issue 4: handle error in julia automatically

Author: KDr2

.gitignore

@@ -57,4 +57,5 @@ src/autogen-methods.jl
 csrc/build/
 csrc/libtorch/
 deps/lib/
+deps/logs/
 deps/build.log

README.md

@@ -14,8 +14,6 @@ A Julia interface for PyTorch's C++ backend.
    ```julia
     julia> using ThArrays
 
-    julia> handle_error_in_julia() # don't crash when error in libtorch
-
     julia> t = Tensor( -rand(3, 3) )
     PyTorch.Tensor{Float64, 2}:
     -0.1428 -0.7099 -0.1446

src/ThArrays.jl

@@ -11,6 +11,7 @@ const PROJECT_DIR = (@__DIR__) |> dirname
 function __init__()
     push!(Libdl.DL_LOAD_PATH, joinpath(PROJECT_DIR, "deps/lib"))
     Libdl.dlopen(joinpath(PROJECT_DIR, "deps/lib/libtorch_capi"))
+    @async handle_error_in_julia()
 end
 
 function handle_error_in_julia()

test/issues.jl

@@ -4,7 +4,6 @@ using Test
 @testset "Issues Regression" begin
 
     @testset "Issue 4" begin
-        handle_error_in_julia()
         Test.@test_throws ErrorException Tensor(3, requires_grad=true)
     end