<kares>
but I see headius did a work-around for 1.8
<kares>
nokogiri tries to support JRuby 1.7+ in latest release
<kares>
but 1.7.x only in 1.9+ compat so this should have been fixed there (also)
<kares>
anyway, JRuby's fix makes perfect sense for supporting 'older' exts such as Nokogiri
mistergibson has quit [Ping timeout: 260 seconds]
sidx64 has joined #jruby
mkristian has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
mkristian has quit [Ping timeout: 248 seconds]
sidx64 has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
claudiuinberlin has joined #jruby
mkristian has joined #jruby
sidx64 has joined #jruby
Puffball has quit [Remote host closed the connection]
drbobbeaty has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
drbobbeaty has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
sidx64 has quit [Client Quit]
sidx64 has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
mkristian has quit [Quit: This computer has gone to sleep]
drbobbeaty has joined #jruby
<kares>
headius: don't think master (jruby-head) will be green - that one jit target has been failing consistently
<kares>
maybe it points to a (JIT) bug since 9.1 is fine
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
sidx64 has joined #jruby
sidx64 has quit [Client Quit]
bbrowning_away is now known as bbrowning
sidx64 has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
mkristian has joined #jruby
sidx64 has joined #jruby
sidx64 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Puffball has joined #jruby
chrisseaton has joined #jruby
<headius>
kares: yeah I'm reluctantly accepting that...but it looks like a JVM bug
<headius>
I'll have a look today
<enebo>
lopex: I think we will just comment out the warning in joni and put out a release with just that commented out
<enebo>
lopex: in order to properly fix this we need to change joni so we can register the default warningcallback and not just use a raw System.err.println default.
<enebo>
lopex: I do not see not displaying this warning as a big compatibility issue but seeing random warnings which may not be your fault will not go over well
<enebo>
as you said native extensions could use a regexp that causes this so we cannot really fix this globally without changing joni itself
<kares>
headius: incidentally on the rails issue - its been reported that now it works
<kares>
so maybe travis-ci does accept a 'red' build
<kares>
(as their jruby-head)
<headius>
yeah I did not expect that
xardion has quit [Remote host closed the connection]
<enebo>
lopex: yeah I was wondering if joni had reasonable display method in it
<lopex>
enebo: it has one in logging code, but onigmo has another
<enebo>
lopex: but without knowing joni very well I suspect we should provide the regex instance itself too so more robust warning callback can may do more than display a string
<enebo>
lopex: can we lookup RubyRegexp from joni Regex?
<lopex>
enebo: well, onigmo does it on it's own without mri
<lopex>
enebo: we could via userObject that's on Regex
<lopex>
otherwise
<enebo>
lopex: oh you mean there is a field with the RubyRegexp on Regex?
<lopex>
enebo: not yet
<lopex>
enebo: well, not user for that purpose now
<lopex>
*used
<lopex>
enebo: there should be not leaks though if used
<enebo>
lopex: yeah that sounds wrong to me
<enebo>
lopex: for that reason
<lopex>
enebo: wrt leaks ?
<enebo>
lopex: but if we get Regex instance could we have something to get RubyRegexp back?
<enebo>
yeah leaks
<enebo>
too many times have we stored Ruby* object and leaked
<lopex>
enebo: I think that user Object should be ok though
<enebo>
I guess I don't know if joni has any static fields or something which could cause leaks but it is simpler to not worry
<lopex>
enebo: no, no statics
<enebo>
lopex: but that will not fix old extensions which are just passing in a String
<lopex>
enebo: well, joni should display the pattern on it's own anyways
<enebo>
but OTOH if someone is just using a Regex outside of RubyRegexp then there is nothing to look up
<lopex>
enebo: so it's a non issues thing
<enebo>
lopex: but how?
<enebo>
lopex: DEFAULT is just a mistake. It should have never existed
<lopex>
enebo: it's during parse
<lopex>
enebo: so it's has access to the pattern bytes
<enebo>
oh I see you present some CharSequence version of the regexp in parsing
<lopex>
yes
<enebo>
for errors
<enebo>
ok
<lopex>
enebo: I was worrying for anything during mathcing in the past
<lopex>
but it's not there
<lopex>
jeese
<lopex>
forgive me my grammar
<lopex>
lol
<lopex>
I just read what I've written above
<enebo>
haha don't worry
<lopex>
enebo: but the thing is still a mistery
<lopex>
enebo: and I agree the default warning should be null warning
<lopex>
enebo: mri got it so much easier using globals
<lopex>
enebo: well
<lopex>
enebo: actually mri should use external encoding for warnings
<enebo>
lopex: I wondered if it was real now a little bit
<enebo>
lopex: Do you recall how I tested?
<lopex>
enebo: but it does not
<lopex>
enebo: puma and ctrl thing ?
<enebo>
I said if I hit ^V a few times I would see the string...DERP
<lopex>
yeah
<enebo>
I don't know that I had copied that to my paste buffer on windows but I MIGHT have
<enebo>
in any case I am still happy I eliminated the possibility for now until we get better warning support
<lopex>
enebo: I thought that message got IO blocked and the warnings could be true when forking
<lopex>
so that's two issues right ?
<enebo>
lopex: so things we need are: 1) setDefaultWarningCallback(...) 2) new methods so we can warn with Regex object 3) See if my problem was real or not
<enebo>
lopex: I doubt it was forking but IO was not flushed until I did ^C and it is windows so I have no idea how windows flushes
<lopex>
enebo: I'm opting for NONE warnings by default
<enebo>
lopex: yeah DEFAULT should be NONE (e.g. we should not use DEFAULT any more)
<enebo>
lopex: but if you set a default it will use it
<lopex>
ok
<lopex>
well, we can change it
<lopex>
we have that :P
<enebo>
oh yeah?
mistergibson has joined #jruby
chrisseaton has left #jruby [#jruby]
<lopex>
enebo: we can change that in joni
<lopex>
enebo: I'm suspecting misunderstanding
<lopex>
quoted
<enebo>
lopex: ok yeah let's add it
<enebo>
lopex: and make a new callback API which adds Regex as field in addition to String
<enebo>
lopex: I think String by itself is weak sauce but we can obviously support both for backwards compat
<lopex>
enebo: ScanEnvironment already has access to the patter, but yeah
<enebo>
warn(String), warn(String, Regex)
<enebo>
Regex is the root of the object which is being warned about so who knows what someone would want to examine on a warning some day
<lopex>
yeah, for potential future use from jruby runtime
<enebo>
String only APIs for warning/errors is also killing me on bytelist_love too
<lopex>
enebo: can you mention that in the issue?
<enebo>
I did I thought
<GitHub196>
[jruby] ChrisBr opened pull request #5056: Time#at receives 3rd argument which specifies the unit of 2nd argument (ruby-2.5...ruby25/time#at) https://git.io/vAuPX
<enebo>
wait you mean saying new interface with backwards compat?
<lopex>
enebo: we could implement a default
<enebo>
what would it do?
<lopex>
dont use the pattern
<lopex>
err, I beggining to loose it
<enebo>
haha
<lopex>
beginning
<lopex>
doh
<enebo>
default == default warning?
<enebo>
or default impl for String, Regexp warn method?
<lopex>
the one without regexp right ?
<enebo>
no?
<enebo>
???
<lopex>
yes!
<lopex>
confusion!
<lopex>
restart
<enebo>
real problem in a joni default for warn impl is what does it do. System.err.println seems like a poor idea
<lopex>
so let's change it to null
<lopex>
er, empty
<lopex>
but we begin to circle
<lopex>
enebo: it will never break any user unless they test for warns
<lopex>
right ?
<enebo>
lopex: well it won't
<enebo>
lopex: but for jruby we will change things so we call a registered default handler
<enebo>
so nothing should break
<lopex>
enebo: how it will change anything ?
<enebo>
lopex: other users may break but I doubt anyone tests warning strings
<enebo>
lopex: we change the DEFAULT constructor paths to ask for default handler
<enebo>
instead of DEFAULT
<lopex>
enebo: stopped using that System.err yesterday right ?
<lopex>
*jruby
<enebo>
lopex: only for .16
<lopex>
potentially
<enebo>
lopex: sorry no we did not disable DEFAULT
<enebo>
lopex: we just commented out one warning in joni
<lopex>
I know
<lopex>
I'm lost
<lopex>
I just dont see the problem
<lopex>
except the unknown warn from jruby runtime
<enebo>
lopex: as I said maybe I was just wrong about the unknown problem
<lopex>
just change the default and then log the pattern
<lopex>
and then see
<enebo>
lopex: I used ^V as the key which showed the string
<enebo>
lopex: if I had that string in my paste buffer then it would appear to be a bug not just me being dumb
<enebo>
lopex: but I don't remember copying that string to a paste buffer
<enebo>
lopex: so I don't know?
<lopex>
so let's log it during rails tests first as we concluded yesterday
<lopex>
the pattern
<enebo>
lopex: well I will but I would like to fix warn API so I can see the regexp
<enebo>
lopex: that would make it trivial to figure out
<lopex>
enebo: so let's just output it without any escape code
<enebo>
lopex: you just mean modify joni to print it out if it happens
<enebo>
and try it
<lopex>
enebo: yes
<enebo>
lopex: I was thinking we could kill two birds with one stone and fix this broken callback warning api at same time
<enebo>
lopex: you did see test:mri does have some tests for this
<enebo>
lopex: I did not correct since I would like this to be visible as something we need to resolve
<lopex>
enebo: yes, I was targetting those
<lopex>
and they passed
<lopex>
enebo: actually the tests could be more specific since they could match against concerete patterns
<lopex>
enebo: would be picked up immediately
<lopex>
and save us few days :D
<lopex>
and verobsity
drbobbeaty has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]