Segfault with the Vararg datatype

[jdlangs]

I was playing around with the new tuple syntax and came across this crash. Is the method definition itself nonsensical here?

julia> f(x::Vararg{Int}) = length(x)
f (generic function with 1 method)

julia> f()
ERROR: AssertionError: la == 0
 in typeinf_uncached at ./inference.jl:1664
 in typeinf at ./inference.jl:1304
 in typeinf_ext at ./inference.jl:1248

julia> f(1)

signal (11): Segmentation fault
unknown function (ip: -924344832)
unknown function (ip: -924134164)
unknown function (ip: -924133876)
jl_trampoline at /home/jdlangs/Documents/Projects/julia/usr/bin/../lib/libjulia.so (unknown line)
jl_apply_generic at /home/jdlangs/Documents/Projects/julia/usr/bin/../lib/libjulia.so (unknown line)
unknown function (ip: -924032813)
unknown function (ip: -924035646)
unknown function (ip: -923959389)
jl_toplevel_eval_in at /home/jdlangs/Documents/Projects/julia/usr/bin/../lib/libjulia.so (unknown line)
eval_user_input at REPL.jl:60
jlcall_eval_user_input_43978 at �% (unknown line)
jl_apply_generic at /home/jdlangs/Documents/Projects/julia/usr/bin/../lib/libjulia.so (unknown line)
anonymous at task.jl:89
unknown function (ip: -924008416)
unknown function (ip: 0)
Segmentation fault (core dumped)

[mbauman]

There's a little magic the parser does to prepare methods with the ... syntax that's not yet incorporated into the Vararg type. It'll be fixed with #10911.

[jdlangs]

Cool, I'll look forward to trying that out once it's merged.

[StefanKarpinski]

I feel like a segfault on master is still a bug, no?

[pao]

Yes.

[mbauman]

I think @timholy's commit 76b2b26 in #10911 will stand on its own and can be committed before the rest of that PR. This is also an issue on 0.3, but I don't think anyone ever tried using Vararg before the tuple overhaul.

[jdlangs]

I wasn't sure, so thanks for clarifying the issue policy there.

[timholy]

Yes, 76b2b26 probably would stand on its own. If people are in a rush I can cherry-pick it.

[JeffBezanson]

Sounds like a good idea to me. Fixing segfaults is always good!