segfault after overloading LinearAlgebra.mul! twice

[simeonschaub]

I can reliably segfault Julia with this reproducer: https://gist.github.com/simeonschaub/ff5930f6e5a368d0caf44731811ddb20, which encounters an illegal instruction. I have uploaded an rr trace here.

julia> versioninfo()
Julia Version 1.6.0-DEV.1489
Commit 13aac259a5 (2020-11-13 17:37 UTC)
Platform Info:
  OS: Linux (x86_64-pc-linux-gnu)
  CPU: Intel(R) Core(TM) i7-7500U CPU @ 2.70GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-11.0.0 (ORCJIT, skylake)
Environment:
  JULIA_NUM_THREADS = 2

This doesn't happen on 1.5.

[maleadt]

FWIW, reduced to:

struct a{b,d<:b} <: AbstractMatrix{b}
    e::d
end
Base.size(c) = 1
using LinearAlgebra
LinearAlgebra.mul!(AbstractVector, ::a, f, α, β) = return
try
  a([1 3 ]) * (1:2)
catch
end
LinearAlgebra.mul!(::AbstractVector, ::a, ::AbstractVector, ::Number, ::Number) = return
a([1 3 ]) * (1:2)

[vtjnash]

Yes, I worked out what was wrong when I self-assigned it. Here's a simplified example:

julia> f(::Int, ::Any) = 1
f (generic function with 1 method)

julia> f(::Any, ::Int) = 2
f (generic function with 2 methods)

julia> g(x) = f(x, x)
g (generic function with 1 method)

julia> g(1)
ERROR: MethodError: f(::Int64, ::Int64) is ambiguous. Candidates:
  f(::Int64, ::Any) in Main at REPL[1]:1
  f(::Any, ::Int64) in Main at REPL[2]:1
Possible fix, define
  f(::Int64, ::Int64)
Stacktrace:
 [1] g(x::Int64)
   @ Main ./REPL[3]:1
 [2] top-level scope
   @ REPL[4]:1

julia> f(::Int, ::Int) = 3
f (generic function with 3 methods)

julia> g(1)
signal (4): Illegal instruction: 4