bug#67005: 30.0.50; improve nadivce/comp/trampoline handling

[Stefan Monnier]

> So IIUC what you a resuggesting would be to pass the symbol holding the
> function that we're calling to a generic (per calling convetion)
> trampoline?

I'm not suggesting anything in particular, really.
I was just pointing out that in order to fix the absence of the function
from the backtrace, we'll need to push the function's symbol onto the
backtrace, which means some piece of the code will now have access to
this symbol.  Depending on how it's done it might allow the trampoline
to find this symbol without having to hardcode it into its code, thus
making it possible to use a small set of precompiled trampolines.


        Stefan