont even know if cec applies to them
<lopex> and cec was always dead code for us
<lopex> and I vaguely understand the concept :P
<headius> ahh interesting
<headius> so there's some heuristic in there for analysing complexity and bailing out?
<lopex> he was here in the channel
<lopex> on that very issue
<lopex> but we havent been able to pin point the case
<lopex> we had a suspicious pattern though
<lopex> but for me it was more like a|a*|+ etc oniguruma supports
<lopex> but there;s nothin I'm aware for look/ahead/behind to make it happen
<lopex> headius: though there might be some bad look-* interactions
<lopex> I think we should check against engines taht support those perf wise
<lopex> headius: I'm clueless otherwise
<lopex> or just lazy :P