Crash with empty let bindings

[didierverna]

This works fine:

didier(s001)% julia
               _
   _       _ _(_)_     |  A fresh approach to technical computing
  (_)     | (_) (_)    |  Documentation: http://docs.julialang.org
   _ _   _| |_  __ _   |  Type "?help" for help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 0.4.5 (2016-03-18 00:58 UTC)
 _/ |\__'_|_|_|\__'_|  |  Official http://julialang.org/ release
|__/                   |  x86_64-apple-darwin13.4.0

julia> let x
       x
       end
ERROR: UndefVarError: x not defined

This doesn't:

didier(s001)% julia
               _
   _       _ _(_)_     |  A fresh approach to technical computing
  (_)     | (_) (_)    |  Documentation: http://docs.julialang.org
   _ _   _| |_  __ _   |  Type "?help" for help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 0.4.5 (2016-03-18 00:58 UTC)
 _/ |\__'_|_|_|\__'_|  |  Official http://julialang.org/ release
|__/                   |  x86_64-apple-darwin13.4.0

julia> let x
       foo() = x
       foo()
       end

signal (11): Segmentation fault: 11
jl_f_tuple at /Users/osx/buildbot/slave/package_osx10_9-x64/build/src/builtins.c:661
eval_user_input at REPL.jl:65
jlcall_eval_user_input_21170 at  (unknown line)
jl_apply at /Users/osx/buildbot/slave/package_osx10_9-x64/build/src/gf.c:1691
anonymous at REPL.jl:92
jl_apply at /Users/osx/buildbot/slave/package_osx10_9-x64/build/src/./julia.h:1331
zsh: segmentation fault (core dumped)  julia

[Keno]

Seems to be fixed on master:

julia> let x
          foo() = x
          foo()
          end
ERROR: UndefVarError: x not defined
 in eval(::Module, ::Any) at ./boot.jl:237

maybe somebody knowns which commit this was and we can slate it for backporting.

[StefanKarpinski]

Seems likely to have been fixed by jb/functions, which is not backportable. I wonder if inference is incorrectly assuming that let variables are initialized.

[vtjnash]

i think this is #265. the result of calling foo() here is supposed to be a MethodError, but it incorrectly returns an answer instead

[Keno]

How is it supposed to be a MethodError? x is undefined, so it's an UndefVarError.

[StefanKarpinski]

Regardless of what kind of error it's supposed to be, segfault is not ok.

[vtjnash]

ah, sorry, i was thinking foo was global, in which case this is odd edge case for parsing such that foo would be defined, but not have any methods until the let block returns.

[tkelman]

should be fixed by #15851