14:21 <enebo[m]> kares: I put it from master

14:22 <enebo[m]> kares: and I actually pushed it :)

14:23 <enebo[m]> I guess I forgot to push last friday

14:37 <headius> I hate the Maven release plugins automatically pushing things but I suppose it avoids issues like this

14:37 <headius> Good morning all

14:50 <basshelal[m]> Good morning!

15:34 <adam12> I swore I saw links to M1 preview on Twitter but can't seem to find them again. Could anybody link me to them?

15:43 <basshelal[m]> headius: Have you seen my latest work on JNR-FFI? I'm planning a ton of work over the next while for it and want to get your opinion

15:48 <headius> I was heads down last week getting M1 support fixed up but I definitely would like to see what you've been up to

15:48 <headius> I'm planning to spend this week improving our pass rate on M1 which will undoubtedly mean more FFI work

15:52 <basshelal[m]> Yeah I noticed, M1 hasn't been easy it seems, it has shown some of the cracks in JNR

16:05 <headius> Yeah the unusual handling of variadic function calls shakes things up a little bit

17:32 <bensheldon[m]> > <@jamesgecko:matrix.org> I have a bizarre issue on my M1 MacBook where jruby shows permission issues whenever it installs a gem. This is what happens when I try to install jruby via ruby-build, but I get similar errors when using jgem after installing jruby from homebrew.... (full message at https://libera.ems.host/_matrix/media/r0/download/libera.chat/87de42cb4b56b3a974cc437d8888a2560b409ab9)

17:35 <headius> yeah that was probably fixed by another related change in jnr-posix but the full set of fixes will be out in 9.3.3 (next week-ish)

17:36 <headius> the jnr-posix change was fixing a stat struct layout to match M1 Darwin's definition... they use 64-bit inode by default which shifts other fields around

18:20 <headius> omg facepalm of the day

18:21 <headius> couldn't figure out why my fixes on M1 weren't working, or working in the wrong order (fix => broken, revert => fixed)

18:21 <headius> turns out that tabbing away from a screen sharing session on my M1 mini does not trigger IntelliJ to save the file like it does if you tab away on a local window

18:22 <headius> I was running the test over an ssh session in a local terminal window

18:22 <headius> also: screen sharing is pretty miserable to work with due to it handling cmd+tab app switching very poorly

21:09 <headius> ok I got test:jruby to complete by excluding three socket-related tests

21:09 <headius> on M1

21:09 <headius> 1096 tests, 76033 assertions, 0 failures, 6 errors, 1 pendings, 1 omissions, 0 notifications

21:10 <headius> of those six errors, one is a known problem with this suite running as an administrator (expects a permission error) and the other five are unexpected file permission errors

21:11 <headius> I'm going to run this with x86 and see how that compares... some of this might be altered security on newer MacOS

21:13 <headius> yup, same errors on an x86 run

21:13 <headius> that is great news

21:13 <headius> I think these might be due to different umask on MacOS

21:13 <headius> enebo: what's your umask

21:14 <enebo[m]> 0022

21:15 <headius> hmmm same here

21:16 <headius> well at least it is equally broken on M1 and x86

21:16 <headius> enebo: hmm I wonder if the same excludes hang on x86

21:25 <headius> nope no hangs

21:25 <headius> so there's some marching orders here to fix

21:25 <headius> I will attach this work to my GHA PR

21:35 <headius> ah at least rubyspec/language is green so far

21:36 <headius> enebo: I will look into the different failures on arm64 and newer macos but we might want to consider using the exclude system for JRuby tests or deleting those tests in favor of specs that cover the same

21:37 <headius> we have a handful of things excluded by platform, usually manually with "if WINDOWS\ndef test_blah" kinds of gross excluding

21:38 <enebo[m]> I am in favor of anything which reduces our tests too but that I guess is different. I don't have any issues with using excludes though

21:38 <headius> hmm yeah probably best to find or write equivalent specs

21:38 <headius> so we can exclude using tags

21:38 <headius> there's very little in test:jruby that is jruby-specific

21:38 <headius> mostly JI that should be in JI specs