<enebo[m]>
fun new bisect experience. bash shell script changes for a while was not properly emitting --dev so I cannot use launcher to find an interp point (now switching to runnign a java command-line)
<headius>
Ha
<headius>
Yeah we transitioned to the shell script just this year
<headius>
Which I have to say is pretty nice given how old the bash on Mac OS is and the growing prevalence of zsh
<enebo[m]>
yeah. I was just surprised to see the bisect end on some launcher changes then realize it was running things differently during some sequence of stuff
<enebo[m]>
It is fine. It is just code which changed just something I did not think would happen sort of scenario
<enebo[m]>
fun stuff...this wrong backtrace happens in 9.3 as well if CompileMode=OFF
<enebo[m]>
but now I am wondering if our current script is emitting what it should here
<enebo[m]>
not that we should run differently with that toggled
<enebo[m]>
no ignore that
<enebo[m]>
9.3 launcher is emitting the same stuff
<enebo[m]>
So this backtrace regression is quite old
<enebo[m]>
If I had to lay money I would put it onto my work to consolidate all runtime to using StaticScope over IRScope
<enebo[m]>
but this will take more digging since it is much older than I thought
<enebo[m]>
Yeah so when in interp the backtrace is wrong for define_method with a block but if it is forced or jitted it is fine
<enebo[m]>
With that said with normal JIT settings 9.3 runs mocha's unit tests without hitting this
<enebo[m]>
Even with --dev
<enebo[m]>
but who knows what is loaded which may change a 2.6 vs a 3.1 impl
<enebo[m]>
Possibly things are being called differently
<enebo[m]>
I guess I will investigate the obvious mistake I found and see if all is well
<enebo[m]>
I really wanted to just see a bisect commit