StackOverflowError in `JavaCall.init` for Julia 1.1.0

[KenziTrader]

               _
   _       _ _(_)_     |  Documentation: https://docs.julialang.org
  (_)     | (_) (_)    |
   _ _   _| |_  __ _   |  Type "?" for help, "]?" for Pkg help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 1.1.0 (2019-01-21)
 _/ |\__'_|_|_|\__'_|  |  Official https://julialang.org/ release
|__/                   |

julia> using JavaCall

julia> JavaCall.init(["-Xmx128M"])

signal (11): Segmentation fault: 11
in expression starting at no file:0
unknown function (ip: 0x122b242b3)
Allocations: 2501680 (Pool: 2501183; Big: 497); GC: 4
ERROR: StackOverflowError:
Stacktrace:
 [1] init(::Array{String,1}) at /Users/rene/.julia/packages/JavaCall/toamy/src/jvm.jl:176
 [2] top-level scope at none:0

[karikoskinen]

Similar problem here, but instead of segfault, is just hangs indefinitely. By putting those lines in a file and running julia, it works.

$ cat initjavacall.jl
using JavaCall
JavaCall.init(["-Xmx128m"])
println("JavaCall initialized")

$ julia -i initjavacall.jl
JavaCall initialized
               _
   _       _ _(_)_     |  Documentation: https://docs.julialang.org
  (_)     | (_) (_)    |
   _ _   _| |_  __ _   |  Type "?" for help, "]?" for Pkg help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 1.1.0 (2019-01-21)
 _/ |\__'_|_|_|\__'_|  |  Official https://julialang.org/ release
|__/                   |

julia>

[aviks]

This seems to be a bigger issue than I thought. For now, I would advice all users of JavaCall and its derivatives stick to Julia 1.0.x

[ExpandingMan]

Great news all, Jeff just merged a PR he made for 1.3. So it looks like we'll be back up and running with 1.3. I'm not quite sure what the implications are of setting this environment variable, at the very least I think it means somewhat inefficient multi-threading. However, at least it will get applications that need to use the JVM unstuck.

[kcajf]

Hadn't noticed that PR - great news. I would also be interested in hearing from someone with a bit more knowledge about the system as to how JULIA_COPY_STACKS=1 will affect performance on different threading workloads.

[aviks]

Guess what? It looks like JDK 11 works with all versions of Julia. Proof: https://travis-ci.org/JuliaInterop/JavaCall.jl/builds/588468061

[mkitti]

Using openjdk version 11.0.6 and Julia 1.3.1 makes JavaCall work with scripts. However, once the REPL starts, things do not work.

I slightly expanded the script above to include a test from the documentation.

$ cat initjavacall.jl 
using JavaCall
JavaCall.init(["-Xmx128m"])
println("JavaCall initialized")
jlm = @jimport java.lang.Math
out = jcall(jlm, "sin", jdouble, (jdouble,), pi/2)
println(out)

The test appears to work correctly on startup. However, once the REPL gets going, the call no longer seems to work correctly.

$ julia --banner=no -i initjavacall.jl 
JavaCall initialized
1.0
julia> jcall(jlm, "sin", jdouble, (jdouble,), pi/2)
0.0

Here is the openjdk version I am using on Ubuntu 18.04.

openjdk version "11.0.6" 2020-01-14
OpenJDK Runtime Environment (build 11.0.6+10-post-Ubuntu-1ubuntu118.04.1)
OpenJDK 64-Bit Server VM (build 11.0.6+10-post-Ubuntu-1ubuntu118.04.1, mixed mode, sharing)

[mkitti]

Setting JULIA_COPY_STACKS=yes makes things work properly.

$ JULIA_COPY_STACKS=yes julia --banner=no -i initjavacall.jl 
JavaCall initialized
1.0
julia> jcall(jlm, "sin", jdouble, (jdouble,), pi/2)
1.0

[aviks]

So yes, to use JavaCall, you need to use Julia 1.0, or Julia 1.3 with JULIA_COPY_STACKS=1. This is due to how Java uses the stack, and a change to Julia's stack mechanism in 1.1 to improve threading performance. This is now documented, and using JavaCall will warn you when running in an unsupported environment. Thanks for bearing with us on this - but I think now the situation is settled.