shellac has quit [Quit: Computer has gone to sleep.]
<GitHub197>
[jruby] perlun opened issue #5145: Issue with prefixing "system" commands with ENV variables in Dir.chdir https://git.io/vp3Jk
guilleiguaran has quit [Quit: Connection closed for inactivity]
mfunkmann has joined #jruby
<mfunkmann>
Hello folks! I am using Warbler to create a JAR file out of my JRuby program. I am using some command line arguments like -J-splash to display a splash image. Is it possible to somehow use these inside of a JAR?
AnandJambhulkar has quit [Ping timeout: 256 seconds]
shellac_ has quit [Client Quit]
AnandJambhulkar has joined #jruby
AnandJambhulkar has quit [Quit: This computer has gone to sleep]
AnandJambhulkar has joined #jruby
AnandJambhulkar has quit [Quit: This computer has gone to sleep]
AnandJambhulkar has joined #jruby
bbrowning_away is now known as bbrowning
shellac_ has joined #jruby
AnandJambhulkar has quit [Quit: This computer has gone to sleep]
shellac_ has quit [Quit: Computer has gone to sleep.]
<kares>
mfunkmann: would try specifying as a resource - similar to how Java folks do (if they do - since most apps get unpacked and run from FS)
<kares>
e.g. classpath://... (it might not work)
<GitHub60>
jruby/handle_exception_cause_with_null_message 4009259 Uwe Kubosch: Handle exception cause with null message
<GitHub60>
[jruby] donv created handle_exception_cause_with_null_message (+1 new commit): https://git.io/vp3Xa
<GitHub85>
[jruby] donv opened pull request #5146: Handle exception cause with null message (master...handle_exception_cause_with_null_message) https://git.io/vp31l
<headius>
looks like bdortch started the removal of etend_proxy
<GitHub168>
[jruby.github.io] enebo pushed 1 new commit to master: https://git.io/vpsfQ
<GitHub168>
jruby.github.io/master 372c383 Thomas E. Enebo: Update for 9.1.17.0
<kares>
your memory is in pretty good shape - going back 10+ years :)
<headius>
yeah, so it would attach this "proxy extender" to the name of the interface, and then when looading the actual interface it would find them and apply them
<headius>
kares: I've been reminded of it a few other times when thinking about startup time
<kares>
guess JRuby is good at Alzheimer's prevention :)
<headius>
FWIW I'm on board with anything we can make lazy as long as it's concurrency-safe
<headius>
JI is a big part of startup because it's all eager now
<kares>
it should be
<kares>
since you made proxy-class concurrency safe
<kares>
and I am planning to initialize right where that happens
<headius>
well it's more a matter of two users both accessing a proxy-extended type at the same time
<headius>
they can't see a partially initialized view
<kares>
TL:DR: load exts as the proxy class/module is generated
<headius>
dortch had a much more complicated locking mechanism around lazy Java class proxies
<kares>
they can not - as far as I understand proxy init
<headius>
that went away as the proxy generation got simpler
<headius>
right, it will block now
<kares>
also I wonder - do we not switch to class-values?
<kares>
they somehow got disabled and the map fallback impl is used atm
<kares>
I think
<headius>
we are using class values for some things but they currently are not "weak" like we want them to be
<kares>
... do not have JRuby open right now
<headius>
they don't work like ThreadLocal, where if the thread goes away the value goes away
<headius>
so they anchor the class
<headius>
and the value
<headius>
I have argued this is wrong
<kares>
hmm but for initializing proxy class for a java class
<kares>
we do not care for that do we?
<headius>
well we need to hold the ClassValue object, and that will hold the Java class forever
<kares>
aaah
<headius>
instead of a proper "class-related value" utility all we got was a glorified hash
<kares>
got it
<kares>
that seems like it nails its usability
<kares>
wonder if they 'fixed' that in 9
<headius>
I think Groovy eventually chose to use ClassValue+weak ref
<ChrisBr>
I'm not 100% sure about the fix because it always uses now the extension of the link and the the extension of the expanded path gets always ignored
<headius>
yeah I don't know how else to explain that failure though
<headius>
as far as I know MRI doesn't do any other content-guessing...I think it might be correct that if you request .rb in the require it loads a .rb regardless of what it resolves to
<headius>
last time I tried to read through all that code was maybe 2.0 or 2.1 timeframe when the started adding absolute paths
<ChrisBr>
hmm
<headius>
but even then it had an early short-circuit if your require ended with .so or .rb
<ChrisBr>
didn't check with mri actually
<headius>
it just went directly into the .so or .rb loading logic
<headius>
oh well actually we should confirm with a simple test of MRI
<headius>
I forgot the failing test is ours
<headius>
should be easy enough to symlink a .so to something without an extension and see what MRI does with it
<ChrisBr>
k, will do that and comment on the PR
<ChrisBr>
btw: what is the state of 2.5 support?
<headius>
I'm hacking away at it right now, but specs are down to 5 and MRI suite is down to like 90
<ChrisBr>
great
<headius>
there's a ton of minor string behavior changes still failing that might be easy to sort out
<headius>
like 20 failures or something
claudiuinberlin has joined #jruby
<headius>
little things like chomping gets and whatnot
<headius>
enebo: hey you
<headius>
how's bytelist love looking
<enebo>
heh
<enebo>
I think ok
<enebo>
I just caused a few errors trying to consolidate something but this is all very minor
<enebo>
largely just making all error messages be m17n