Unreachable reached, Julia 1.2.0-rc2.0

[quinnj]

julia> CartesianIndex(BigInt(1), 1)
Unreachable reached at 0x117746b26

signal (4): Illegal instruction: 4
in expression starting at REPL[4]:1
Type at ./multidimensional.jl:68
do_call at /Users/julia/buildbot/worker/package_macos64/build/src/interpreter.c:323
eval_stmt_value at /Users/julia/buildbot/worker/package_macos64/build/src/interpreter.c:362 [inlined]
eval_body at /Users/julia/buildbot/worker/package_macos64/build/src/interpreter.c:758
jl_interpret_toplevel_thunk_callback at /Users/julia/buildbot/worker/package_macos64/build/src/interpreter.c:884
unknown function (ip: 0xfffffffffffffffe)
unknown function (ip: 0x11907c70f)
unknown function (ip: 0x0)
jl_interpret_toplevel_thunk at /Users/julia/buildbot/worker/package_macos64/build/src/interpreter.c:893
jl_toplevel_eval_flex at /Users/julia/buildbot/worker/package_macos64/build/src/toplevel.c:815
jl_toplevel_eval_flex at /Users/julia/buildbot/worker/package_macos64/build/src/toplevel.c:764
jl_toplevel_eval at /Users/julia/buildbot/worker/package_macos64/build/src/toplevel.c:824 [inlined]
jl_toplevel_eval_in at /Users/julia/buildbot/worker/package_macos64/build/src/toplevel.c:844
eval at ./boot.jl:330
eval_user_input at /Users/julia/buildbot/worker/package_macos64/build/usr/share/julia/stdlib/v1.2/REPL/src/REPL.jl:86
run_backend at /Users/jacobquinn/.julia/packages/Revise/agmgx/src/Revise.jl:949
#75 at ./task.jl:268
jl_apply at /Users/julia/buildbot/worker/package_macos64/build/src/./julia.h:1606 [inlined]
start_task at /Users/julia/buildbot/worker/package_macos64/build/src/task.c:596
Allocations: 8132905 (Pool: 8131258; Big: 1647); GC: 18
[1]    10138 illegal hardware instruction  julia
➜  dev

[bkamins]

On 1.3.0-DEV.542 Win it works OK (at least for me):

julia>  CartesianIndex(BigInt(1), 1)
CartesianIndex(1, 1)

[fredrikekre]

I can reproduce this on Linux (1.2.0-rc2) so I don't think it is related to the OS.

[vchuravy]

Can't reproduce on Version 1.3.0-DEV.468 (2019-06-30) Linux, so might be fixed on master?

[goggle]

I can confirm this.

On v1.2.0-rc2:

julia> CartesianIndex(BigInt(1), 1)
Unreachable reached at 0x7fe48d9c584f

signal (4): Illegal instruction
in expression starting at REPL[1]:1
Type at ./multidimensional.jl:68
do_call at /home/alex/Projects/github/goggle/julia/src/interpreter.c:323
eval_value at /home/alex/Projects/github/goggle/julia/src/interpreter.c:411
eval_stmt_value at /home/alex/Projects/github/goggle/julia/src/interpreter.c:362 [inlined]
eval_body at /home/alex/Projects/github/goggle/julia/src/interpreter.c:754
jl_interpret_toplevel_thunk_callback at /home/alex/Projects/github/goggle/julia/src/interpreter.c:884
unknown function (ip: 0xfffffffffffffffe)
unknown function (ip: 0x7fe4a77a558f)
unknown function (ip: (nil))
jl_interpret_toplevel_thunk at /home/alex/Projects/github/goggle/julia/src/interpreter.c:893
jl_toplevel_eval_flex at /home/alex/Projects/github/goggle/julia/src/toplevel.c:815
jl_toplevel_eval_flex at /home/alex/Projects/github/goggle/julia/src/toplevel.c:764
jl_toplevel_eval_in at /home/alex/Projects/github/goggle/julia/src/toplevel.c:844
eval at ./boot.jl:330
eval_user_input at /home/alex/Projects/github/goggle/julia/usr/share/julia/stdlib/v1.2/REPL/src/REPL.jl:86
macro expansion at /home/alex/Projects/github/goggle/julia/usr/share/julia/stdlib/v1.2/REPL/src/REPL.jl:118 [inlined]
#26 at ./task.jl:268
jl_apply at /home/alex/Projects/github/goggle/julia/src/julia.h:1606 [inlined]
start_task at /home/alex/Projects/github/goggle/julia/src/task.c:596
unknown function (ip: 0xffffffffffffffff)
Allocations: 1883706 (Pool: 1883236; Big: 470); GC: 3

On the current master (1.3.0-DEV.557):

julia> CartesianIndex(BigInt(1), 1)
CartesianIndex(1, 1)

Both version are build from source on a 64-bit Linux machine.

[Keno]

Bisect says this got fixed with #31698.

[JeffBezanson]

Bisecting in the other direction, the first bad commit was 8c44566.

[JeffBezanson]

cc @vtjnash

[vtjnash]

Reduced to:

julia> typeintersect(Type{<:Tuple{Any,Vararg{Any}}}, Type{Tuple{Vararg{Any,A}}} where A)
Union{}

(which should have solutions for A > 1)

[JeffBezanson]

So this was not caused by the commit pointed to by bisect, just exposed by it due to an improvement in inference.