Possible dispatch error leading to Julia freezing

[mlhetland]

I've come across what seems to me to be a bug in Julia, related to dispatch, I guess. I've reduce the problem to a couple of simple examples; one is given below, and another (with more undefined variables etc. that go undetected) can be found in the discussion at Julia-Users. In my original code, the problem is that the dispatch seems to go to the wrong function, creating infinite mutual recursion. In the simplified code below, I've removed the alternative function (and other scaffolding code), so the expected behavior would be a method error, I guess?

type B
end

type X{T}
    baz::Vector{T}
end

function foo(a, c)
    bar(b, c)
end

function bar{T}(d::(X{T}, Int), val::T)
    foo("err", d[1].baz)
end

b = B()

function test()
    bar(b, "wat")
end

test()

When I run this, I would have expected:

ERROR: no method bar(B, ASCIIString)

Instead, the code just hangs. I'm not sure if it's really looping between foo and bar—I haven't been able to produce any evidence of that, really. Printing (with flushing) doesn't work, and undefined variables are ignored (c.f., the longer example on Julia-Users). Could it be some loop/mutual recursion going on in Julia itself (in the inference/dispatch)?

Here's my versioninfo():

Julia Version 0.3.0-prerelease+3884
Commit 3e6a6c7* (2014-06-25 10:41 UTC)
Platform Info:
  System: Darwin (x86_64-apple-darwin12.5.0)
  CPU: Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz
  WORD_SIZE: 64
  BLAS: libgfortblas
  LAPACK: liblapack
  LIBM: libopenlibm

[jmert]

I think I might have found the same issue through a different means. I've attached the minimum working example I've been able to hobble together, a snippet of logs that my kernel spits out, and an strace of the process right before I enter the last line of the MWE through the hung point.

The process hangs to the point that I've only been able to kill the process by rebooting my computer entirely. Trying to get the PID through ps ends up hanging itself, and when I take the PID before hanging julia, I also can't kill -9 the process successfully.

strace: http://pastebin.com/iSJ75P1V
kernel messages: http://pastebin.com/8tUVJdMp
MWE: http://pastebin.com/FZLsUk3A

[ihnorton]

Original test case is fixed on latest master. jmert's case stackoverflows, but line 29 doesn't really make sense.

needs-test before closing?