<chrisseaton>
headius: the safepoints are already there from the GC - I think this will simply add a dependency - which doesn't modify the code
<lopex>
headius: one of those
<lopex>
headius: 100 ibu etc
<headius>
chrisseaton: well, we will end up emitting more of them, but yeah, JVM should be able to coarsen and move things around
<chrisseaton>
headius: I'm writing something on this stuff at the moment - I'll keep you in the loop and give you a look look
<headius>
my main concern at this point is that it's still a global safepoint, so it kills all sites when you have to send an event across threads
<chrisseaton>
yeah
<headius>
but I'm not sure how else to do it
<headius>
it could trigger at any of those polling sites
<lopex>
you mean deopt granularity ?
<headius>
I have not watched the assembly to see how it evolves in the presence of events
<headius>
lopex: right
<headius>
oh I just had one idea
<headius>
per-thread SwitchPoint
<headius>
that would at least reduce it to the code that thread is running
<lopex>
that would require some machinery
<headius>
oh, but I can't do that either because another thread could run the same code
<headius>
bother
<headius>
yeah, I have not been able to figure out how to keep it from being a global hit
<chrisseaton>
you say that you emit more safepoints - I don't think that's how it works - safepoints get put in the same place every time - using a SwitchPoint doesn't add a new one, it will use the existing ones - so your use of a SwitchPoint here should theoretically produce the same machine code and should be zero-overhead
<headius>
chrisseaton: we can certainly confirm that :-)
<chrisseaton>
but not sure in practice - recording the state for deoptimization may interfere in some way
<headius>
that's why I figured it would need more...it has to be able to branch from any of the specified sites into the interpreter
<chrisseaton>
thanks for that pointing out how you're doing it though - very helpful
<headius>
and if it can't see through side effects in the intervening code...
<lopex>
chrisseaton: you have an idea how truffle manages in js cases like class schemes etc ?
<lopex>
I guess it resembles ruby ivar cases right ?
vyorkin has quit [Quit: WeeChat 1.0.1]
<chrisseaton>
lopex: there's a paper on that - http://dl.acm.org/citation.cfm?id=2647517 - pay wall I'm afraid but I can send you a pre-print - PM me if you want it
fivebats has joined #jruby
<lopex>
chrisseaton: just a genuine curiosity after reading some v8 and MS papers
<chrisseaton>
we use a hidden-shape system - similar to V8
<lopex>
chrisseaton: MS impl is quite far away too
<lopex>
ah
<lopex>
well, I mean quite advanced
<lopex>
chrisseaton: does jruby-truffle benefit from taht ?
yfeldblu_ has joined #jruby
<chrisseaton>
JRuby+Truffle currently has a much simpler version
<lopex>
due to semantics ?
<lopex>
headius: anyways, US beers rock!
<lopex>
headius: and stupid europe wants to go blind
<lopex>
headius: you know you have 3000 breweries ?
josh-k has quit [Remote host closed the connection]
josh-k has joined #jruby
<chrisseaton>
lopex: it's just older - we'll use the new Truffle one at some point
yfeldblum has quit [Ping timeout: 244 seconds]
<chrisseaton>
headius: do you want to stop talking about 9k now that we're 9.0? Or are we keeping the shorthand?
josh-k has quit [Read error: Connection reset by peer]
<headius>
lopex: I'm not surprised...there's a new one in my area every other week it seems
<lopex>
chrisseaton: will the js-truffle impl land under OracleLabs ?
<Antiarc>
"This is a technique for speeding up floating point calculations on 64 bit machines by using immediate values for some floating point vales, similar to using Fixnums for integers. The pattern for Flonums is ...xxxx xx10 (i.e. the last two bits are 10, where for fixnums the last bit is 1). The object_ids of other immediate values have been changed to accomodate this change."
<headius>
nice description
<headius>
it has never been clear to me how important it is to match those object IDs
<Antiarc>
In the RubyBasicObject constructor, I've added this.setFrozen(this.isImmediate()); -- is this acceptable overhead, or should I manually assign the frozen flag for known immediate values?
<headius>
but we have generally done so
<Antiarc>
(working on TestObject#test_freeze_immediate, for reference)
<Antiarc>
Fixnum/Bool/nil, really
<headius>
probably should do in each
<Antiarc>
I guess it's better to do it in each
<Antiarc>
yeah
<headius>
hmmm
<headius>
are all numerics frozen now?
<Antiarc>
1.23.frozen? => true in 2.2.0
<Antiarc>
but the test only covers fixnums
<Antiarc>
Bigdecimals are not frozen
<Antiarc>
haven't tested rationals yet
<headius>
complex isn't
<headius>
that seems wrong
<Antiarc>
Rational(4, 6).frozen? => false
<headius>
nor BigDecimal
<headius>
oh, I did 1.0r and that was true
<Antiarc>
Looks like just floats, maybe a side effect of their tagging?
<headius>
recentish trunk build
<Antiarc>
Ah, I'm behind
<Antiarc>
I'm running preview1
<headius>
well, still...it seems like all the core numeric types should be implicitly frozen
josh-k has joined #jruby
<headius>
anyway...that's another debate
<Antiarc>
Yeah, but can't do it in RubyNumeric because of BigDecimal. Bleh
x1337807x has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<Antiarc>
undef_method(:method_missing) can actually remove it from a class, so calling Bar.undef_method :method_missing; Bar.new.method_missing => NoMethodError. What's the right way to handle that?
toshe_tm has joined #jruby
<Antiarc>
Perhaps if trying to undef method_missing, just remove it instead?
toshe_tm has quit [Read error: Connection reset by peer]
toshe_tm has joined #jruby
toshetm has quit [Ping timeout: 255 seconds]
Hobogrammer has quit [Ping timeout: 256 seconds]
noop has joined #jruby
erikhatcher has quit [Ping timeout: 244 seconds]
pchalupa has joined #jruby
benlovell has joined #jruby
snovity has joined #jruby
rsim has joined #jruby
erikhatcher has joined #jruby
dumdedum has joined #jruby
yfeldblum has quit [Remote host closed the connection]
thsig has joined #jruby
anaeem1 has joined #jruby
anaeem1 has quit [Remote host closed the connection]
marr has joined #jruby
shellac has joined #jruby
yfeldblu_ has joined #jruby
yfeldblu_ has quit [Ping timeout: 255 seconds]
skade has joined #jruby
vtunka has joined #jruby
drbobbeaty has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
ephemerian has joined #jruby
josh-k_ has joined #jruby
ephemerian has quit [Client Quit]
josh-k has quit [Ping timeout: 265 seconds]
yfeldblum has joined #jruby
yfeldblum has quit [Ping timeout: 258 seconds]
benlovell has quit [Ping timeout: 258 seconds]
yfeldblum has joined #jruby
benlovell has joined #jruby
yfeldblum has quit [Ping timeout: 264 seconds]
etehtsea has joined #jruby
erikhatcher has quit [Quit: erikhatcher]
justinmcp has joined #jruby
drbobbeaty has joined #jruby
kotk has quit [Quit: Leaving...]
diegoviola has joined #jruby
benlovell has quit [Ping timeout: 265 seconds]
skade has quit [Quit: Computer has gone to sleep.]
yfeldblum has joined #jruby
yfeldblum has quit [Ping timeout: 258 seconds]
tcrawley-away is now known as tcrawley
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] eregon pushed 1 new commit to master: http://git.io/NyZ1Jw
<JRubyGithub>
jruby/master 4bf104a Benoit Daloze: [Truffle] Fix constant lookup Mod::Const: no lexical scope in that case....
<JRubyGithub>
[jruby] eregon pushed 1 new commit to master: http://git.io/421a_Q
<JRubyGithub>
jruby/master 8cc02f8 Benoit Daloze: Revert "[Truffle] Reove BodyTranslator.translateCPath()": We need it for cases other than unqualified....
<nirvdrum>
Unfortunately, the timing was such that I couldn't get a RubyConf ticket before they sold out, so I won't be able to join you guys next week.
<enebo>
nirvdrum: ruby-talk generally has some tickets pop up
<nirvdrum>
I didn't fully understand the process, but I think tickets became non-transferable after 10/31. The only way to get one is to get on the RubyCentral waiting list.
<nirvdrum>
At least that's my understanding.
<enebo>
oh … must have been a problem at some point
Aethenelle has joined #jruby
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/master 6140817 Benoit Daloze: [Truffle] Kernel#pretty_inspect should call #inspect rather than #to_s....
<JRubyGithub>
[jruby] eregon pushed 3 new commits to master: http://git.io/9xO9Lg
<JRubyGithub>
jruby/master 9c70781 Benoit Daloze: [Truffle] Use LexicalScope.NONE for readibility and flexibility.
<JRubyGithub>
jruby/master 15b169b Benoit Daloze: [Truffle] Fix module/class constructors so they automatically set the constant in the lexical parent....
JRubyGithub has left #jruby [#jruby]
iamjarvo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
fivebats has quit [Quit: quit]
<headius>
chrisseaton: can you give me the short version of how you reify instance variables into different shape objects right now?
<headius>
I have a flag in standard JRuby to do that using a pre-created set of 10 classes (0-10 ivars), by inspecting the method hierarchy for possible ivars
<headius>
I believe Rubinius does something similar, but I've never made this code live
Aethenelle has quit [Quit: Aethenelle]
<headius>
nirvdrum: one big area enebo and I both want to see more work on is sharing between jruby proper and jruby truffle
<headius>
some of that will come in the form of moving code to Ruby, which we want for both Truffle and IR
<enebo>
nirvdrum: For IR we want to selectively load Ruby versions for JIT but likely continue using Java ones for interp
<enebo>
nirvdrum: Eventually hopefully Java versions can go away but startup is affected quite a bit from loading Ruby
<headius>
exciting times :-)
zeroecco has joined #jruby
joast has joined #jruby
Aethenelle has joined #jruby
<nirvdrum>
enebo: Well, if you want faster startup times, we should really discuss the require semantics :-)
DrShoggoth has quit [Quit: Leaving]
<enebo>
nirvdrum: yeah startup time anyway we can get it
<enebo>
nirvdrum: I know require semantics are a large chunk
<nirvdrum>
A very large chunk.
<enebo>
nirvdrum: I hope you have some more time to look at stuff like that in addition to pure Truffle code
iamjarvo has joined #jruby
triple_b has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<nirvdrum>
That's up to chrisseaton :-) But I suspect that won't be a problem.
<nirvdrum>
enebo: But some of this stuff you and headius would have to buy into. E.g., we've tossed around the idea of needing a flag to add .class file scanning during require.
<enebo>
nirvdrum: Interestingly I wonder if we can delineate when we do that?
<chrisseaton>
headius: hi
<enebo>
nirvdrum: Like perhaps for jruby-complete we need to do it from front of Ruby.init() but for normal jruby perhaps we enable it after kernel
<nirvdrum>
As far as I can tell, it's only needed for AOTC stuff. .class files in JARs should just be picked up from the classpath.
<enebo>
nirvdrum: ah true
zeroecco has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<nirvdrum>
enebo: And getting to the bottom of why .jar.rb is a specially-handled case. I guess I need to grab donV whenever he shows up.
vyorkin has joined #jruby
<enebo>
nirvdrum: hmm perhaps we need require_class and then AOTC generates an empty wrapper containing that
<chrisseaton>
headius: there's a shape object that maps names to either primitive fields or array elements. every object has a shape. when you add or remove an instance variable, or change the type, you get a new shape and the arrays are reallocated if needed
<enebo>
nirvdrum: Actually that idea is not too unreasonable
<nirvdrum>
enebo: And all of this is to satisfy transparent lookup when doing "require 'x'". "require 'x.class'" would work in either case.
frobs has quit [Quit: Leaving]
<nirvdrum>
Cutting down the suffix list would cut out a lot of the FS accesses my crazy load service PR was addressing.
<chrisseaton>
headius: I don't currently look at the number of statically visible instance variables - but that's mainly just a philosophy thing - I don't do static optimisations in JRuby+Truffle if I can possibly avoid it, for simplicity
<enebo>
nirvdrum: Well perhaps we should consider changing the requirement for 9000? I am not so against some of these early decisions
<headius>
chrisseaton: so the type shape changes but the active object does not...but when you go to allocate a new instance that shape is used to determine which one to use
<enebo>
nirvdrum: It is a major version and we have not put out a preview release. I am just trying to think of impact to existing users vs balance of improvement to general experiene
<headius>
sure...it wouldn't be difficult for what I have now to work lazily and subsequent objects would reify variables we've actually seen
<chrisseaton>
headius: I think currently we try to force all instances of an object so share the same shape, but our new implementation (not yet released) treats every instance on its own, and performs much better
<nirvdrum>
enebo: Yeah, I was trying to force it before 9k is released. It seemes this represents a good opportunity to revisit some sins of the past :-)
<chrisseaton>
instances of a class I mean
<headius>
my "shape" only involves the names and offsets of ivars, but it's there in the table of ivars
<enebo>
.class is pretty exceptional case but I think having a .rb shim as part of AOT could make removing it backwards compat by having them run jrubyc with —shim (just made up the option for sake of migration argumnt)
<nirvdrum>
enebo: The other one would be the service class lookup.
<chrisseaton>
so every object starts out empty - the class doesn't inform the layout at all (in the new faster version)
<headius>
ah...does the object reshape itself in memory on the fly?
<headius>
obviously I'm limited in what I can do to in-flight objects
<chrisseaton>
headius: no, there's some primitives and then an extension array, but we don't do any kind of class generation or things like that
<chrisseaton>
headius: I guess you're worried about indirection - you want actual fields in your RubyBaiscObject instances, right?
<headius>
Aethenelle: that *seems* ok :-)
<Aethenelle>
fixes the one error but creates a different one
<headius>
chrisseaton: indirection and allocation; that array is a whole extra object to hassle with
<enebo>
Aethenelle: you have to have a typo there :)
<enebo>
!= null is what you have?
<Aethenelle>
nope ==
thsig has joined #jruby
<headius>
Aethenelle: do you know why it fixes the error?
<headius>
enebo: if it's null he wants to return it
<chrisseaton>
headius: well in our new version we have our RubyBasicObject, then we have a DynamicObject as field in that, then instance variables go in fields in that, or in an extension array, so up to three levels of indirection, and it's still faster than what we have currently, as the shape system is more sophisticated
<chrisseaton>
headius: so the indirection isn't the key problem
<headius>
but we really should figure out if that can ever actually be null, and why
<enebo>
ah :) heh
<enebo>
I should have read the right side of that expression
<chrisseaton>
headius: I can show you the object stuff in depth at RubyConf of course
<headius>
chrisseaton: yeah, we should have some sessions while we're all there
<headius>
a separate DynamicObject inside the ruby object isn't really much different than having an array
<Aethenelle>
it seems like te implementation class in the error is stale. I wanted to try putting it there because it would prevnt me from having to sprinkle getMethodLocation() all over the code, but it seems to mess up prepend_features somehow
<chrisseaton>
yeah but then we have an array in the dynamic object as well
<headius>
allocation rate should be similar
<enebo>
headius: Aethenelle: yeah in which case is impl class null on method.
<enebo>
For years we have had some of that logic but subbu killed a lot of that when he killed rubyclass field
<headius>
chrisseaton: do all those objects fold away somewhere? because that's now two levels of indirection and one additional object
zeroecco has joined #jruby
towski has quit [Ping timeout: 255 seconds]
<Aethenelle>
enebo: headius: dunno, I didn't look into it... blindly adding getMEthodLocation there took the error count from 33/12 to 32/25
<chrisseaton>
headius: if the root object escapes, then no they don't get folded - but the careful way we guard and type the reads gets us the performance back and more
<enebo>
Aethenelle: ah curious which methods do it if not too big a problem
<enebo>
Aethenelle: probably a single case
<headius>
ok...my designs are influenced heavily by the lack of good EA so I have to meet JVM halfway...that means direct object fields
<headius>
it works, and it definitely reduces alloc rates
anaeem1 has joined #jruby
<Aethenelle>
i think it's something like InterpretedIRMethod
tcrawley is now known as tcrawley-away
<chrisseaton>
headius: and about doing more stuff in Ruby - the code we're importing from Rubinius is being carefully left unmodified, so it isn't being made Truffle specific and could be shared
<headius>
with indy, the ivar access sites bind pretty much down to field accesses, and if EA was working they might fold away
<enebo>
Aethenelle: well that is all Ruby loaded code
bbrowning_away is now known as bbrowning
<enebo>
Aethenelle: which perhaps we don’t use the field any more
<Aethenelle>
it gets used
<headius>
chrisseaton: yeah we'll have to figure out how sharing that code will work, collaboration and licensing-wise
<enebo>
Aethenelle: I mean from IR compiled code
<enebo>
Aethenelle: e.g. pure-Ruby code
<Aethenelle>
let me see if I can find one of the tests that failed to try...
<enebo>
Aethenelle: We might use it for Java impld methods perhaps?
pchalupa has quit [Quit: Leaving]
<enebo>
Aethenelle: I will take a quick peek also
<Aethenelle>
it's used for super resolution among other things
paulswil_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<enebo>
we use it to invalidate cache hierarchies
iamjarvo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<enebo>
but IR assumes it is non-null and is accessing the field directly so I think maybe this is a special case
tcrawley-away is now known as tcrawley
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] eregon pushed 3 new commits to master: http://git.io/c2xwwQ
<JRubyGithub>
jruby/master 0738040 Benoit Daloze: [Truffle] Add an alias and error on initialize_copy with different classes.
<JRubyGithub>
jruby/master 3037dce Benoit Daloze: [Truffle] Split Kernel#dup and #clone as they will differ.
JRubyGithub has left #jruby [#jruby]
iamjarvo has joined #jruby
<Aethenelle>
UndefinedMethod
<enebo>
Aethenelle: ah hmm :)
<enebo>
Aethenelle: not sure anything should be seeing that directly
<Aethenelle>
populateInstanceMethodNames
avsej has left #jruby [#jruby]
<Aethenelle>
doesn't really matter... changing that line isn't the right thing to do anyway... creates other problems that are almost as bad... just seeing if it rings a bell or something liek that...
e_dub has joined #jruby
<enebo>
Aethenelle: Hmm UndefinedMethod I thought we just used as a singleton null object pattern. I know in the past we tried saving it to indicate cached nomethod but I did not think we were doing that anymore
<Aethenelle>
enebo: it does get cached on lookup and it's used for method undef
<enebo>
Aethenelle: ok. I guess I also forgot undef
<Aethenelle>
... which, now that i think about it, may mean there's a missing test for prepend ...
dumdedum has quit [Ping timeout: 255 seconds]
<Aethenelle>
hrmm.... MRI fails the test I would have written anyway...
<enebo>
Aethenelle: Not to invite too much work but the abstraction for impl class I think is broken for this method.getImplementationClass() == smoe_type should really be method.isImplementedBy(some_type)
<enebo>
Aethenelle: Then it is compatible with null object pattern we are using for UndefinedMethod
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] headius pushed 1 new commit to master: http://git.io/T9zbUA
<JRubyGithub>
jruby/master 77a71b8 Charles Oliver Nutter: Tag Range#max failure for now.
JRubyGithub has left #jruby [#jruby]
<Aethenelle>
enebo: I'll add that to the cleanup list...
<enebo>
Aethenelle: thanks in advance :)
<Aethenelle>
np, it's a pretty simple refactor... almost a regex even...
imperator has quit [Ping timeout: 244 seconds]
<dfr|work>
morning folks
yfeldblum has joined #jruby
havenwood has quit [Remote host closed the connection]
<headius>
I can't reproduce that test:mri error from travis yet
tcrawley is now known as tcrawley-away
skade has quit [Quit: Computer has gone to sleep.]
tcrawley-away is now known as tcrawley
havenwood has quit [Remote host closed the connection]
noop has joined #jruby
vtunka has quit [Quit: Leaving]
nirvdrum_ has quit [Ping timeout: 258 seconds]
havenwood has joined #jruby
yfeldblum has joined #jruby
erikhatcher has joined #jruby
<rtyler>
headius: Lookout is going to have #2017 and #2032 fixed for the 1.7.17 release, FYI
tmio has joined #jruby
yfeldblum has quit [Ping timeout: 240 seconds]
<rtyler>
do you know of any other jruby-complete or jruby-core/stdblib bugs that we should take on?
phrinx has joined #jruby
postmodern has joined #jruby
etehtsea has quit [Ping timeout: 255 seconds]
postmodern has quit [Ping timeout: 265 seconds]
anaeem1_ has quit [Remote host closed the connection]
zeroecco has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
Aethenelle has quit [Ping timeout: 272 seconds]
phrinx has quit [Read error: Connection reset by peer]
tmio has quit [Quit: tmio]
paulswilliamsesq has joined #jruby
<headius>
rtyler: oh nice!
<headius>
rtyler: issues with lots of comments that we haven't addressed are always good :-) hard to say top priorities without some auditing
<headius>
there's always masked specs and tests to be filled out as well, but that's just a big grab-bag
zeroecco has joined #jruby
phrinx has joined #jruby
lanceball is now known as lance|afk
benlovell has joined #jruby
ludyte has joined #jruby
benlovell has quit [Ping timeout: 245 seconds]
pietr0 has joined #jruby
erikhatcher has quit [Quit: erikhatcher]
<Antiarc>
headius: the MRI test suite has a test which verifies that implicit calls to respondTo? (via typecasting) pass the second "true" parameter, testing that it can see private methods. JRuby can see private methods, but doesn't pass "true" so the spec fails. In this case, is it better to make JRuby match MRI's behavior, or just exclude the test? I think it's a bad test in that it tests implementation rather than return values, personally
<Antiarc>
But then you might have cases where people override respond_to? or something, maybe, and would want that boolean passed?
<headius>
Antiarc: do they check arity of the target method and only pass it when it will not error, or something?
<headius>
I don't see how you can pass that without breaking single-arg respond_to? impls everywhere
<Antiarc>
They do check the arity (which needs a fix in JRuby)
<Antiarc>
Right now it enforces 1..2 params, JRuby will accept 1..n
skade has joined #jruby
<Antiarc>
But I guess that's inbound
<Antiarc>
I need to test what you're asking
zeroecco has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<Antiarc>
yeah actually, that's what the test is testing
<headius>
the logic for the new test/unit hits io/console to get windows size, and it works on OS X 10.9, Ubuntu 14.04, but not on Travis in this configuration
cajone has joined #jruby
iamjarvo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<headius>
I've checked that the struct layout and the params are all correct for linux
<headius>
and it works on local linux installs
<asarih>
hmmm. I can give you a debug VM…
<headius>
that would be very helpful
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] mkristian closed issue #1549: openssl issue with rails and installed default gems http://git.io/QsTA9A
JRubyGithub has left #jruby [#jruby]
<headius>
bbiab, snow
diegoviola has quit [Remote host closed the connection]
<asarih>
kk
yfeldblum has joined #jruby
auxbuss has joined #jruby
<Antiarc>
headius: last night you said that rational was frozen in head - I just tested a head build and it doesn't appear to be. Can you doublecheck your install?
cajone has quit [Remote host closed the connection]
cajone has joined #jruby
cajone has left #jruby [#jruby]
<nirvdrum>
headius: Dunno how helpful this is, but I saw a very similar error way back in the early 1.7 days and Wayne had some fix for it.
triple_b has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
calavera has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
tmio has quit [Quit: tmio]
yfeldblu_ has joined #jruby
yfeldblum has quit [Ping timeout: 265 seconds]
cajone has joined #jruby
noop has quit [Ping timeout: 245 seconds]
paulswilliamsesq has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
vyorkin has joined #jruby
Antiarc has quit [Read error: Connection reset by peer]
zeroecco has joined #jruby
oblutak has quit [Ping timeout: 246 seconds]
cajone has quit [Remote host closed the connection]
cajone has joined #jruby
cajone has quit [Remote host closed the connection]
benlovell has joined #jruby
benlovell has quit [Ping timeout: 265 seconds]
calavera has joined #jruby
lance|afk is now known as lanceball
cajone has joined #jruby
cajone has quit [Remote host closed the connection]
postmodern has quit [Ping timeout: 264 seconds]
josh-k has joined #jruby
postmodern has joined #jruby
subbu has joined #jruby
cajone has joined #jruby
auxbuss has quit [Quit: I'm gone.]
rsim has quit [Quit: Leaving.]
anaeem1_ has quit [Remote host closed the connection]
x1337807x has joined #jruby
iamjarvo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
cajone has quit [Remote host closed the connection]
vyorkin has quit [Quit: WeeChat 1.0.1]
yfeldblu_ has quit [Remote host closed the connection]
havenwood has quit []
iamjarvo has joined #jruby
yfeldblum has joined #jruby
havenwood has joined #jruby
bbrowning is now known as bbrowning_away
<headius>
qmx: I'm getting rid of the "JRuby Wiki" team on Github, since you're the only one in it and you're a committer too
subbu has quit [Ping timeout: 245 seconds]
ludyte has quit [Quit: ludyte]
zeroecco has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<rtyler>
headius: OH, one other thing I wanted to discuss with you and enebo was jruby 1.7 backports
<rtyler>
I'm wondering if there are performance or bug fixes in master that you guys have neglected to backport to 1.7 due to time constraints
<rtyler>
we're going to have to support 1.7 internally for a while moving forward, and backports could/would be part of it
iamjarvo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<headius>
nirvdrum: yeah that does seem somewhat relevant
<headius>
I thought maybe the fileno was not getting pulled out properly for whatever stdio it's using, but when I gave it a totally bogus one the error's different
<enebo>
rtyler: we might have missed some backports that are simple but some like IO are completely different code now
<headius>
when we can't get a real fileno we map it to some bogus number > 10k
cajone has joined #jruby
<enebo>
rtyler: same for most performance issues since it is a completely different runtime
<rtyler>
right
cajone has left #jruby [#jruby]
<enebo>
rtyler: but let us know if we are missing something
<headius>
yeah we have continued to do most behavioral fixes against 1.17
<enebo>
rtyler: Our only caveat to backporting is solely risk
<headius>
up until maybe a few weeks ago when Antiarc and I have been hacking on 2.2 behavior
<rtyler>
that's fine by me, just wondering if I should expend time on it :)
<headius>
with 9k coming along, we'd certainly appreciate help maintaining 1.7 line
Antiarc has joined #jruby
<enebo>
rtyler: we would love it if time was spent on it. Personally, I predict 1.7 released until 2016
zeroecco has joined #jruby
<enebo>
released=releases
<headius>
yeah...maintaining it for another year is not unreasonable
Antiarc has quit [Remote host closed the connection]
<headius>
asarih: so, debug vm...show me how
Antiarc has joined #jruby
<headius>
Antiarc: checking
<asarih>
headius: send email to support@travis-ci.com, please.
<headius>
looks like I'm 700 commits behind MRI trunk at the moment
<headius>
asarih: can do
<asarih>
Danke
tcrawley-away is now known as tcrawley
<rtyler>
enebo: that wouldn't surprise me, the big thing that might make this challenging for me is whether we need to support many versions of the ruby language between 1.8, 1.9 and 2.x
<enebo>
rtyler: Feels like if you are on 1.9.3 then 2.2 will not be a big deal
<enebo>
rtyler: 1.8.7 is a much larger visible gap
<rtyler>
yeahhh... :/
johnsonch is now known as johnsonch_afk
<qmx>
headius: okie dokie
<headius>
rtyler: we have not had patches for 1.8 behavior come in for a very long time
josh-k has quit [Remote host closed the connection]
<headius>
we might be willing to accept patches to get 2.0 mode working better in JRuby 1.7, but there's a lot of features only supported on master
josh-k has joined #jruby
* rtyler
nods
<rtyler>
I don't have nay requirements around anything 2.x right now thankfully
<headius>
blast, I thought running through maven would reproduce this issue
<Antiarc>
still on the ioctl issue?
<Antiarc>
I couldn't reproduce it with nohup or a tty-less SSH session
postmodern has quit [Ping timeout: 260 seconds]
josh-k has quit [Ping timeout: 264 seconds]
iamjarvo has joined #jruby
drbobbeaty has joined #jruby
<headius>
Antiarc: I'm requesting help from travis folks, but I discovered I can patch the test logic to gracefully fail when we have this error
<headius>
I field #2168
<headius>
filed
cajone has joined #jruby
tcrawley is now known as tcrawley-away
drbobbeaty has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<Antiarc>
headius: I'm +13 commits right now, all centered around test_object. Each is small, but they're a decent chunk of code all together. Do you want them as separate PRs?
calavera has joined #jruby
toshe_tm has quit [Ping timeout: 245 seconds]
<headius>
Antiarc: if they're separate commits a single PR is ok
<Antiarc>
ok
<headius>
asarih: no reply yet :-\
<asarih>
headius: hit reload button incessantly.
<headius>
I shall do so
<asarih>
or I can just give it to you.
<asarih>
hang on.
<headius>
that's an option too :-)
<headius>
maybe my gmail tab has gone stale
<asarih>
pm
<asarih>
plz
ludyte has joined #jruby
cajone has quit [Remote host closed the connection]
mkristian has quit [Quit: bye]
zeroecco has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
nirvdrum has quit [Ping timeout: 265 seconds]
cajone has joined #jruby
cajone has quit [Remote host closed the connection]