Internal errors caused by broken Varargs

[maleadt]

I'm not sure we care, given that this Vararg is very wrong, but I guess it maybe shouldn't throw an internal error:

f(x=whatever) = 1
f(::Vararg{1}) = 2
f()

julia> f()
Internal error: encountered unexpected error in runtime:
MethodError(f=Core.Compiler.widenconst, args=(1,), world=0x00000000000015a5)
jl_method_error_bare at /Users/julia/.julia/scratchspaces/a66863c6-20e8-4ff4-8a62-49f30b1f605e/agent-cache/default-honeycrisp-R17H3W25T9.0/build/default-honeycrisp-R17H3W25T9-0/julialang/julia-release-1-dot-10/src/gf.c:2203
jl_method_error at /Users/julia/.julia/scratchspaces/a66863c6-20e8-4ff4-8a62-49f30b1f605e/agent-cache/default-honeycrisp-R17H3W25T9.0/build/default-honeycrisp-R17H3W25T9-0/julialang/julia-release-1-dot-10/src/gf.c:2221
jl_lookup_generic_ at /Users/julia/.julia/scratchspaces/a66863c6-20e8-4ff4-8a62-49f30b1f605e/agent-cache/default-honeycrisp-R17H3W25T9.0/build/default-honeycrisp-R17H3W25T9-0/julialang/julia-release-1-dot-10/src/gf.c:3052 [inlined]
ijl_apply_generic at /Users/julia/.julia/scratchspaces/a66863c6-20e8-4ff4-8a62-49f30b1f605e/agent-cache/default-honeycrisp-R17H3W25T9.0/build/default-honeycrisp-R17H3W25T9-0/julialang/julia-release-1-dot-10/src/gf.c:3067
tuple_tfunc at ./compiler/tfuncs.jl:1927
most_general_argtypes at ./compiler/inferenceresult.jl:151
most_general_argtypes at ./compiler/inferenceresult.jl:109 [inlined]
matching_cache_argtypes at ./compiler/inferenceresult.jl:27 [inlined]
InferenceResult at ./compiler/types.jl:88 [inlined]
typeinf_edge at ./compiler/typeinfer.jl:919
...
ERROR: UndefVarError: `whatever` not defined

[maleadt]

A similar case:

f(x=whatever...) = 1
f(x::T, y::Vararg{1}) where {T} = 2
f()

Internal error: encountered unexpected error in runtime:
TypeError(func=:UnionAll, context="", expected=Type, got=1)
ijl_type_error_rt at /cache/build/default-amdci5-3/julialang/julia-release-1-dot-10/src/rtutils.c:119
ijl_type_error at /cache/build/default-amdci5-3/julialang/julia-release-1-dot-10/src/rtutils.c:127
ijl_type_unionall at /cache/build/default-amdci5-3/julialang/julia-release-1-dot-10/src/jltypes.c:861
rewrap_unionall at ./essentials.jl:388
rewrap_unionall at ./essentials.jl:388
rewrap_unionall at ./essentials.jl:396 [inlined]
most_general_argtypes at ./compiler/inferenceresult.jl:140
most_general_argtypes at ./compiler/inferenceresult.jl:109 [inlined]
matching_cache_argtypes at ./compiler/inferenceresult.jl:27 [inlined]
InferenceResult at ./compiler/types.jl:88 [inlined]
typeinf_edge at ./compiler/typeinfer.jl:919
...
ERROR: UndefVarError: `whatever` not defined

[maleadt]

Naively adding a bail-out path that returns Bottom if the param isn't a Type (which I'm not sure is legal) gets me a little further, but it then fails during inlining:

ERROR: TypeError: in <:, expected Type, got a value of type Int64
Stacktrace:
  [1] ir_inline_unionsplit!(compact::Core.Compiler.IncrementalCompact, idx::Int64, argexprs::Vector{…}, union_split::Core.Compiler.UnionSplit, boundscheck::Symbol, todo_bbs::Vector{…}, params::Core.Compiler.OptimizationParams)
    @ Core.Compiler base/compiler/ssair/inlining.jl:589
  [2] batch_inline!(ir::Core.Compiler.IRCode, todo::Vector{…}, propagate_inbounds::Bool, params::Core.Compiler.OptimizationParams)
    @ Core.Compiler base/compiler/ssair/inlining.jl:712
  [3] ssa_inlining_pass!
    @ Core.Compiler ./compiler/ssair/inlining.jl:81 [inlined]
  [4] run_passes_ipo_safe(ci::Core.CodeInfo, sv::Core.Compiler.OptimizationState{…}, caller::Core.Compiler.InferenceResult, optimize_until::Nothing)
    @ Core.Compiler base/compiler/optimize.jl:765
  [5]
    @ Core.Compiler base/compiler/typeinfer.jl:970
  [6] code_ircode_by_type(tt::Type; world::UInt64, interp::Core.Compiler.NativeInterpreter, optimize_until::Nothing)
    @ Base ./reflection.jl:1586
  [7] code_ircode_by_type
    @ Base ./reflection.jl:1572 [inlined]
  [8] code_ircode(f::Any, types::Any; kwargs::@Kwargs{})
    @ Base ./reflection.jl:1563
  [9] code_ircode(f::Any, types::Any)
    @ Base ./reflection.jl:1558
 [10] top-level scope
    @ REPL[6]:1

I guess the compiler is just not great at handling these ill-formed types? Should/can we prevent construction of them?

[vtjnash]

I added that change, and it seems to pass tests: https://github.com/JuliaLang/julia/compare/jn/51228. It is probably a reasonable fix, though it means we need to fix printing of NTuple