Liothen_ has quit [Read error: Connection reset by peer]
Liothen_ has joined #jruby
marcheiligers[m] has quit [Quit: Idle for 30+ days]
<headius[m]> yo
<headius[m]> enebo: so weirdly enough I switched my apple aarch64 machine to Java 8 and the variadic tests that failed now work
<headius[m]> it was running 16 before
<enebo[m]> uh oh
<headius[m]> I am wondering if 16 changed something about how it reports platform (ptr width or arch or something)
<headius[m]> oh wait... I'm stupid... the 8 I have installed is x86_64 so it went back to that native backend
* headius[m] shakes fist at rosetta
<headius[m]> back to trying to untangle how variadic args work in jnr
<enebo[m]> headius: It maybe has had enough time that someone else is doing something FFI-like on Java on this platform
<enebo[m]> Although I guess you need to understand how jnr does it first
<headius[m]> yeah I am trying to see what ffi gem does but I don't see any changes in the variadic logic related to Apple aarch64
<headius[m]> currently I am trying to determine where it does the assembly of varargs on stack because this varadic logic may be the same on apple x86_64
<headius[m]> I am hoping this is just not choosing the right logic because apple + arm64 is not a known config previously
<enebo[m]> is this broken on MRI too?
<headius[m]> FFI gem supports apple arm64 currently so I assume varargs work
<headius[m]> Okay, the gem has a completely separate path for varargs versus jnr that does it from Java. I think I will need to add native code to call the right ffi varargs logic
<headius[m]> It is strange though, because it works fine doing the mapping from java on other platforms
<headius[m]> There must be a way to set up the stack from java without adding more code native side
<headius[m]> Based on my reading of the Apple documentation this seems like it's just a size or alignment issue