skade has quit [Quit: Computer has gone to sleep.]
yfeldblum has joined #jruby
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/master 3092c8f Charles Oliver Nutter: Use invokeExact to avoid varargs and conversions on handle invoke.
<JRubyGithub>
jruby/master 98f7b3f Charles Oliver Nutter: Jit logging should log successful target script compile.
<JRubyGithub>
[jruby] jrubyci pushed 3 new commits to master: http://git.io/hVCL3g
<JRubyGithub>
jruby/master 6f40ed6 Charles Oliver Nutter: Missing visitor call from keyrest instr.
JRubyGithub has left #jruby [#jruby]
nirvdrum has joined #jruby
yfeldblum has quit [Ping timeout: 272 seconds]
nirvdrum_ has quit [Ping timeout: 245 seconds]
<cpuguy83>
So if I want to build 9000, is that master?
<headius>
cpuguy83: indeed!
<cpuguy83>
headius: For automated build updates... should I run the test suite? Seems like master tests are failing often.
<headius>
they are at the moment but we get them green occasionally
<headius>
a good quick test on master is rake spec:ruby:fast
<headius>
it takes about a minute on my machine but it hits a lot of stuff
<headius>
there were four failures a few days ago but I think enebo may have fixed them
drbobbeaty_ has quit [Ping timeout: 244 seconds]
rsim has quit [Quit: Leaving.]
nirvdrum has quit [Ping timeout: 245 seconds]
nirvdrum has joined #jruby
<cpuguy83>
headius: So for the docker official images, I have to point at a specific commit hash. Would really like to get 9000 up though so people can test it out, play with it easily.
<cpuguy83>
So maybe weekly we can find a clean commit and update?
<headius>
cpuguy83: ahh sure...we could do a pass to get things a bit greener and I could get the JIT running without extra flags
tesmar has joined #jruby
<cremes>
sounds like it might be time for a “release candidate” of jruby9k
<headius>
yeah, we plan to do that no later than RubyConf
<cpuguy83>
That'd be cool. "Want to play with 9000? docker run jruby:9000"
<headius>
current schedule looks like RC for RubyConf, and final some time around MRI 2.2
rsim has joined #jruby
<headius>
cpuguy83: can you open an issue asking for a stable hash to use?
<cremes>
i like your idea of synchronizing major releases with MRI at Christmas
<cpuguy83>
headius: Sure.
<headius>
cremes: we'll finally have parity with this release, so it's a good time to start
<cremes>
indeed. it’s incredible how much work y’all have done. jruby is the “alternate runtime” to beat, no question.
<headius>
and yet it still feels like we've only just begun!
<cremes>
ha! i look forward to jruby kicking some ruby ass for years to come.
<JRubyGithub>
[jruby] jrubyci pushed 1 new commit to jruby-1_7: http://git.io/Y2B_mw
JRubyGithub has left #jruby [#jruby]
<JRubyGithub>
jruby/jruby-1_7 87bf291 Charles Oliver Nutter: Clean up some verbose warnings about arg lists.
drbobbeaty has quit [Ping timeout: 260 seconds]
drbobbeaty has joined #jruby
tesmar has joined #jruby
johnsonch_afk is now known as johnsonch
erikhatcher has joined #jruby
iamjarvo has joined #jruby
noop has quit [Ping timeout: 258 seconds]
<enebo>
lopex: any chance to look at that weird thang today?
vtunka has joined #jruby
jaffachief has quit [Quit: Computer has gone to sleep.]
drbobbeaty has quit [Ping timeout: 258 seconds]
subbu has joined #jruby
e_dub has quit [Quit: ZZZzzz…]
yfeldblum has joined #jruby
yfeldblum has quit [Ping timeout: 240 seconds]
<cpuguy83>
Does the pre-built package have anything special about it, or is just a cleaned up "mvn package"?
subbu has quit [Quit: Ex-Chat]
<headius>
it's an mvn -Pdist package, but shouldn't be a whole lot different from a bootstrapped mvn package
<headius>
you probably want the -Pdist to be sure though
<cpuguy83>
And for the actual releases, would you prefer me to use the pre-built ones or build from source. Currently using the pre-builts but was thinking to just build from source since I'm doing this 9k one anyway.
subbu has joined #jruby
mje113__ has joined #jruby
Aethenelle has joined #jruby
<enebo>
cpuguy83: I think it is fine either way but definitely do -Pdist
<cpuguy83>
Cool
<Aethenelle>
headius: I looked at MRI's include code over the weekend. I have no idea how jruby's is any different.
<headius>
Aethenelle: hmm
<Aethenelle>
i'm working on stepping through jruby's right now in case i see something i'm missing in the code but...
jaffachief has joined #jruby
joast has quit [Quit: Leaving.]
<Aethenelle>
wait... i think i just found it...
<Aethenelle>
YES! let me get this onto the 1_7 branch and create a PR
benlovell has quit [Ping timeout: 272 seconds]
colinsurprenant has quit [Quit: colinsurprenant]
jaffachief has quit [Quit: Computer has gone to sleep.]
<enebo>
lopex: UNSAFE was the issue
<lopex>
enebo: hiya
<lopex>
enebo: I was about to respond
<lopex>
enebo: then there's a bug in it
<lopex>
enebo: is the coderange valid ?
<enebo>
lopex: yeah np. So knowing this code is responsible is a fix to remove but it would be nice to make it work
<enebo>
lopex: the CR is valid
<enebo>
lopex: well it claims to be valid
<enebo>
lopex: but I would expect ooo * 15 to be valid since ooo shold be valid
e_dub has joined #jruby
<enebo>
oh
<enebo>
so I think I can see one pattern
joast has joined #jruby
<enebo>
The thing returns 10 and it should return 4
joast has quit [Client Quit]
<enebo>
So I wonder if it is walking 2 and then hitting unsafe which increments by a whole long which ends up being 10?
<lopex>
enebo: I think the += 4 made sense in previous algorithm
<lopex>
enebo: at what time was that ?
<enebo>
Date: Thu Sep 18 12:53:25 2008 +0000
<enebo>
:)
<lopex>
enebo: I remember that
<headius>
[Bug #4366][ruby-dev:43170]
<enebo>
So wrong algo for 6 years
<headius>
lopex: I should have waited longer!
<lopex>
headius: I just came home
<headius>
other changes in that method are older
<lopex>
if it's the same change as that in mri the it's safe to ship
<headius>
the UNSAFE block there is a direct re-port from MRI
<lopex>
it's too hot method to not show early
<headius>
in the patched version
<headius>
yeah
<headius>
that's true
<headius>
though this bug didn't show :-)
<lopex>
how so!
<lopex>
it did
<lopex>
:)
<headius>
well, eventually!
<headius>
heheh
<headius>
enebo: test/spec?
<headius>
I fixed it, you can add a test somewhere or figure out if there's an exclude or tag :-)
<headius>
enebo: unrelated... I started playing with the CodeSizeEvaluator in ASM
<lopex>
headius: I'm still of signedness issues there
<lopex>
for longer strings
<lopex>
*afraid
<enebo>
headius: yeah is it pretty good?
<headius>
lopex: yeah, MRI is using VALUE, which is void* it wraps MethodVisitor and gives you an
<headius>
oops
<lopex>
headius: we'd have to mask longs with 0xffffffff
<headius>
last part of that is for enebo, plus "estimate of bytecode size"
<headius>
enebo: it looks ok
<headius>
lopex: the code that uses the longs appears to do the right thing, no?
<headius>
masking, >>>, etc
<lopex>
headius: isnt VALUE just unsigned long ?
<headius>
lopex: oh, it might be... unsigned either way though
<enebo>
headius: I was talking to subbu about code esitmation after you left. I was wondering if 2 compile attempts makes more sense
<lopex>
then you have discouraed me from todays hacking
<enebo>
headius: first failed attempt can return actual bytecode count and then chunk up knowing
<headius>
enebo: we can get some metrics about code size after compiling with and without JIT passes
mkristian has joined #jruby
<enebo>
headius: for me I was wondering whether we want to save estimates per instr but then have to maintain that info or fail on attempt and gather sizes as input (which requires no code maintenance)
<headius>
yeah
<headius>
I thought about doing that too but it will get out of sync immediately
<enebo>
headius: because in the former it means any changes to bytecode layer means updating the estimates
<enebo>
yeah
<headius>
we could gather metrics though....a set of code examples that hit all instrs and use them to estimate
<headius>
this is more weird offline processing though
<enebo>
headius: as a one-time thing?
<headius>
if it's a one-time thing, it would be a weird offline processing thing :-)
<enebo>
headius: yeah. The 2 compile attempt way seems like it would just work forever at the cost of large things taking longer
<headius>
I filed a bug for the failure I was able to get running jit passes against interpreted scope
<enebo>
headius: but huge things will never be fast anyways
<enebo>
headius: and really it is mostly for AOT
<headius>
yeah
<headius>
if int can run jitted code that does jit passes jit can just fail
<enebo>
headius: So I like the 2 compile idea
<headius>
I can certainly rig that up again
<enebo>
Each BB compiled can generate instrs per BB and if failure then a pass can recieve that data to chunk up the method
josh-k has quit [Remote host closed the connection]
<headius>
now what to do about concurrency? :-D
<headius>
we're back to needing to be able to clone a scope
<enebo>
we also talked a tiny bit about making synthetic calls since we need them for literals and general code size problem
mkristian has quit [Client Quit]
josh-k has joined #jruby
<enebo>
When I was more actively hacking on IR scope I was working it towards having a cleaner reset
<headius>
I can do more to embed logic behind indy sites for stuff that becomes constant (bytelists, symbols, etc)
<headius>
but there's still an upper limit
<enebo>
Since much of the data stored should be thrown out in cases
<headius>
strftime etc
<enebo>
So flags are not an enum set and easily changed
<enebo>
headius: yeah large single value literals is another case I forgot about
<enebo>
headius: Has a >32k string literal ever been encountered in the wild?
<headius>
not to my knowledge!
<enebo>
It is an issue though :)
<chrisseaton>
enebo: maybe a >32k here string
<headius>
there could be another pseudo-JIT mode that just emits a purpose-built interpreter for that method
<chrisseaton>
someone embedding a payload like a binary in base64
Aethenelle has quit [Ping timeout: 244 seconds]
<enebo>
chrisseaton: yeah it is definitely possible. I am just unsure if I have ever seen it or not
<headius>
chrisseaton: it would be unusual to see a 32k heredoc without some string interpolation that would break it up
<headius>
binary encoded content is one case I suppose :-P
zorak8 has joined #jruby
<headius>
chrisseaton: what do you think you'll be able to show for rubyconf btw?
josh-k has quit [Ping timeout: 260 seconds]
<headius>
it's great you'll be there so we can focus on non-truffle in our talk
<enebo>
headius: chrisseaton: So if we do need to do that then it should be the same as we talked about yesterday for hash/array literals. We add a new enum and extend our RubytimeHelperCall for appends
benlovell has joined #jruby
<enebo>
headius: I guess benefit for string data is it does not contribute to overall method bytecode size in the same way as a list/array
colinsurprenant has joined #jruby
<headius>
yeah
<headius>
I'd be curious about persisted IR size for some of those monster methods
<enebo>
add a new enum == ADD_TO_STRING_LITERAL, A DD_TO_HASH_LITERAL, ...
<headius>
enebo: so I wasn't clear yesterday...is it a lot of work to make scopes cloneable?
<headius>
we have to do something since the passes aren't thread-safe while interpreting
<enebo>
headius: for mixed literals (literal values + calls) it gets big because it uses lots of temp vars
<enebo>
headius: Less the amount of work than determining what to throw out at clone time
<enebo>
headius: for a pure clone we could try and clone everything but we really don’t need to if we know we plan on running passes
<enebo>
headius: If we know the fallback will be to run other passes than many of those data structures will get replaces
<headius>
only need to clone the mutable bits obviously
<enebo>
headius: also true
<enebo>
headius: I am thinking a clone can wipe out all the pass data and maybe just be flags, cfg
<enebo>
linearizedcfg and live var stuff can just not survive but my round about answer if we need to examine every field in IRScope and see hwat should be saved or not
<enebo>
because cloning it all is a lot
<enebo>
and probably not useful
<enebo>
headius: Did you commit that StringSupport diff?
joast has joined #jruby
<headius>
I did not!
rsim has joined #jruby
<headius>
I merely create the patches
drbobbeaty has quit [Ping timeout: 250 seconds]
skade has quit [Quit: Computer has gone to sleep.]
<headius>
enebo: anything else for 1.7?
<headius>
er 1.7.16
<enebo>
headius: I don’t think so but I have not spun up any builds yet
<enebo>
headius: this string bug was last one in list
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/test-load-from-jruby-classloader 594390c Christian Meier: each required jar will be added to the LOAD_PATH removing the need to handle...
<JRubyGithub>
[jruby] mkristian force-pushed test-load-from-jruby-classloader from 9214912 to 3c40d9a: http://git.io/FIl8Cw
<JRubyGithub>
jruby/test-load-from-jruby-classloader de09868 Christian Meier: load resources from jruby-classloader
<JRubyGithub>
jruby/test-load-from-jruby-classloader 37e3e66 Christian Meier: putting the jruby-classloader as context classloader means that all attached "jars" can be found via...
JRubyGithub has left #jruby [#jruby]
vtunka has quit [Quit: Leaving]
<cremes>
how about some win32ole love for 1.7.16? ;)
Aethenelle has joined #jruby
<Aethenelle>
this time w/ a test
<enebo>
cremes: going to CA for Javaone on saturday. Gotta get this one out today or tomorrow at latest
<cremes>
enebo: alas
<enebo>
:(
jwinter_ has quit [Quit: Connection closed for inactivity]
<enebo>
headius: this patch does not work correctly
<headius>
enebo: madness!!
<headius>
what's wrong with it?
<enebo>
it prints out a lot of o’s now
<enebo>
not just 2
drbobbeaty has joined #jruby
calavera_ has joined #jruby
<enebo>
headius: p ('øøø' * 15).slice(43, 230)
<bbrowning>
enebo: running TB3 integs against jruby 1.7.16-SNAPSHOT build has 10 failing integ tests - looking into those now
<enebo>
"øøøøøøøøøøøø"
<enebo>
MRI only shows 2 (albeit in unicode escape form — different bug)
<headius>
enebo: hmmm
<headius>
yeah it should only be 2...sorry I didn't confirm output
<headius>
strange though
<enebo>
bbrowning: thanks good buddy
<headius>
maybe you should just comment out unsafe logic for .16 and we'll figure it out right
<headius>
I may just rewrite the logic because I've done long-stride byte[] walking elsewhere
<headius>
actually no, I'm not committing to doing that :-)
<enebo>
headius: yeah removing unsafe probably is not end of world
<enebo>
headius: regardless of speed this is hinky O(n) crap
Aethenelle has quit [Ping timeout: 245 seconds]
<headius>
yeah
<bbrowning>
hmm a NoMethodError (to_ary) in org/jruby/RubyKernel.java:386:in `Array'
benlovell has quit [Ping timeout: 260 seconds]
<headius>
hmm, that would be updated coercion logic
<enebo>
bbrowning: any of your gems change recently?
<headius>
for Array
<bbrowning>
enebo: I'm looking into the gems - we have most of the deps pinned down 100% but we may be missing a few
<bbrowning>
but since there were changes around to_ary recently, I thought it may be related to those ;)
<enebo>
headius: so I will land your patch but disable it for 1.7.16
<enebo>
headius: what you have might be slightly wrong maybe but probably more right than what we had
<enebo>
I will add a fIXME and reference github bug
<headius>
wait
<headius>
don't land it if it's completely wrong, it may be screwing up stuff that worked
toady00 has quit [Remote host closed the connection]
<enebo>
headius: it will be false && UNSAFE
<enebo>
with a FIXME right above
yfeldblum has joined #jruby
<enebo>
oh I might see a difference in port
<headius>
ok
yfeldblum has quit [Ping timeout: 258 seconds]
benlovell has joined #jruby
tylersmith has joined #jruby
<enebo>
hmm getLong should use ’s’ but it still did not fix it
JohnBat26 has joined #jruby
<enebo>
heh found another difference
<enebo>
man I dislike reading C :)
<headius>
2014-09-25T11:11:34.055-05:00: JITCompiler: could not compile:Format.Date::Format.Format at /Users/headius/projects/jruby/lib/ruby/2.1/date/format.rb:5 because of: "bytecode size 44159 too large in MODULE_BODY Format[/Users/headius/projects/jruby/lib/ruby/2.1/date/format.rb:5]"
<headius>
that's my active check rather than failing at verification time
<headius>
using existing jit.maxsize value
<enebo>
funsville
<enebo>
ok can I find a third difference
<enebo>
p++ was happening twice per loop for second change
<headius>
woo: 2014-09-25T11:12:46.133-05:00: JITCompiler: could not compile:Rake.Rake.Rake at /Users/headius/.rvm/gems/ruby-2.1.2/gems/rake-10.3.2/lib/rake/application.rb:10 because of: "bytecode size 110343 too large in MODULE_BODY Rake[/Users/headius/.rvm/gems/ruby-2.1.2/gems/rake-10.3.2/lib/rake/application.rb:10]"
<headius>
big class body in main rake.rb
<headius>
oooooo
<headius>
bytecode size 207251 too large in CLASS_BODY OptionParser[/Users/headius/projects/jruby/lib/ruby/2.1/optparse.rb:212]
<headius>
enebo: more good news: the only bodies that fail to compile with threshold=0 during rake -T are all code size
benlovell has quit [Ping timeout: 250 seconds]
<headius>
so at least for that scenario, JIT is handling all syntax and not breaking
<headius>
yay
<enebo>
cool
<enebo>
yeah so screw this code
<enebo>
unless SIZEOF_VALUE for some reason should be 4
andreanastacio has joined #jruby
<enebo>
headius: also I think if this method is wrong then we probably have at least one other method in here which should be changed
<enebo>
headius: so perhaps I will just delete this altogether with a FIXME saying we are missing unsafe version
drbobbeaty has quit [Read error: Connection reset by peer]
statonjr has joined #jruby
<enebo>
I will refer to this bug so people can find the patch as a starting point
toady00 has joined #jruby
drbobbeaty has joined #jruby
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/master 82d2b91 Charles Oliver Nutter: Try JIT twice....
<JRubyGithub>
jruby/master 10375b0 Charles Oliver Nutter: Get JRuby.compile using new compiler stuff.
<JRubyGithub>
[jruby] jrubyci pushed 3 new commits to master: http://git.io/MOUk4Q
<JRubyGithub>
jruby/master e5cf31f Charles Oliver Nutter: Remove static reference to handle for removed method.
JRubyGithub has left #jruby [#jruby]
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/jruby-1_7 01eebe6 Thomas E. Enebo: Fixes #1986. String#slice on multibyte chars raise Exception
<JRubyGithub>
[jruby] jrubyci pushed 1 new commit to jruby-1_7: http://git.io/5ATwig
JRubyGithub has left #jruby [#jruby]
<headius>
enebo: ok
<headius>
enebo: that commit lands two-try JIT
<headius>
given that stuff seems to be working ok with JIT I'm gong to work on non-indy backend
<headius>
before optimizing indy backend
calavera_ has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<enebo>
headius: yeah non-indy is exciting stuff
<enebo>
headius: I am really curious to see the delta
<enebo>
headius: in both eventual speed and warmup
<bbrowning>
so I need to figure out what it is we're now calling to_ary on that has no to_ary method
calaver__ has joined #jruby
<enebo>
bbrowning: yeah that would be helpful. headius fixed a problem but it might have exposed another problem and you obviously have code which hits it
calavera_ has quit [Ping timeout: 240 seconds]
<bbrowning>
the code that hits is just a fairly basic rails4 app ;)
<enebo>
bbrowning: :)
<enebo>
yay
xcv has joined #jruby
calaver__ has quit [Client Quit]
<bbrowning>
it looks like the code in question is calling Array.new(format) where format is an object of type Mime::Type
<headius>
enebo, bbrowning: went through the Array logic in MRI pretty closely
<headius>
Array.new() does something different I believe
yfeldblum has joined #jruby
<headius>
but it may be hitting the new coercion logic
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/master 21ccb97 Chris Seaton: [Truffle] Workaround for bug in 0.5 Truffle splitter/inliner.
JRubyGithub has left #jruby [#jruby]
<JRubyGithub>
[jruby] jrubyci pushed 1 new commit to master: http://git.io/XPOoJA
<headius>
Array.new takes a number
vyorkin1 has joined #jruby
<headius>
bbrowning: did you mean Array(format)
<bbrowning>
err crap, yeah
<headius>
enebo: I don't know if you saw...I added some "isBooting" logic to JIT on master to not JIT until we've prepped the Ruby instance
<headius>
might be worth doing in 1.7 as well to avoid jit overhead during boot
<enebo>
headius: yeah
<enebo>
headius: eventually it would be nice if that became rate over time perhaps since we might be able to generalize that
nanoyak has joined #jruby
<bbrowning>
headius: so what is Array(...)? that's not a constructor - Array is a named method on Kernel?
<enebo>
headius: I am pretty sure rate of method addition stays nearly constant during boot
nanoyak has quit [Read error: Connection reset by peer]
nirvdrum has joined #jruby
<enebo>
headius: his broken test is exactly the method you changed
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/truffle-head 7be5b73 Chris Seaton: Merge branch 'master' into truffle-head...
<JRubyGithub>
[jruby] jrubyci pushed 2 new commits to truffle-head: http://git.io/rmgRMA
<JRubyGithub>
jruby/truffle-head a48af12 Chris Seaton: Merge branch 'master' into truffle-head
JRubyGithub has left #jruby [#jruby]
nanoyak has joined #jruby
subbu has quit [Quit: Ex-Chat]
subbu has joined #jruby
<bbrowning>
yeah I'm tracing the logic now and it seems like it shouldn't be calling to_ary if to_ary isn't defined
<bbrowning>
but yet it's calling it
subbu is now known as Guest22493
phrinx has joined #jruby
<enebo>
bbrowning: unless we are messing up and not adding a to_ary to Mime::Type :)
<enebo>
lunch!
johnsonch is now known as johnsonch_afk
<bbrowning>
hey this looks interesting - ruby's Mime::Type class declares private empty methods to_ary and to_a
Guest22493 has quit [Ping timeout: 272 seconds]
vyorkin1 has quit [Ping timeout: 240 seconds]
yarker has quit [Remote host closed the connection]
zorak8 has joined #jruby
colinsurprenant has quit [Quit: colinsurprenant]
<bbrowning>
so is the fact that they are defined but private the key? whatever checks to see if they're defined passes but then the call fails because private?
<enebo>
nirvdrum: ok. I definitely want us to match MRI so for 1.7 it would be a temporary revert until the release is out but for master it would not be reverted at all
<nirvdrum>
Maybe not the same thing then.
wojtczak has joined #jruby
calavera_ has joined #jruby
<headius>
I did have a commit that was calling to_ary blindly, but I fixed that
yfeldblum has joined #jruby
<enebo>
rb_method_call_status seems to allow private and protected calls is not marked CALL_PUBLIC
<enebo>
our method does not have this extra field to say what caller is
<enebo>
That feels a bit invasive for a release today or tomorrow
<enebo>
but fixing the private invocation issue I don’t think is the primary issue but it seems to be a difference
yfeldblum has quit [Ping timeout: 244 seconds]
<enebo>
bbrowning: I am going to revert this commit only for this release and we will reapply after release and fix whatever is broken
<bbrowning>
enebo: kk - as far as I can tell, it would impact most rails 4 apps as-is
skade has joined #jruby
skade has quit [Client Quit]
<wojtczak>
how can I change Dir.tmpdir default /tmp to something else? I tried setting env TMPDIR, TMP, adding -Djava.io.tmpdir option with no luck
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/jruby-1_7 6ba08e6 Thomas E. Enebo: Revert only temporarily because this breaks something in Rails 4 involing Array. Figure out for 1.7.17
<JRubyGithub>
[jruby] jrubyci pushed 1 new commit to jruby-1_7: http://git.io/QnLpRw
JRubyGithub has left #jruby [#jruby]
<enebo>
bbrowning: can you retest again with that revert in place?
nanoyak has quit [Read error: Connection reset by peer]
nanoyak has joined #jruby
<chrisseaton>
a few weeks ago someone disabled a test for me as Truffle was breaking Jetty - but now I can't remember who that was to tell them to turn it back on - does this remind anyone?
<chrisseaton>
or rather Jetty was breaking itself when Truffle happened to be there
JohnBat26 has joined #jruby
yfeldblum has joined #jruby
asuka has joined #jruby
calavera_ has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
nanoyak has quit [Quit: Computer has gone to sleep.]
yfeldblum has quit [Ping timeout: 272 seconds]
drbobbeaty has quit [Read error: Connection reset by peer]
dumdedum has quit [Quit: foo]
drbobbeaty_ has joined #jruby
havenwood has joined #jruby
<bbrowning>
enebo: everything passed after the revert
<enebo>
bbrowning: cool. thanks for testing
diegoviola has quit [Read error: Connection reset by peer]
<headius>
enebo: ok I lied, I'll have basic working non-indy in a few minutes
<headius>
simplest possible impls...no caching of anything
<headius>
super needs to be filled in but the rest is there
<enebo>
coolio
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/master 1d38f42 Chris Seaton: [Truffle] Update to 0.5.
<JRubyGithub>
[jruby] jrubyci pushed 2 new commits to master: http://git.io/6eZttA
<JRubyGithub>
jruby/master cab2272 Chris Seaton: Merge branch 'truffle-head'
JRubyGithub has left #jruby [#jruby]
<headius>
the call sites are a big scam...I am generating a method per as we discussed, so I just generate them with the literal args and then do callMethod
<headius>
when we actually optimize the non-indy version that will change to a lookup + cache + DynamicMethod.call
<enebo>
headius: well working is a good first swipe
<headius>
yeah
<enebo>
headius: non-indy vms can run 9000
<headius>
they can!
<enebo>
well I guess they could before too but hey it will be faster now :)
<headius>
hah yes
benlovell has joined #jruby
benlovell has quit [Client Quit]
nanoyak has joined #jruby
Aethenelle has joined #jruby
nanoyak has quit [Client Quit]
nirvdrum has quit [Ping timeout: 244 seconds]
nanoyak has joined #jruby
lanceball is now known as lance|afk
toady00 has quit [Remote host closed the connection]
Guest22493 has joined #jruby
toady00 has joined #jruby
<cpuguy83>
headius: You mentioned there are some extra flags for 9k?
<headius>
just temporarily to enable JIT, since it doesn't work in all modes
<headius>
that will change shortly
iamjarvo has joined #jruby
<cpuguy83>
Is that documented somewhere?
nirvdrum has joined #jruby
e_dub has quit [Quit: ZZZzzz…]
toady00 has quit [Ping timeout: 245 seconds]
wojtczak has quit [Quit: Leaving...]
drbobbeaty has quit [Ping timeout: 272 seconds]
<headius>
the flags are all documented but not specifically for 0k
<headius>
9k
<headius>
compile.invokedynamic=true, since we default it to off and JIT only works with indy right now
<headius>
and then the rest are the normal JIT flags for threshold, etc
<cpuguy83>
excellent, thanks.
yfeldblum has joined #jruby
yfeldblum has quit [Remote host closed the connection]
yfeldblum has joined #jruby
Guest22493 is now known as subbu
JRubyGithub has joined #jruby
<JRubyGithub>
jruby/1.7.16 575b395 Thomas E. Enebo: Bump for release
<nirvdrum>
Yes, but you need to use their hostname or you'll get a certificate mismatch.
<enebo>
nirvdrum: ok well we can do that but then you are also saying we should make download page itself as https
<enebo>
sorry two people :)
<mpapis>
:)
<nirvdrum>
Yeah. In that case, you'd probably need to switch to CloudFront hosting. Then your options are SNI-only SSL for free or non-SNI SSL for $600 / year (IIRC)
<enebo>
mpapis: that line to nirvdrum was directed at you
<nirvdrum>
enebo: Any idea what the original TTL was?
<enebo>
nirvdrum: yeah I thought I changed it in Japan but the UI sucks ass and I had to hit a second save button
elia has joined #jruby
<enebo>
1hour I think
<enebo>
nirvdrum: but headius is also correcting how our ci job saves so it puts it in a location that lines up with what rbenv expects
<nirvdrum>
Okay. I guess I'll just drum my fingers for another 55 minutes.
<headius>
Aethenelle: I think this is a coercion problem
<enebo>
nirvdrum: so we have a few things to iron out
<headius>
like, MRI may to_h but then treat it like a hash and ignore default
<headius>
or something
<nirvdrum>
enebo: No worries. I've been manually building, but now that I'm trying to get others to fix things in 9k, it's helpful if they can actually install it.
<headius>
we can get a raw backtrace and find the exact line where it's calling default, and compare it to what MRI side does
<enebo>
nirvdrum: well good news is we are getting close to having this back in order…sorry it has taken so long
<Aethenelle>
mri just sends it to the internal hash
<Aethenelle>
I'll poke through it on the way home tonight
<headius>
yeah, it's something minor
<headius>
christ our dist builds take a long time
<headius>
at least on an m1.small CI machine
JRubyGithub has joined #jruby
<JRubyGithub>
[jruby] jrubyci pushed 1 new commit to jruby-1_7: http://git.io/ub26cQ
<JRubyGithub>
jruby/jruby-1_7 aafb40b Charles Oliver Nutter: Fix coercion of arrays in Kernel#Array and Array#product....
JRubyGithub has left #jruby [#jruby]
<headius>
enebo: oops
<enebo>
headius: bbrowning: ^ I added this back onto jruby-1_7 so any reproduction on Array(MimeThingy) would be helpful
<enebo>
headius: I did that :)
<headius>
come on cloudbees...you can do it
<Aethenelle>
headius: I think i got it... org.jruby.ext.ffi.Enums < org.jruby.RubyHash but FFI::Enums < Object
<enebo>
headius: I am deleting those top-level saved files in S3
<headius>
Aethenelle: ahh, there ya go
<headius>
hierarchies don't match
<Aethenelle>
I'm tempted to make Enums stop inheriting from RubyHash, doesn't really get us anything anyway I think...
<Aethenelle>
also makes it easier for supporting it when I get the MRI version to C...
drbobbeaty has joined #jruby
<headius>
yeah that might be simplest
<headius>
so it was RubyHash re-dispatching to #default, but default wasn't defined because of the hierarchy mismatch
<Aethenelle>
k...should have a PR by the time i get home...
<Aethenelle>
yeah, unless there's some annotation magic somewhere I don't see, there's no way for the ruby side of things to know Enums is a Hash under the covers.
johnsonch is now known as johnsonch_afk
CaptainJet has joined #jruby
<headius>
nah, there's nothing special about that...the metaclass still has to be present for dispatch
<Aethenelle>
k
<Aethenelle>
reminds me... you want that last prepend test to match current MRI or future MRI? (updating included classes not showing up without re-including them)
<Aethenelle>
not sure which is better... fixing MRI or unfixing my code...
postmodern has joined #jruby
CaptainJet has quit [Client Quit]
Xzyx987X_ has quit [Quit: Leaving]
enebo has quit [Quit: enebo]
colinsurprenant has quit [Quit: colinsurprenant]
Aethenelle has quit [Quit: Aethenelle]
jaffachief has quit [Quit: Computer has gone to sleep.]
DomKM has joined #jruby
Specialist has quit [Read error: Connection reset by peer]
havenwood has quit [Remote host closed the connection]
drbobbeaty has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]