<GitHub193>
jruby/master f83a55f kares: review the old RubyDateFormat - no 1.9 lef-overs even is deprecated code
<GitHub193>
jruby/master 3bf9d77 kares: align RubyArray's values_at method with its RubyStruct copy-pasta
olle has quit [Ping timeout: 240 seconds]
olle has joined #jruby
olle has left #jruby [#jruby]
SqREL has quit [Remote host closed the connection]
SqREL has joined #jruby
SqREL has quit [Ping timeout: 240 seconds]
vtunka has quit [Quit: Leaving]
nowhereFast has joined #jruby
<headius>
jit + concurrency bugs are so much fun
<nowhereFast>
I'm trying to use the google_drive gem which makes use of the google_api_ruby_client gem, which in turn makes use of httpclient, which breaks with:
<nowhereFast>
RangeError: too big for int: 100000000000
<GitHub188>
[jruby] headius created jit-conc-fixes (+3 new commits): https://git.io/v7jxg
<GitHub188>
jruby/jit-conc-fixes 9fbb25e Charles Oliver Nutter: Try again to bail out if someone has beaten us to the punch.
<GitHub188>
jruby/jit-conc-fixes c203a88 Charles Oliver Nutter: Ensure once-only acquisition of scope constructor....
<GitHub188>
jruby/jit-conc-fixes 7d91bdd Charles Oliver Nutter: Synchronize bytecode JIT to avoid concurrency issues....
subbu has joined #jruby
subbu is now known as Guest28382
Guest28382 is now known as subbu
<headius>
nowhereFast: is it still broken?
<headius>
I suspect this is an encoding problem we've fixed recently
<headius>
maybe only on master
<headius>
enebo: jrubyc encoding related?
subbu has quit [Client Quit]
<enebo>
headius: oh yeah?
<enebo>
headius: no I don't think it is related to encoding
<headius>
yeah neither do I now
<headius>
it still fails in 9.1.13
<headius>
I haven't tested master
<enebo>
headius: probably the compiler is asking for the name of the krestarg and choking on ""
<headius>
maybe
<enebo>
but it is all ascii in repro so I doubt it is encoding
<headius>
yeah
<enebo>
nowhereFast: headius: I can look at this one quickly.
<headius>
ok
<enebo>
I keep meaning to go over to windows but this will be a more enjoyable one so I will do it first
subbu has joined #jruby
subbu is now known as Guest12021
Guest12021 is now known as subbu
subbu has quit [Client Quit]
<nowhereFast>
enebo: I'm using the dry-container gem doing SomeContainer.register(:some_class, ->(**kwargs) { SomeClass.new(kwargs) }), works fine in ruby form, but not when compiled I think due to ^this issue
<enebo>
nowhereFast: oh so name does not matter then at all
<enebo>
nowhereFast: it is no ** bare which is the issue
Guest18220 is now known as subbu
<nowhereFast>
oh, okay. Let me know if there's any other info I can help with
<enebo>
nowhereFast: I was asking but I have realized it is even with a name
<nowhereFast>
ahh.. thought you were having a moment of revelation
<nowhereFast>
yup, it happens in both cases
<headius>
nowhereFast: I think this might qualify as a bug in httpclient
<headius>
I didn't realize they are just calling the Java socket APIs here directly...I suspect it's the logic for calling those APIs that doesn't like that they're passing a huge number in
<headius>
I also have to wonder if this is still needed...we've had a working SSLSocket for a long time now
<headius>
it must be more complete than this
<headius>
I'll comment more on the bug
<GitHub84>
[jruby] headius closed issue #4744: Too big for int https://git.io/v7jS0
<nowhereFast>
hmm.. so its either google's api client gem or the google_drive gem that need to bring their timeout value down..
<nowhereFast>
headius: thanks, I'll log it with httpclient
<headius>
nowhereFast: or httpclient should handle this case
<headius>
that seems like the best option to me
<enebo>
well this jrubyc thing is really weird
<enebo>
IR generated in both are identical
<enebo>
but kwargs pushed down through ** in an instance super is {:'' => restkwarg_var}
<enebo>
and it blows up looking for '' in the current scope
<nowhereFast>
:-) glad I could provide some entertainment then
<enebo>
but when run as a script the exact same thing happens
<enebo>
nowhereFast: The beauty of our AOT design is it loads the code as IR and then will compile or not based on how hot it is
<headius>
enebo: persistence
<headius>
I bet it's losing something
<enebo>
headius: at this point without stepping I would say the scope has no '' entry
<enebo>
although I would not expect it to
<headius>
persistence has to rebuild the scope so it's probably mucking up the signature somehow
<enebo>
hmmm that gives me an idea
<enebo>
the repro has both classes loaded through persistence
<headius>
or SIgnature itself doesn't round trip restkwarg properly
bbrowning is now known as bbrowning_away
nowhereFast has left #jruby [#jruby]
<GitHub81>
[jruby] enebo pushed 1 new commit to jruby-9.1.13.0: https://git.io/v5eWH
<GitHub81>
jruby/jruby-9.1.13.0 ce64002 Thomas E. Enebo: Fixes #4621. jrubyc compiled representation breaks but source rb works fine....