01:16 <headius[m]> mspec ci, you mean?

01:17 <headius[m]> Just running mspec defaults to "run" which doesn't pick up tags

01:19 <headius[m]> There's a few tags but not much I'd expect people to care about

01:19 <headius[m]> awk mode and old encoding flags and such

01:22 <headius[m]> I believe some of those might depend on bin being in path, not sure. I always have bin in path

07:40 <ChrisSeatonGitte> @headius do you know what I run to run specs with tags? I tried `bin/jruby spec/mspec/bin/mspec ci :command_line` but that still seems to run tagged specs. It's not about `command_line` specifically - just can't work out the proper way to run specs with tags at all.

07:42 <headius[m]> ci should do it

07:47 <headius[m]> Is it not picking up jruby config or something?

07:48 <ChrisSeatonGitte> That's possibly it... `mspec` is too magic in some ways, and then also relies in manual configuration for things which should be magic like finding tags.

07:48 <headius[m]> I run mspec ci on almost a daily basis, so I know that's the correct way to do it

07:48 <ChrisSeatonGitte> Do you use `ci :foo` though?

07:49 <headius[m]> Yes

07:50 <headius[m]> Is bin in PATH? Perhaps it is picking up some other ruby during setup

07:50 <headius[m]> Is there some spec you are trying to confirm?

07:51 <ChrisSeatonGitte> I was looking at updating specs but can't figure out how they all run - I think I'll leave it to Benoit.

07:51 <ChrisSeatonGitte> `PATH=`pwd`/bin:$PATH bin/jruby spec/mspec/bin/mspec ci :command_line` doesn't fix it - still get `The --internal-encoding command line option sets Encoding.default_internal if given an encoding as a separate argument` for example

07:51 <headius[m]> In order to pass as many specs as possible, we usually run it through rake so it passes additional flags like object space etc

07:52 <ChrisSeatonGitte> What `rake` command would you run?

07:52 <headius[m]> But mspec ci ought to be fine

07:52 <headius[m]> rake spec:ruby:slow probably has the cli specs

07:53 <headius[m]> I haven't looked in awhile but I assume they launched a subprocess for almost everything

08:04 <ChrisSeatonGitte> To be really specific, `bin/jruby spec/mspec/bin/mspec spec/ruby/command_line/dash_internal_encoding_spec.rb` fails for me, and that's not about tags because that's not tagged.

08:04 <ChrisSeatonGitte> Tried `-t bin/jruby` as well

08:08 <ChrisSeatonGitte> Yes they do. Ok never mind I'll just leave these specs.

08:13 <headius[m]> I'll give it a try in a bit

08:13 <ChrisSeatonGitte> Cheers

08:14 <headius[m]> Did you get any further playing with JRuby native image?

08:14 <headius[m]> I'm planning to start working on that after RailsConf

08:15 <headius[m]> I need to figure out a good way to isolate places we use method handles directly but otherwise I'm hoping JRuby proper can be shoehorned in easily enough

08:17 <headius[m]> Serbia so many different run time configurations, need to figure out a way to pair that down

08:17 <headius[m]> Pare

08:17 <headius[m]> Ha, JRuby

08:17 <headius[m]> Not Serbia

08:17 <ChrisSeatonGitte> No, I got as far as some failures in your static initialisers but didn't know as much about your internal architecture to easily figure out how to fix it.

08:19 <headius[m]> Oh right, the file descriptor cracking stuff.

08:19 <ChrisSeatonGitte> Yes that was it. I guess there must be some fallback for other JVMs?

08:19 <headius[m]> Most of that could be commented out for early experiments. We generally run okay without it, but some native operations don't be quite right

08:19 <headius[m]> The fallback is that they just allowed to fail to crack open the file descriptor

08:20 <ChrisSeatonGitte> I thought maybe if you turned on all options for using as basic Java as possible, no unsafe, it may work?

08:20 <headius[m]> In which case we use other tricks to pretend channels are file descriptors

08:20 <headius[m]> Yeah, that's what I'm thinking as well

08:20 <headius[m]> Disabling native features should cover most of that, but the static initializers we just allowed to run and then ignore it

08:21 <headius[m]> Might need to look at a jnr back end for the graal ffi as well

08:21 <headius[m]> I don't suppose you guys ever made one of those

08:23 <ChrisSeatonGitte> Native Image used to support JNR, not sure about now.

08:23 <ChrisSeatonGitte> (Because we moved to Truffle NFI.)

08:23 <headius[m]> Ah, well that's a good start

08:23 <headius[m]> Maybe one less thing to worry about

08:24 <ChrisSeatonGitte> You unlikely ever want to use Graal FFI by the way - that's something very specialist. Ask if you're trying to use that for something when you get to that point.

08:26 <headius[m]> Okay thanks

08:26 <headius[m]> If JNR works or can be made to work then no big deal

10:07 <xmj> moin

10:07 <xmj> is there a workaround to otto.takari.io timing out and maven fetching some artifact instead from maven central?

10:09 <xmj> it looks like i've jumped into a time machine, https://irclog.whitequark.org/jruby/2015-08-20

15:25 <headius[m]> xmj: yeah I remember this

15:26 <xmj> i've had a bunch of timeouts with that box today ... before stuff ended up cached

15:26 <xmj> took quite a while, and the settings.xml where i set timeout to 5s was ignored =(

15:27 <headius[m]> Are you building a current version of JRuby? my memory is fuzzy on this but I feel like we just removed that source because the plugins could get those dependencies elsewhere

15:28 <headius[m]> Chris Seaton (Gitter): I get this for command_line: 29 files, 137 examples, 155 expectations, 22 failures, 0 errors, 10 tagged

15:28 <headius[m]> so it's picking up tags but either the right flags aren't getting passed to JRuby or we stopped running command_line specs at some point and have not picked up the newer specs

15:29 <headius[m]> none of these look critical...mostly stuff that was added that we aren't redirecting to existing flags

15:30 <headius[m]> I don't think I even knew about --internal-encoding for example...but it would just be the same as a -E flag

15:30 <xmj> not quite, i'm trying to port HBase 2.1 to FreeBSD and we had patched jruby-complete into the predecessor

15:31 <headius[m]> aha...so that might be an older jruby-complete with a takari dep?

15:31 <xmj> so that patch was carried over to the new one, and the timeouts were only happening with jruby dependencies

15:31 <xmj> 9.0.3.0, can't be too recent as i've worked on that predecessor some three years ago

15:31 <xmj> maybe four :-)

15:31 <headius[m]> wow yeah

15:31 <headius[m]> that's like 3 years old

15:32 <headius[m]> so things have definitely changed

15:32 <headius[m]> current JRuby builds shouldn't have dependencies in the takari repo

15:32 <xmj> oh sweet

15:34 <xmj> 9.1.13.0, which comes in the newer pom, seems more recent

15:35 <headius[m]> what pom is that?

15:35 <headius[m]> that might be better but it's still over a year old

15:36 <xmj> headius[m]: https://github.com/apache/hbase/blob/rel/2.1.4/pom.xml#L1406

15:37 <headius[m]> So hbase has not updated jruby in awhile

15:38 <headius[m]> That JRuby was released Sept 2017

15:40 <xmj> hmmm yes, at least that version downloads fast :)

15:40 <headius[m]> ok cool, so it's past that problem anyway!

15:40 <xmj> yep, thank you!

15:41 <headius[m]> if you have the time you could try to file an issue to update hbase to a newer jruby but I'm glad we found a path for you

15:47 <headius[m]> Chris Seaton (Gitter): if you could, file a bug about the command_line specs not lining up with tags...I suspect it fell out of the complicated mspec magic and we just missed some new flag features and spec additions