[Stacktraces] Name of the macro instead of macro expand

[vchuravy]

julia> f() = @time error()
f (generic function with 1 method)

julia> f()
ERROR: 
Stacktrace:
 [1] error() at ./error.jl:42
 [2] macro expansion at ./util.jl:175 [inlined]
 [3] f() at ./REPL[2]:1
 [4] top-level scope at REPL[3]:1

Would be great if [2] could have been [2] @time expansion at ...

[c42f]

I had a quick look into how to do this. Currently, the fact that a given part of the code came from a macro seems to be lost after macro expansion.

So why do you see macro expansion in the backtrace? Well, this information is re-inferred during lowering in the compile-body and compact-ir passes, but the inference is imperfect: it relies on the file name where the macro was defined being different from that of the function itself. Certainly, the exact macro name is lost by this point!

So to do this correctly is actually somewhat involved: it requires adding information to the AST which comes out of macro expansion, and correctly preserving this through the initial lowering passes. (LineNumberNode won't do the trick since it only has file and line information.)

Still, this seems directly related to what I'm working on at the moment, so it's probably something I should tackle as part of the larger frontend work I'm doing.

[vtjnash]

May have been mostly implemented in #44995