Method definition overwritten warning could be more helpful...

[ericphanson]

...in the case that a file is included twice. In that case, it shows up as

WARNING: Method definition (::Type{PseudoArcLengthContinuation.DefaultLS})() in module PseudoArcLengthContinuation at /Users/rveltz/work/prog_gd/julia/dev/dev1/PseudoArcLengthContinuation.jl/src/LinearSolver.jl:62 overwritten at /Users/rveltz/work/prog_gd/julia/dev/dev1/PseudoArcLengthContinuation.jl/src/LinearSolver.jl:62.
  ** incremental compilation may be fatally broken for this module **

(example from Slack). This is actually really helpful if you know the only way a method at LinearSolver.jl:62 can overwrite a method at LinearSolver.jl:62 is if the code is being evaluated twice, i.e., a duplicated include statement, but is pretty confusing if you don't realize that is possiblity. I propose adding a check to see if the file path and line number is the same for the two methods, and in that case changing the warning to something like

WARNING: Method definition (::Type{PseudoArcLengthContinuation.DefaultLS})() in module PseudoArcLengthContinuation at /Users/rveltz/work/prog_gd/julia/dev/dev1/PseudoArcLengthContinuation.jl/src/LinearSolver.jl:62 evaluated twice. Check for duplicate `include` statements.
  ** incremental compilation may be fatally broken for this module **

The warning is generated here

julia/src/gf.c

Lines 1372 to 1387 in 3b53f54

	if (jl_options.warn_overwrite == JL_OPTIONS_WARN_OVERWRITE_ON) {
	jl_method_t *method = (jl_method_t*)newentry->func.method;
	jl_module_t *newmod = method->module;
	jl_module_t *oldmod = oldvalue->module;
	JL_STREAM *s = JL_STDERR;
	jl_printf(s, "WARNING: Method definition ");
	jl_static_show_func_sig(s, (jl_value_t*)newentry->sig);
	jl_printf(s, " in module %s", jl_symbol_name(oldmod->name));
	print_func_loc(s, oldvalue);
	jl_printf(s, " overwritten");
	if (oldmod != newmod)
	jl_printf(s, " in module %s", jl_symbol_name(newmod->name));
	print_func_loc(s, method);
	jl_printf(s, ".\n");
	jl_uv_flush(s);
	}

but I don't really know C, so I filed this issue instead of making a PR.

[vtjnash]

The risk here (with the proposed change) is that having f() = 1; f() = 2 on a line would print that same message. I'll make a PR with something similar however