Here is a MWE that leads Julia to silently exit:
run(`git clone https://github.com/mohamed82008/juliaexits`)
using Pkg
Pkg.activate("juliaexits")
Pkg.instantiate()
using Distributions, LinearAlgebra
A = rand(3, 3); A = Matrix{Real}(A + A' + 3I)
MvNormal(rand(3), A)
I am using a custom PDMats.jl. My best guess is that the error happens in this function https://github.com/mohamed82008/PDMats.jl/blob/pdmat_constr/src/pdmat.jl#L9.
Can someone reproduce this?
Did you try with --check-bounds=yes?
Yes, same result.
Bump.
Bump.
@mohamed82008 sorry to hear you are running into this. I'm not a Julia maintainer, so I can't help you, but I can advice you that many people will be hesitant to clone + run code from a git repository in this way.
It's not overly paranoid to worry: after all, the Julia maintainers have commit access to this repository, and are able to upload Julia binaries. If those get compromised, that gives access to hundreds of thousands of systems, some of which process sensitive data.
Here's some other things you can do:
gdb and paste a backtrace (assuming the silent exit is a crash)Hope that helps!
Oh wow, that kind of defeats the purpose of Manifest files for bug reports. I understand the concerns. I will try to reduce this to a MWE. Thanks for your advice.
Most helpful comment
@mohamed82008 sorry to hear you are running into this. I'm not a Julia maintainer, so I can't help you, but I can advice you that many people will be hesitant to clone + run code from a git repository in this way.
It's not overly paranoid to worry: after all, the Julia maintainers have commit access to this repository, and are able to upload Julia binaries. If those get compromised, that gives access to hundreds of thousands of systems, some of which process sensitive data.
Here's some other things you can do:
gdband paste a backtrace (assuming the silent exit is a crash)Hope that helps!