most of remaining issues are getting things in the right places so that native-image can see them (psych jar, openssl jar, jffi binary) and then I think things will largely work fine
Hi! The nightly builds are not accessible: https://www.jruby.org/nightly links to https://repo.aws.oss.sonatype.org/content/repositories/snapshots/org/jruby/jruby-complete/ which gives a 404 error page.
It's not a 404 for me
Oh but the jar isn't there
Well that's strange, the nightly snapshot job disappeared
Afternoon/Morning. I have this project: https://github.com/fidothe/saxon-rb which wraps the Saxon Java library. We vendor a specific version of Saxon in to the gem with jar-dependencies. There's a mechanism to use different Saxon JARs if you need them (for an older version, or for the EE version). That has worked fine. I just came to use it again to test something and I am getting StackOverflow errors. I have a simple example gist at https://gist.github.com/fidothe/5247bb8236e6dcd42d6bbc1d1b63ad2f showing what's needed. All the loading code is in
lib/saxon/loader.rb.
At this point I'm really not sure if this is a bug in my code or a problem somewhere else
I will check with a JRUBY nightly and see
I will open an issue later this evening, once I get home. Hopefully it's just me doing something dumb.
I'd tested the ability to load other versions of Saxon by loading a SaxonPE version
which has a different JAR name
So I imagine that despite having not required any of the
jar_dependencies stuff on the code path that loads an external Jar, trying to load another version of the Saxon HE jars causes a horrendous pile up because the jar_dependencies vendors the Jars under lib, and so on the $LOAD_PATH
So, this is probably one to take up with @mkristian on the
jar_dependencies repo issues list...
(I have checked, and current code example will load an external Saxon PE jar fine)
Yeah, if you think it is related to jar-deps then file it there, but filing on main JRuby repo is fine too
Do you have a stack trace gist handy?
@headius, better reproduction with a non-1-line stacktrace at the updated https://gist.github.com/fidothe/5247bb8236e6dcd42d6bbc1d1b63ad2f
Will file an issue. Not sure there's a simpler reproduction than to clone my project and grab the external JAR, alas.
@fidothe Thank you!
@donv It seems like jruby-complete jars started snapshotting again 🤷♂️
@headius @enebo I don't know if anyone is tackling #5905, but it looks like something I could tackle in a reasonable time. Looking the code over, it seems fairly straightforward. The main thing I don't really understand is the role of RubyString's
gsubFast - the only use of that I can find is in RubyDate.java, which uses a Regexp, so it's not directly affected by behaviour that uses a string pattern arg.
Yeah gsub fast may not be used by this logic...but I suppose it could be?
I think I linked the MRI commit that made this optimization in that bug, so that would be a good rough template to follow (even though this logic isn't a line-for-line port)
is this expected behavior?
I've been setting up my machine for hacking on #5905, and I have a couple of questions. just running
mvn complains about the .polyglot.pom.rb, ./mvnw is fine, and ./mvnw -Pbootstrap runs fine, but ./mvn still complains after. (This is on JDK 1.8 on macOS 10.14) Is this expected? I ran jruby -S rake spec:ruby:fast which reported 3 failures and 2 errors (Signal.list redefines CLD with CHLD if defined FAILED, plus some networking stuff that seems like it might be a platform thing). I can see that's not happening in the last Travis build. Is that expected/a problem?
for MRI Ruby
So that looks like a bug
I wish notifications worked on here
@fidothe What's the complain? I think you might be hitting a recent Maven bug :-(
3.6.4 or something broke polyglot maven
I think if you delete pom.rb it will work fine since we keep the pom.xml up to date too
@fidothe takari/polyglot-maven#200
@jlahtinen That is a bug
Looks like it's probably just a bug in Exception#inspect