<headius>
looking at ways of preserving the load path where a file was found... it will be involved but would help us better represent the source in JVM bytecode (for debugging)...
<headius>
seems like this logic is the heart of the problem... it takes the base load path and the target file and slams them together; we'd have to plumb logic all the way down to here to instead return an e.g. java.nio.file.Path (which preserves parent path and relative path)
<headius>
if we could alter the file searching and parsing logic to pass the Path all the way through, I can emit the bytecode using the relative path. That will allow Java debuggers to just add LOAD_PATH to the list of source locations and step through Ruby code
<headius>
I think this would also help with some of our absolutizing/canonicalizing bugs since we can preserve the original requested filename and the canonicalized full path
<headius>
yeah so this thing eventually creates a JRubyFile so that is part of the problem too. File (and JRubyFile) do not preserve the parent
<headius>
probably need to start from the bottom replacing JRubyFile with Path or similar and move up from there
<headius>
it's very involved
<headius>
byteit101: I think I'm having problems with io-console/pty on MacOS M1
<headius>
not sure if it is relayed to your subspawn stuff yet
<headius>
related
<headius>
wow yeah io-console seems totally busted on M1
<headius>
ok yeah something is wrong with ttymode
<headius>
so when it tries to reset terminal settings it fails
subbu has quit [Ping timeout: 246 seconds]
subbu has joined #jruby
subbu has quit [Ping timeout: 260 seconds]
<headius>
this might be an FFI on M1 bug
<byteit101[m]>
headius: keep me posted on what it turns out to be
<headius>
I'm pretty sure it's an FFI thing right now because the same code runs with only a couple of errors in C Ruby
<headius>
Which means it's going to be a bloody nightmare to figure out because it's some weird low level C calling thing
<headius>
It's super broken though
<headius>
I guess I could try to run CRuby with our FFI console code and subspawn but I have very little confidence that's where the problem is
<headius>
I just need to sit down and make the M1 FFI support perfect and go from there
ldywicki[m] has quit [Quit: You have been kicked for being idle]
razetime has quit [Remote host closed the connection]
<headius>
enebo: I think we should pull off a 9.4 branch now
<enebo[m]>
why?
<enebo[m]>
I saw the 3.2 issue but I am really hoping we do not see 3.2 work rather than people helpnig with 3.1 work
<enebo[m]>
If we see more 3.1 PRs then having it on another branch is going to cause us to manually change branches
<enebo[m]>
maybe we can 3.2-devel or some name like that and then it will become master in a while
<enebo[m]>
then when it is old we will not have consumed the maint branch name
<enebo[m]>
The other issue I just thought of is all the people using jrbuy-head will not currently be hitting what we want the most coverage of
subbu has joined #jruby
subbu has quit [Ping timeout: 260 seconds]
sagax has quit [Remote host closed the connection]