Reproducible "invalid age range update" in v0.6.0

[rdeits]

I'm seeing an issue that looks similar to #22351 and #22355 but I happen to have a non-private reproducible example, so I thought I would share it. It's annoyingly complicated, but at least it should be something you can just run and immediately reproduce locally. The complete code to reproduce is:

#!/bin/bash

set -ex

export JULIA_PKGDIR="`pwd`/packages"
echo $JULIA_PKGDIR
mkdir packages
julia -e "Pkg.init()"
julia -e "Pkg.add(\"BenchmarkTools\")"
julia -e "Pkg.add(\"MultivariatePolynomials\")"
julia -e "Pkg.clone(\"git@github.com:rdeits/TypedPolynomials.jl\")"
pushd packages/v0.6/MultivariatePolynomials
git checkout 777f71792a2c25dc4227886e125af4b2ecda1332
popd
pushd packages/v0.6/TypedPolynomials
git checkout e8eea2c0ae45a52c516dcaee6555daaa67933973
popd

julia -e "Pkg.test(\"TypedPolynomials\")"

gives:

WARNING: An error occurred during inference. Type inference is now partially disabled.
Base.MethodError(f=typeof(Core.Inference.convert)(), args=(Base.AssertionError, "invalid age range update"), world=0x0000000000000ac2)

This occurs for me when using:

➜ julia -E "versioninfo()"
Julia Version 0.6.0
Commit 903644385b (2017-06-19 13:05 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin13.4.0)
  CPU: Intel(R) Core(TM) i7-2860QM CPU @ 2.50GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Sandybridge)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, sandybridge)
nothing

please let me know if there's anything I can do to help debug further. The full output is at https://gist.github.com/rdeits/425db367e198d186b9dfadca0086b8cb .

[rdeits]

BTW, this issue only cropped up when I started using https://github.com/simonster/Reexport.jl . The TypedPolynomials commit 5b9cc52dbd33a7f933795538ed3136107eb5451c is identical except for the lack of Reexport.jl and does not exhibit the crash.

[tkelman]

Does it have the same bisect conclusion as #22355 (comment) ?

(would be preferable to use a gist for really long backtraces) done thanks

[rdeits]

Currently building Julia to try and find out. And I've moved the output to a gist 😄

[rdeits]

Unfortunately I can't run the test code on master to try to bisect because I can't precompile StaticArrays:

fatal: error thrown and no exception handler available.
Base.ArgumentError(msg="No StaticArray found in argument list")
rec_backtrace at /Users/rdeits/Projects/julia/src/stackwalk.c:86
record_backtrace at /Users/rdeits/Projects/julia/src/task.c:246 [inlined]
jl_throw at /Users/rdeits/Projects/julia/src/task.c:568
same_size at /Users/rdeits/Downloads/inference_bug/packages/v0.7/StaticArrays/src/traits.jl:117 [inlined]
map at /Users/rdeits/Downloads/inference_bug/packages/v0.7/StaticArrays/src/mapreduce.jl:10
stacktrace at ./stacktraces.jl:151
display_error at ./client.jl:128
display_error at ./client.jl:142
do_call at /Users/rdeits/Projects/julia/src/interpreter.c:70
eval at /Users/rdeits/Projects/julia/src/interpreter.c:262
eval_body at /Users/rdeits/Projects/julia/src/interpreter.c:539
jl_toplevel_eval_body at /Users/rdeits/Projects/julia/src/interpreter.c:511
jl_toplevel_eval_flex at /Users/rdeits/Projects/julia/src/toplevel.c:602
jl_toplevel_eval_in at /Users/rdeits/Projects/julia/src/builtins.c:505
eval at ./sysimg.jl:48
_start at ./client.jl:429
true_main at /Users/rdeits/Projects/julia/usr/bin/julia (unknown line)
main at /Users/rdeits/Projects/julia/usr/bin/julia (unknown line)
ERROR: LoadError: Failed to precompile StaticArrays to /Users/rdeits/Downloads/inference_bug/packages/lib/v0.7/StaticArrays.ji.
Stacktrace:
 [1] compilecache(::String) at ./loading.jl:598
 [2] _require(::Symbol) at ./loading.jl:387
 [3] require(::Symbol) at ./loading.jl:318
 [4] include_relative(::Module, ::String) at ./loading.jl:464
 [5] include(::Module, ::String) at ./sysimg.jl:14
 [6] anonymous at ./<missing>:2

[KristofferC]

Please comment if this is still reproducible on newer Julias.