Segmentation fault when creating CuArray of CuArray

[DesmondZhong]

Describe the bug
Segmentation fault when creating CuArray of CuArray

To Reproduce
The Minimal Working Example (MWE) for this bug:

using CuArrays
x1 = cu(rand(5))
x2 = cu(rand(5))
x = [x1, x2] # Array of CuArray
convert(CuArray, [x1, x2])

I got the following error:

2-element CuArray{CuArray{Float32,1},1}:
 Float32[0.669814, 0.810592, 0.705412, 0.670489, 0.720598]
signal (11): Segmentation fault
in expression starting at no file:0
jl_alloc_array_1d at /usr/bin/../lib/libjulia.so.1 (unknown line)
unknown function (ip: 0x7fcb38be7d40)
Type at ./dict.jl:110
d at /home/desmond/.julia/packages/Lazy/ZAeCx/src/macros.jl:241 [inlined]
render at /home/desmond/.julia/packages/Atom/E4PBh/src/display/view.jl:31
render at /home/desmond/.julia/packages/Media/ItEPc/src/system.jl:173
render at /home/desmond/.julia/packages/Media/ItEPc/src/system.jl:173
jl_fptr_trampoline at /usr/bin/../lib/libjulia.so.1 (unknown line)
jl_apply_generic at /usr/bin/../lib/libjulia.so.1 (unknown line)
render at /home/desmond/.julia/packages/Atom/E4PBh/src/display/view.jl:49
render at /home/desmond/.julia/packages/Atom/E4PBh/src/display/display.jl:19 [inlined]
displayandrender at /home/desmond/.julia/packages/Atom/E4PBh/src/display/showdisplay.jl:127
jl_fptr_trampoline at /usr/bin/../lib/libjulia.so.1 (unknown line)
jl_apply_generic at /usr/bin/../lib/libjulia.so.1 (unknown line)
JuliaGPU/CuArrays.jl#117 at /home/desmond/.julia/packages/Atom/E4PBh/src/eval.jl:102
jl_apply_generic at /usr/bin/../lib/libjulia.so.1 (unknown line)
jl_f__apply at /usr/bin/../lib/libjulia.so.1 (unknown line)
jl_f__apply_latest at /usr/bin/../lib/libjulia.so.1 (unknown line)
macro expansion at ./essentials.jl:742 [inlined]
JuliaGPU/CuArrays.jl#113 at /home/desmond/.julia/packages/Atom/E4PBh/src/eval.jl:86
jl_apply_generic at /usr/bin/../lib/libjulia.so.1 (unknown line)
handlemsg at /home/desmond/.julia/packages/Atom/E4PBh/src/comm.jl:164
unknown function (ip: 0x7fcb2af2376c)
jl_apply_generic at /usr/bin/../lib/libjulia.so.1 (unknown line)
jl_f__apply at /usr/bin/../lib/libjulia.so.1 (unknown line)
JuliaGPU/CuArrays.jl#19 at ./task.jl:259
unknown function (ip: 0x7fcb2afc6574)
jl_apply_generic at /usr/bin/../lib/libjulia.so.1 (unknown line)
unknown function (ip: 0x7fcb43df3617)
unknown function (ip: 0xffffffffffffffff)
Allocations: 72891018 (Pool: 72878366; Big: 12652); GC: 153

Julia has exited. Press Enter to start a new session.

Apparently, it tried to create a CuArray of CuArray, but some segmentation fault is hit along the way. When I run this code, sometimes the error occurs before the line 2-element CuArray{CuArray{Float32,1},1}: Float32[0.669814, 0.810592, 0.705412, 0.670489, 0.720598] shows up in the REPL.

Expected behavior
expect to create a 2-element CuArray{CuArray{Float32,1},1}:

Build log

Building ZipFile ─────────→ `~/.julia/packages/ZipFile/YHTbb/deps/build.log`
  Building Arpack ──────────→ `~/.julia/packages/Arpack/UiiMc/deps/build.log`
┌ Error: Error building `Arpack`: 
│ ERROR: LoadError: LibraryProduct(nothing, ["libarpack"], :libarpack, "Prefix(/home/desmond/.julia/packages/Arpack/UiiMc/deps/usr)") is not satisfied, cannot generate deps.jl!
│ Stacktrace:
│  [1] #write_deps_file#152(::Bool, ::Function, ::String, ::Array{LibraryProduct,1}) at /home/desmond/.julia/packages/BinaryProvider/TcAwt/src/Products.jl:414
│  [2] (::getfield(BinaryProvider, Symbol("#kw##write_deps_file")))(::NamedTuple{(:verbose,),Tuple{Bool}}, ::typeof(write_deps_file), ::String, ::Array{LibraryProduct,1}) at ./none:0
│  [3] top-level scope at none:0
│  [4] include(::String) at ./client.jl:403
│  [5] top-level scope at none:0
│ in expression starting at /home/desmond/.julia/packages/Arpack/UiiMc/deps/build.jl:74
└ @ Pkg.Operations /build/julia/src/julia-1.1.0/usr/share/julia/stdlib/v1.1/Pkg/src/Operations.jl:1075
  Building SpecialFunctions → `~/.julia/packages/SpecialFunctions/fvheQ/deps/build.log`
  Building Rmath ───────────→ `~/.julia/packages/Rmath/Py9gH/deps/build.log`
  Building Conda ───────────→ `~/.julia/packages/Conda/CpuvI/deps/build.log`
  Building CUDAdrv ─────────→ `~/.julia/packages/CUDAdrv/3cR2F/deps/build.log`
  Building LLVM ────────────→ `~/.julia/packages/LLVM/tg8MX/deps/build.log`
  Building CUDAnative ──────→ `~/.julia/packages/CUDAnative/wU0tS/deps/build.log`
  Building FFTW ────────────→ `~/.julia/packages/FFTW/p7sLQ/deps/build.log`
  Building CuArrays ────────→ `~/.julia/packages/CuArrays/PwSdF/deps/build.log`
  Building CodecZlib ───────→ `~/.julia/packages/CodecZlib/9jDi1/deps/build.log`

Environment details (please complete this section)
Details on Julia:

Julia Version 1.1.0
Commit 80516ca202 (2019-01-21 21:24 UTC)
Platform Info:
  OS: Linux (x86_64-pc-linux-gnu)
  CPU: Intel(R) Core(TM) i7-3740QM CPU @ 2.70GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-6.0.1 (ORCJIT, ivybridge)
Environment:
  JULIA_EDITOR = atom -a
  JULIA_NUM_THREADS = 4

Julia packages:

• CUDAdrv
• CUDAnative
• CuArrays
• Flux

CUDA version: 10.1
Driver Version: 418.56

Let me know if there's any other information I should provide! Thank you!

[kshyatt]

FWIW when I try to run this MWE now:

julia> using CUDA

julia> x1 = cu(rand(5));

julia> x2 = cu(rand(5));

julia> convert(CuArray, [x1, x2])
ERROR: CuArray only supports bits types
Stacktrace:
 [1] error(::String) at ./error.jl:33
 [2] CuArray{CuArray{Float32,1,Nothing},1,P} where P(::UndefInitializer, ::Tuple{Int64}) at /home/kshyatt/.julia/dev/CUDA/src/array.jl:118
 [3] CuArray{CuArray{Float32,1,Nothing},N,P} where P where N(::UndefInitializer, ::Tuple{Int64}) at /home/kshyatt/.julia/dev/CUDA/src/array.jl:127
 [4] similar(::Type{CuArray{CuArray{Float32,1,Nothing},N,P} where P where N}, ::Tuple{Int64}) at ./abstractarray.jl:671
 [5] convert(::Type{CuArray}, ::Array{CuArray{Float32,1,Nothing},1}) at /home/kshyatt/.julia/packages/GPUArrays/X4SqE/src/host/construction.jl:95
 [6] top-level scope at REPL[9]:1

Which seems like a reasonable error to throw? Can we close this?

[maleadt]

Yup, that seems entirely reasonable to me!