Puffball has quit [Quit: Puffball]
rdubya has quit [Ping timeout: 264 seconds]
drbobbeaty has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
shellac has joined #jruby
drbobbeaty has joined #jruby
rdubya has joined #jruby
jmalves_ has joined #jruby
jmalves_ has quit [Remote host closed the connection]
jmalves has joined #jruby
jmalves_ has joined #jruby
jmalves has quit [Ping timeout: 245 seconds]
jmalves_ has quit [Ping timeout: 252 seconds]
sgeorge has joined #jruby
sgeorge has quit [Remote host closed the connection]
sgeorge has joined #jruby
<headius> lopex: hey what are you seeing?
<headius> hmm
<headius> I only released jcodings so far
<headius> I'll check locally...thought I ran with the final jcodings
<headius> lopex: building jruby master on 10 seems to be fine here
<headius> I can reproduce "u" error
sgeorge has quit [Remote host closed the connection]
sgeorge has joined #jruby
<headius> lopex: I get it with previous jcodings
<headius> I think it's the bytelist move
<headius> yeah
<headius> oy vey
sgeorge has quit [Remote host closed the connection]
sgeorge has joined #jruby
<headius> enebo: ah I remember running into this before
<headius> I think it's some logic I "fixed" to make ByteList.length conform to CharSequence and accurately represent chars
<enebo> headius: yeah I think I also hit this at some point and started using realSize()
<headius> I think I started on a patch to fix anywhere in JRuby that called length but there's still a ton of them
<headius> return filenameBytes.length() == 1 && filenameBytes.get(0) == '-';
<headius> etc
<enebo> zillions in native extensions if they work on bytelists too...although that may not be a ton
<enebo> tno of extensions but the ones which do probably call it a lot
<enebo> headius: although generally unless it is utf-16le or something like that it will still work
<enebo> not perfect but a lurking killer
<headius> well this was hit #2 out of 189
<headius> to be fair this code would be broken under utf16 even if it were using byte length since it probably should be checking the first *character*
<enebo> yeah
<enebo> be would work though :P
<enebo> 16-be that is
<enebo> paths are difficult in our codebase because they need to end up as String/Charset and behave
<enebo> so generally by the time we do stuff like this we usually are working with utf-8 or ascii
<enebo> although fuzzing ruby string data for paths would probably uncover some interesting bugs
<headius> well with bytelist folding back into JRuby I'm contemplating just rolling back that patch
<headius> the library is dead if we want to work properly in 9 and this change was never released anywhere
<enebo> yeah makes sense to me. I suspect nokogiri may be a consumer of bytelist and might use length
<enebo> funny though we almost always use realSize internally
<headius> yeah I have for a long time
<headius> these could all be audited but that won't help exts
<headius> there's a path forward though, since realSize() is there...all current uses of length() in exts would likely be .realSize()
<enebo> yeah assuming there are any length usages. I assume so since it is the more obvious choice without knowing the difference
<headius> well things look fine with that commit reverted
<headius> I paused the work of getting this reintegrated because you were finishing bytelist identifiers
<headius> enebo: I'm going to push this...ByteList pre-2.0 is part of JRuby 9.2 API so this change is too invasive anyway.
<enebo> ok
<headius> yeah, seems ok now
xardion has quit [Remote host closed the connection]
<lopex> headius: the first npe during build was fixed by surefire plugin update
<lopex> I didnt investigate the second since I wanted to work on joni for that sunday search thing
<headius> perhaps scanning for tests
<headius> tripped over the module-info
<headius> someone mentioned testing, maybe it was even surefire
<lopex> headius: but I had a clean copy of jruby for that 'u' thing
<headius> surefire may be running against the non-shaded artifact
<headius> yeah it was the length change
<lopex> headius: moved most joni constants into constants.internal
<lopex> the ones we dont want to expose
<headius> right, sounds good to me
<headius> I hope we had no references to them in JRuby?
<lopex> jcodings is a bit trickier, everything is more interconnected
<lopex> headius: well, I there wont
<lopex> headius: we we'd also like to hide jcodings.util
<lopex> headius: there is few references to IntHash though
<headius> hah
<lopex> should be changed to jruby internal collections IntHash
<headius> yeah
xardion has joined #jruby
<headius> there's only a couple it seems
<lopex> headius: use from core in String in #tr methods mostly
<lopex> headius: it still can be useful though since it can hash on byte, beg, len pair
<lopex> er, tripple
<headius> ah, util.CaseInsensitiveBytesHash is exposed as part of Encoding's API
<lopex> yeah, whole packake
<lopex> there's threee concrete hashes
<headius> along with that util.Hash{.HashEntryIterator}
<lopex> it's specialized fast lookup, so maybe we still need that
<lopex> case insensitive
<lopex> well, we need it in EncodingSupport afaik
<headius> Hash could move somewhere public and the Encoding API modified to return it
<headius> none of these changes are backward compat obviously
<lopex> enebo: sunday search seems to pass joni tests, but there's tow new case insensitive versions for that too
<lopex> *two
iceman_22_ has quit [Quit: Connection closed for inactivity]
<enebo> lopex: you before thought it might not yield much in the way of results
rdubya has quit [Ping timeout: 252 seconds]
subbu is now known as subbu|away
<lopex> enebo: yes
shellac has quit [Read error: Connection reset by peer]
rdubya has joined #jruby
jmalves has joined #jruby
jmalves has quit [Ping timeout: 246 seconds]
jmalves has joined #jruby
<lopex> enebo: headius: but what caught my thought is that graal was now surprisingly stellar for that too
<lopex> which means graal suffers from the same bigs switch issues ?
<headius> stellar for what exactly?
<lopex> headius: these additional regexp examples for that "match?" issue
<lopex> enebo: benchmarked it on graal too
<headius> stellar with the switch or with the new code?
<lopex> with old switch
<lopex> I said it was not stellar
<lopex> well, it might be the case that jruby internals swamp most of it anyways
<headius> ah yeah hard to know
<headius> I don't know either way whether graal is better at big switches
jmalves has quit [Ping timeout: 240 seconds]
rdubya has quit [Ping timeout: 250 seconds]
drbobbeaty has quit [Ping timeout: 246 seconds]
rdubya has joined #jruby
shellac has joined #jruby
drbobbeaty has joined #jruby
<lopex> this is insane how much do they copy code in mri
jmalves has joined #jruby
jmalves has quit [Ping timeout: 244 seconds]
shellac has quit [Ping timeout: 244 seconds]
sgeorge has quit [Ping timeout: 245 seconds]
<headius> I wonder if it's a C thing
<headius> enebo noticed it many times in the oj ext too
<headius> is it some fear of making another function?
shellac has joined #jruby
<headius> could force inline them if that were it so I don't understand
<lopex> what I do now is mostly deduping
<lopex> and it gets kind of annoying
<lopex> all the time
drbobbeaty has quit [Ping timeout: 244 seconds]
<headius> yeah I've noticed that too
<headius> fear of declaring another variable?
<lopex> headius: they copy all sets of functions and ifdef all of it
<lopex> in out case it's still free since that setting is static final
<lopex> *our
<headius> yeah I have typically done a pass like this whenever porting something
<headius> make the code a little more sane to follow
<lopex> too bad we have to dup some code since one operates on byte[] and the other on int[]
<headius> ugh, I keep playing whack-a-mole with these spurious failures on travis
<headius> most of them I'm not sure if it's a bug or just travis being slow or otherwise weird
sgeorge has joined #jruby
subbu|away is now known as subbu
jmalves has joined #jruby
subbu is now known as subbu|lunch
jmalves has quit [Ping timeout: 252 seconds]
<enebo> headius: in oj's case though some portion is just a tiny change in what the copy has
<enebo> personally I don't think that tiny change is worth whatever gain it gives but it is at least a rational explanation for it
<enebo> although not all of it is. I half wonder if in those cases it is not trusting inline
shellac has quit [Ping timeout: 252 seconds]
shellac has joined #jruby
jmalves has joined #jruby
jmalves has quit [Ping timeout: 240 seconds]
sgeorge has quit [Read error: Connection reset by peer]
sgeorge_ has joined #jruby
subbu|lunch is now known as subbu
<lopex> headius: I'm logged on sonatype.com whast should I do next to apply for release right ?
<lopex> dont want to poke enebo everytime
shellac has quit [Quit: Computer has gone to sleep.]
<enebo> lopex: from what I recall there is a jira project where you request and you submit a public key
<lopex> yes, already there
<lopex> oh I remember why I got so discouraged
sgeorge_ has quit [Remote host closed the connection]
shellac has joined #jruby
sgeorge has joined #jruby
shellac has quit [Ping timeout: 250 seconds]
shellac has joined #jruby
shellac has quit [Quit: Computer has gone to sleep.]
shellac has joined #jruby
rosenjcb has joined #jruby
<rosenjcb> I have this annoying error I keep getting when I run jruby -S gem install jbundle:
<rosenjcb> It seems your ruby installation is missing psych (for YAML output).
<rosenjcb> LoadError: load error: psych -- java.lang.NoClassDefFoundError: Could not initialize class org.jruby.ext.psych.PsychEmitter
<rosenjcb> To eliminate this warning, please install libyaml and reinstall your ruby.
<rosenjcb> I have libyaml-dev installed and I reinstalled both ruby and jruby
<headius> hmmm
<rosenjcb> I think it's a local issue and not a bug necessarily because it runs fine on my personal computer
<headius> well that almost certainly is due to the JRuby version and a Psych version conflicting
<headius> we don't use libyaml
<headius> the libyaml warning should probably be fixed in Psych since it doesn't apply to JRuby
<rosenjcb> That would make sense.
<headius> how did you install JRuby initially?
<rosenjcb> I think it was automatically installed
<rosenjcb> But I uninstalled it and reinstalled it
<rosenjcb> Maybe I didn't completely uninstall it?
<rosenjcb> So I think maybe my jruby is an older version
<rosenjcb> I'm running 9.1.13.0
<rosenjcb> ruby itself is 2.5.1p57
<rosenjcb> When I type which jruby I get /usr/bin/jruby but I cannot find the directory
shellac has quit [Ping timeout: 240 seconds]
<rosenjcb> headius, I tried installing via rebenv but I'm not getting anywhere with it.
<havenwood> rosenjcb: After setting `rbenv local jruby`, and running `rbenv rehash`, what do you get for?: which jruby
<headius> rosenjcb: automatically installed by your host OS? I'm confused
<headius> rbenv should also work but it's a different process obviously
<headius> if you installed (or had installed for you) JRuby from a Linux package (Fedora or Ubuntu or something) it could mean there's something wrong with that package
<rosenjcb> I resolved it by upgrading to 9.2.0.0
rosenjcb has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
<headius> well, I guess that's that
jmalves has joined #jruby
jmalves has quit [Ping timeout: 246 seconds]
<headius> kares: hey what's this whole load service + bootsnap + rails + whatever issue I'm hearing about?
<headius> I remember one or two issues but I'm not clear where we stand right now
<headius> ok yeah, this is the extension issue
<headius> our internal extensions don't work right with bootsnap because they are not files on disk
<enebo> oh yeah I rememebr this
<headius> in my opinion this is a bootsnap bug since it will also fail for totally value extensions inside jar files
<headius> it should have a fallback that continues to do the long search whenever the loaded resource does not resolve to a filesystem artifact
<headius> totally valid I meant
<headius> not that our internal exts aren't valid but they're weird
<headius> special case
<lopex> enebo: sunday has landed
<lopex> though it's pretty invasive since there's additional heuristics that choose the algos
<enebo> lopex: now start on monday please
jmalves has joined #jruby
jmalves has quit [Ping timeout: 272 seconds]