<riddley>
ok I've got to bail. I'll try to stop in again tomorrow to see if you guys can get to the bottom of it. I realize none of you are here, but thank you so much for the time you spent looking at it already.
<GitHub85>
[jruby] chrisseaton pushed 3 new commits to truffle-head: https://git.io/v1aiR
<GitHub85>
jruby/truffle-head 5a948d5 Chris Seaton: [Truffle] Tidy up dependencies.
<GitHub85>
jruby/truffle-head 17084b7 Chris Seaton: [Truffle] We don't need the classic jar any more.
<GitHub85>
jruby/truffle-head 7313cc3 Chris Seaton: [Truffle] Always run our main.
alxs has quit [Ping timeout: 244 seconds]
_whitelogger has joined #jruby
<GitHub198>
[jruby] chrisseaton created truffle-mx-no-core (+1 new commit): https://git.io/v1aMe
<GitHub198>
jruby/truffle-mx-no-core 2b56607 Chris Seaton: [Truffle] Attempt at removing core from mx.
<GitHub27>
[jruby] chrisseaton opened pull request #4371: [Truffle] Attempt at removing core from mx. (truffle-head...truffle-mx-no-core) https://git.io/v1aMT
<chrisseaton>
headius: right - what reason is there ever to modify the case of text?
<headius>
well one case way back then was for capitalizing or de-capping table/class names for ActiveRecord
<headius>
but they're all pretty contrived
<headius>
I dunno...people still bitch about bad casefolding to this day
<chrisseaton>
case folding language identifiers is fine - that can be defined by the language spec
<chrisseaton>
Case folding people's names and things is crazy
<headius>
yeah the problem is that you can't reliable case fold a lot of languages
<headius>
reliably
<headius>
within a given country you might have multiple ways to capitalize some letter
<headius>
I think Turkey was an example Tim gave back then...don't remember the letter
<headius>
in any case...it shouldn't be hard to port this translation table into jcodings
<headius>
we need to divorce ourself from MRI's C source real soon though
<lopex>
why Turkey ?
<headius>
I don't know
<headius>
Turkey or Hungary, I am not sure
<lopex>
chrisseaton: folding is crazy
<lopex>
the problem is who cares ?
<lopex>
chrisseaton: like jruby always distributes all foldings
<GitHub139>
[jruby] chrisseaton pushed 1 new commit to truffle-head: https://git.io/v1VfK
<GitHub139>
jruby/truffle-head 3113f09 Chris Seaton: [Truffle] Restore some command line functionality I fudged.
<headius>
lopex: yeah I'm not looking forward to shipping more gigantic tables
<headius>
and there's no way this will init in 64k of bytecode
<headius>
Java needs constant pool entries for arrays of constants
<headius>
so badly
<lopex>
headius: the problem is not to ship the rarest things
<headius>
I suppose that's one way to fake it...force the whole blasted thing into one byte[] as a string and process it
<headius>
I agree
<headius>
well actually no...the problem is not witholding the rare bits. The problem is making sure people who need the rare bits don't have a terrible experience
<lopex>
but is there a lazy machanizm to load those reliably ?
<headius>
maybe something in Jigsaw as you mentioned
<lopex>
haha
<headius>
we could always have gems that install stuff
<headius>
but that's still a bad experience
<headius>
I dunno
<lopex>
no idea
<headius>
you were saying Big5 tables could be pulled?
<lopex>
yeah
<lopex>
well I hoped to
<headius>
but that's Taiwan, Hong Kong, etc for traditional chars
<headius>
every encoding you say we could pull, I know people personally that would be affected
<headius>
which sucks :-(
<lopex>
yeah
<headius>
I wish the encoding stuff at the JVM level were a little easier to reuse, because a lot of that is there
<lopex>
that's why I never had opinion on that
<headius>
hmmm
<headius>
I do still have my CharBuffer-based transcoder
<lopex>
wchich whould change whate ?
<lopex>
chrisseaton: but you need to mention how bad 2.4 folding is implemented
<GitHub46>
jruby/master d26784d Thomas E. Enebo: Fixes #4348. File.open with open mode a+ do not work (windows or pure-Java mode)
<GitHub172>
[jruby] enebo closed issue #4348: File.open with open mode a+ do not work https://git.io/v1Gzk
<GitHub31>
[jruby] eregon opened pull request #4373: [Truffle] Integrate the null Array strategy with other strategies. (truffle-head...truffle-null-array) https://git.io/v1Vys
atambo has quit []
atambo has joined #jruby
<eregon>
ebarrett: The Truffle version override 9-Dtruffle.version=0.18) is needed, because in the source it was some different snapshot version, and when building GraalVM 0.18, Truffle 0.19 was enforced, no matter what the sources say.
<eregon>
ebarrett: I see, hope it works well for you then
vtunka has joined #jruby
donValentin has joined #jruby
<ebarrett>
hi
<ebarrett>
let me show you what I ended up doing (although truffle still isn't enabled in jruby when I try to run it)
<ebarrett>
this is a debug trace from our benchmark runner
<ebarrett>
you can see the jruby invocation, and jrucy telling us that there is no JIT
<Aethenelle>
anyone know why rubygems doesn't handle the platforms: options the same as bundler?
<ebarrett>
the benchmark runner then crashes as it attempts to parse the error message as json ;)
<Aethenelle>
riddley: you get things solved?
<ebarrett>
eregon: thanks for your help :)
<Aethenelle>
sorry I wasn't back on yesterday. Ran into an old friend on the train.
<eregon>
ebarrett: it seems you're missing some flags for running Graal
<eregon>
I only see -J-Djvmci.Compiler=graal
<headius>
Aethenelle: still failing
<headius>
he said he'd be around today to keep investigating
<Aethenelle>
bugger...
<Aethenelle>
we have a test case that works like his use case yet?
riddley2 has joined #jruby
<riddley2>
Aethenelle and headius I don't think you saw me say thank you last night. Thanks very much for the time you spent with my issue. If you have any time today, I'm happy to do whatever I can to get it resolved.
<riddley2>
(the Kafka UTF-8) thing
<headius>
riddley2: we were just chatting about it :-)
<riddley2>
perfect timing!
<Aethenelle>
i have time, but not having a test case we can do locally is going to start meaning some less stable test cases to go into your prod env...
<riddley2>
So, the Kafka portion is in production and is being used by various groups in my org, but my use of it is not yet in production because of this issue. All of the other groups use Perl and are working fine, so I don't think we'll have an issue if we just send data through the pipeline
donV has joined #jruby
<Aethenelle>
that's good to hear
<Aethenelle>
... sort of...
<riddley2>
hey I've connected in a stupid way... let me reconnect from a different host... brb
<Aethenelle>
good to know the test cases aren't running in prod
riddley2 has quit [Client Quit]
riddley2 has joined #jruby
<riddley2>
ok, back, sorry about that
<Aethenelle>
it's all good... you using a Gemfile for your dependencies?
<riddley2>
I am
<riddley2>
would you like me to paste it somewhere?
<Aethenelle>
no, just wanted to know
<Aethenelle>
will help if we need to try doing stuff in the gem itself...
<riddley2>
ok cool
<riddley2>
I don't know if you saw what I wrote last night
<riddley2>
but the "hangs" I was seeing were just exceptions somehow not making it up the chain
<riddley2>
so when I ran jruby -d I started seeing them
<GitHub77>
jruby/master 125315f Thomas E. Enebo: Some conversion of obvious FileChannel which can be made into more specific Channel interfaces
<riddley2>
I tried every version of what you guys suggested and most of them just acted like #force_encoding('UTF-8') in that they worked, but didn't fix the issue
<riddley2>
message = java.lang.String.new(message, "utf-8") gave me TypeError: no implicit conversion of Java::JavaLang::String into String
<GitHub113>
[jruby] bjfish pushed 1 new commit to truffle-head: https://git.io/v1Vbj
<Aethenelle>
headius: could something be lost in translation due to the thread creation
<riddley2>
not sure if you were around when I offered to do a webex, but that's an option if it'll help
olle has quit [Quit: olle]
<GitHub190>
[jruby] eregon pushed 1 new commit to truffle-head: https://git.io/v1Vp0
<GitHub190>
jruby/truffle-head 4db07f2 Benoit Daloze: [Truffle] Verbosity is false by default, not nil.
<Aethenelle>
riddley2: is this code in the github issue everything for your kafka stuff? (i.e., if I have a kafka instance and this file it'll work if the worker reference is removed)
<enebo>
yeah weird. I do not see you in the membership list
<Cris_Shupp>
You are an approver right?
<enebo>
Cris_Shupp: yeah I can add you manually … whatever is happening with subscribing for you seems to not be working I guess
<Cris_Shupp>
I did it again:
<Cris_Shupp>
Your subscription request has been received, and will soon be acted upon. Depending on the configuration of this mailing list, your subscription request may have to be first confirmed by you via email, or approved by the list moderator. If confirmation is required, you will soon get a confirmation email which contains further instructions.
<Cris_Shupp>
i am Cris Shupp, cshupp@gmail.com
<Cris_Shupp>
thanks
<enebo>
Cris_Shupp: normally we do not manually approve subscriptions so something is off
<enebo>
Cris_Shupp: added. I will also toggle digest?
<Cris_Shupp>
Yes please
<GitHub122>
[jruby] chrisseaton force-pushed truffle-bytelist from e2a4664 to 3180e2b: https://git.io/v1Vk5
<GitHub122>
jruby/truffle-bytelist 3783b09 Chris Seaton: [Truffle] Don't cache ByteList as a String - we never do that on the fast path.
<GitHub122>
jruby/truffle-bytelist 24f68f9 Chris Seaton: [Truffle] Remove immediately dead code in ByteList.
<GitHub122>
jruby/truffle-bytelist 47206bd Chris Seaton: [Truffle] Privatise ByteList.
<enebo>
Cris_Shupp: done. you will get digest and you should have gotten a welcome email
<enebo>
Cris_Shupp: sorry I am confused why you were not just subscribed but sometimes weird stuff happens
<enebo>
Cris_Shupp: unless you got a confirm email sent to spam or something
<Cris_Shupp>
Thanks. No welcome email yet
<Cris_Shupp>
Ahaha
<Cris_Shupp>
Gmail is moving you to 'promotions' but not spam
<enebo>
Cris_Shupp: hahah…you might have already won!
<Aethenelle>
just got off a call... I've ben able to get stuff going through kafka and see it in the consumer output
<Aethenelle>
not getting any errors yet...
<riddley2>
oh no
<riddley2>
so it's something particular to my environment?
<Aethenelle>
maybe... not sure yet
<riddley2>
and you're sending stuff like Sül
<Aethenelle>
there's no call to exists? in the code you've provided... are you actually seeing the error at that point or in the sidekiq worker?
<riddley2>
both. I started not seeing any jobs for messages with multibyte characters and the first thing in my worker is a guard that looks to see if the path it was handed exists. I did a puts on the message and the chars were screwy, so I did a puts on the message before I handed it to the worker and saw the same thing.
djellemah_ has joined #jruby
<riddley2>
could this have anything to do with shell variables like LC_* ?
djellemah_ has quit [Ping timeout: 250 seconds]
<Aethenelle>
the testing we've done so far... that was in the code you gave me (and any associated prints) or were you watching from the worker?
<Aethenelle>
what's the worker running on jruby?
<Aethenelle>
riddley2: it's possible... what are they set to?