14:09 <headius> yo I'm back in the office

14:10 <rtyler> \o/

14:11 <headius[m]> we release jruby-launcher just rarely enough that I never remember how to build it for Windows

14:11 <headius[m]> oh right mingw and the VM I wiped out because I was getting virus warnings

14:12 <headius[m]> bleh time to figure out the cross-compiling option

14:17 <rtyler> headius[m]: I don't know if you saw the follow up on that JRuby/Gradle ticket, but I was able to figure out that user's issue https://github.com/jruby-gradle/jruby-gradle-plugin/issues/341

14:17 <rtyler> just need to fix some other things before I can release a new version of JRuby/Gradle

14:17 <headius[m]> oh nice

14:18 <rtyler> actually, does this error message look familliar to you? https://github.com/jruby-gradle/jruby-gradle-plugin/pull/343#issuecomment-487388600

14:19 <headius[m]> That is new to me

14:19 <headius[m]> The default_exe_path we should just fix to not do j prefix anymore

14:19 <headius[m]> I never wanted that and nobody ever has both JRuby and CRuby in PATH at the same time because obviously

14:19 <rtyler> heh

14:20 <headius[m]> I dunno I suppose we should ask around

14:20 <headius[m]> enebo: do you run the j versions?

14:24 <rtyler> headius[m]: do you use @jruby on twitter for user research at all? could ask the hoard :D

14:24 <headius[m]> Yeah we do occasionally...I'll run it up the flag pole

14:30 <rtyler> oh interesting! this "installing into parent path" error appears between JRuby 9.1.15.0 and 9.1.16.0

14:37 <rtyler> hah, the rubygems version was only from 2.6.14 to 2.6.14.1 according to the changelog, and rubygems' changelog doesn't have a .1 entry

14:38 <headius[m]> hmmm likely a security release

14:38 <headius[m]> so you can't reproduce with 9.1.15.0 but that's the only RG change

14:38 <rtyler> correct, cannot reproduce the test failure with 9.1.15.0

14:38 <rtyler> kk

14:39 <headius[m]> https://bugs.ruby-lang.org/issues/14481

14:39 <rtyler> derps

14:39 <headius[m]> this appears to be what got backported into 2.6.14.1

14:39 <headius[m]> https://devcenter.heroku.com/changelog-items/1369

14:39 <rtyler> ah, thanks for the link

14:39 <headius[m]> "Prevent path traversal when writing to a symlinked basedir outside of the root."

14:40 <headius[m]> that sounds like a candidate

14:40 <rtyler> oh jinkeys

14:40 <rtyler> that does

14:41 <headius[m]> https://github.com/ruby/ruby/commit/4237809aa3983fca36b520f95b48fdf9e7e88de2#diff-7d891128874ea0e4a2f0c3b809ee2c87R458

14:41 <headius[m]> or maybe the similar block next one down

14:42 <headius[m]> so something they changed in here passes this test and that's where the error started coming from

14:42 <rtyler> the next block down seems like a strong candidate, since we pass --install-dir and a bunch of other flags to the `gem install` invocation

14:43 <headius[m]> so they're doing something when it detects installing in a symlinked path

14:43 <headius[m]> oh, or just fiddling with the parent path

14:44 <rtyler> that's a really good hunch

14:44 <headius[m]> the two paths in your error are identical

14:46 <rtyler> the symlink theory is a good one, since this is not reproducible outside of our integration tests, and we have to do some goofy things with our integration tests in order to set up the environments properly

14:46 <headius[m]> https://github.com/ruby/ruby/commit/4237809aa3983fca36b520f95b48fdf9e7e88de2#diff-3a0baa1500616100ad7469f0f7b20c3cL427

14:47 <headius[m]> at least in this general area

14:47 <headius[m]> maybe we don't put the / on a path or something? This seems brittle

14:51 <rtyler> I don't think this is a JRuby problem, I think it's a JRuby/Gradle problem fwiw

14:51 <rtyler> but yeah, any + strong concatenation like this seems brittle

14:56 <headius[m]> yeah that's fine, I'm glad we have a paper trail at least

17:40 <rtyler> any good JRuby news from Railsconf?