subbu has joined #jruby
subbu has quit [Ping timeout: 240 seconds]
rcrews[m] has quit [Remote host closed the connection]
subbu has joined #jruby
subbu has quit [Ping timeout: 240 seconds]
subbu has joined #jruby
subbu has quit [Ping timeout: 248 seconds]
subbu has joined #jruby
subbu has quit [Ping timeout: 268 seconds]
subbu has joined #jruby
<headius> Tim Uckun: Most of these would probably be best left to defaults, and CMS is largely deprecated now. We will want to revisit them and see if they're worth looping back in
<headius> the only ones I'd say are good to keep would be invokedynamic, ipv4 (presumably needed for this to work right), and maybe the metaspace and codecache settings but only if they're actually needed
<headius> the rest are better left unset or I would need to read up on them more
<headius> also good morning all!
<headius> oh nice someone jumped in and tweaked joni Config to be property-driven: https://github.com/jruby/joni/pull/67
<headius> enebo: any concerns about this? I know all modern JVMs can optimize through lazy static finals like this so I have no concerns at the moment
<enebo[m]> yeah I saw it. No concerns.
<enebo[m]> I asked lopex if hewould review the other PR
<headius> only bit I would have trouble saying yes to is the naming of the properties
<headius> but I don't have a good alternative... there's not really a joni.org and putting the properties under JRuby seems wrong now
<headius> they're also snake_case instead of camelCase but I don't have a strong opinion on that either
<enebo[m]> I don't think there has to be a real org for namespacing
<enebo[m]> org.joni.opt wouldperhaps be more consistent for us but I doubt another joni will pop up in Java lib land
<enebo[m]> I prefer snake case just because it is easier to delineate words
<enebo[m]> but perhaps that is unusuable for java opts
<enebo[m]> heh unusual
<headius> unusual perhaps
<enebo[m]> I don't care because these are rare options
<enebo[m]> and these choices do not make them confusing
<headius> yeah
<headius> someone owns joni.org through godaddy but there's no site
<enebo[m]> probaably not for software
<enebo[m]> 4 letter domains are tough to get
<headius> there's plenty of other tlds now too
<headius> but I'm fine just going with these properties as is
<enebo[m]> 2023-05-22T11:03:06.733-05:00 [main] INFO TraceType : Exception raised: TypeError : superclass must be a Class (Integer given)
<enebo[m]> is it me or are all errors lacking where stuff happens
<headius> that's interesting
<enebo[m]> this is yarp parsing so perhaps not that interesting
<headius> I have not seen that all of them are lacking
<headius> I file an issue about some specific cases swallowing the Java cause when classloading fails but others have been ok
<headius> something like that could be happening during boot before there's any backtrace to present
<headius> final boolean DONT_OPTIMIZE = ConfigSupport.getBoolean("joni.dont_optimize", false);
<headius> that's confusing
<headius> should be OPTIMIZE=true or maaaaybe DISABLE_OPTIMIZATION=false
<enebo[m]> optimise is a nearly meaningless term too
<headius> it's better than turning off "don't optimize"
<headius> but yeah it is pretty vague
<headius> I assume it is configurable for debugging
<enebo[m]> sure I do not use negatives even in if statements 99% of the time much less options
<enebo[m]> protected final void setOptimizedInfoFromTree(Node node) {
<enebo[m]> it is what it is called so I would reverse and default to true I guess
<headius> yeah I just submitted a review with that and a couple other things
subbu has quit [Ping timeout: 256 seconds]
<byteit101[m]> headius: did you see my message from saturday?
<headius> no I will look back
<headius> enebo: not sure what to do about that weird encoding error not erroring under indy on CI... I can't reproduce locally and have no idea how indy would affect it
<headius> byteit101: trying it now
<enebo[m]> headius: worst case we exclude it (even though it tests 10 million other combos)
<headius> require 'subspawn' still works with new gem
<headius> anything else I should try?
<headius> enebo: yeah it is a large test
subbu has joined #jruby
<headius> $ jruby -e 'loaded = $".dup; require "subspawn"; puts ($" - loaded).grep /fix/'... (full message at <https://libera.ems.host/_matrix/media/v3/download/libera.chat/c0834fdc9c82e68269a3a6ef45065dc9de773738>)
<headius> byteit101: ~
<headius> ^
<byteit101[m]> Nope, that's it! I will publish that shortly
<headius> ok!
<headius> you have messed around with this optimization so maybe you have some thought here
<headius> I guess it wasn't actually compiled before because it was static final false
<enebo[m]> weird. I thought my revert to fix perf rgression enabled sunday search
<enebo[m]> I will look into that I guess
<headius> yeah confusing
<headius> enebo: ok yeah reverse what I said... it was true before but now it might be true or false so the old path is followed and does not initialize the variable
<headius> I'm looking into the cve's
<headius> both CVEs should be patched and should be simple gem upgrades
<headius> I have lots track of CRuby security since March but I knew these were floating around
<headius> s/lots/lost/
<headius> enebo: joni PR looks good to me at this point, other than not knowing whether that sunday search patch is right
subbu has quit [Ping timeout: 265 seconds]
<byteit101[m]> Only took me 2 hours to recover from oom/swap hell, but I just published and tagged 0.5.1.1. can be updated in the pom.rb file now
<headius> hah ok
<headius> should any of the other gems updated?
<byteit101[m]> nope, just that one
<byteit101[m]> I didn't actually test it, just used the new-ish github code editor
<byteit101[m]> oops, lets see if the tests fail, I pushed all the native ones except -java :-O
<byteit101[m]> now pushed too
<headius> ok
<headius> none of the failures are your problem so I'll merge
<headius> byteit101: thank you!
<byteit101[m]> Yay!
<headius> is jruby.exe working for you on Windows?
<headius> I'm not sure when this started failing
<headius> I grabbed an updated windows VM, will look into it
<headius> well the exe works fine for me
<headius> oh duh I bet I know what's wrong
<headius> it's restoring a build from linux onto windows and probably messing up the exe somehow
<headius> probably something similar happening on the macOS build
<headius> I'll try reverting these builds to building on their own
<headius> yup that got windows running again
<headius> ok so just down to the one weird failure at this point
<TimUckun[m]> <headius> "the only ones I'd say are good..." <- Thanks.