Printing of function names that are invalid identifiers (`var"..."`) in stacktraces doesn't always work right

[CameronBieganek]

This works as expected:

julia> var"#foo#bar"() = throw(ArgumentError("oops"));

julia> qwer() = var"#foo#bar"();

julia> qwer()
ERROR: ArgumentError: oops
Stacktrace:
 [1] var"#foo#bar"()
   @ Main ./REPL[4]:1
 [2] qwer()
   @ Main ./REPL[5]:1
 [3] top-level scope
   @ REPL[6]:1

But this does not:

julia> var"foo#bar"() = throw(ArgumentError("oops"));

julia> asdf() = var"foo#bar"();

julia> asdf()
ERROR: ArgumentError: oops
Stacktrace:
 [1] foo()
   @ Main ./REPL[1]:1
 [2] asdf()
   @ Main ./REPL[2]:1
 [3] top-level scope
   @ REPL[3]:1

Note that in the second example, the function name does not start with #, and the stacktrace prints foo() instead of var"foo#bar"().

[BioTurboNick]

Was just going to poke at this and... wat?

julia> var"#foo#bar"() = throw(ArgumentError("oops"));

julia> qwer() = var"#foo#bar"();

julia> qwer()
ERROR: ArgumentError: oops
Stacktrace:
 [1] var"#foo#bar"()
   @ Main .\REPL[1]:1
 [2] qwer()
   @ Main .\REPL[2]:1
 [3] top-level scope
   @ REPL[3]:1

julia> var"foo#bar"() = throw(ArgumentError("oops"));

julia> qwer()
ERROR: MethodError: no method matching var"#foo#bar"()
The type `var"#foo#bar"` exists, but no method is defined for this combination of argument types when trying to construct it.
Stacktrace:
 [1] qwer()
   @ Main .\REPL[2]:1
 [2] top-level scope
   @ REPL[5]:1

[Seelengrab]

Almost looks like something in dispatch ignores leading # in symbol names..? Or maybe it thinks this is a closure, since that's about how they are lowered? If so, that's a gnarly bug!

[vtjnash]

foo#bar makes a function type named #foo#bar, so that is fine. In general, # is a reserved character by the runtime, so it can cause confusing behavior if you try to use it directly

[JeffBezanson]

Yes, this is due to name mangling that we use internally, so if you use names that collide with that you can get unexpected results. I don't think we will do anything about this.