CUDA app by PackageCompiler hangs at runtime on Ubuntu / Debian

[paulxshen]

Dummy app made by PackageCompiler with [email protected] (latest) as sole dep hangs at runtime on Ubuntu 24 and Debian 12. Works on Windows.

module Dummy
function julia_main()::Cint
    return 0
end
end

name = "Dummy"
uuid = "60a98398-2d19-4130-ae20-63172a4a42f2"

[deps]
CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"

using PackageCompiler
create_app("Dummy", "CompiledDummy")

Julia Version 1.11.2
Commit 5e9a32e7af2 (2024-12-01 20:02 UTC)
Build Info:
Official https://julialang.org/ release
Platform Info:
OS: Linux (x86_64-linux-gnu)
CPU: 16 × AMD EPYC 7R32
WORD_SIZE: 64
LLVM: libLLVM-16.0.6 (ORCJIT, znver2)
Threads: 1 default, 0 interactive, 1 GC (on 16 virtual cores)

(also failed on Debian 12 w/ Xeon, Nvidia T4)

I made this MWE after discovering a more complicated PackageCompiler app worked after removing CUDA. same thread as https://discourse.julialang.org/t/cuda-app-by-packagecompiler-hangs-at-runtime-on-ubuntu-debian/124995/3

[paulxshen]

Update: problem arises on Google Colab and not in AWS VMs. Problem persists with julia11.4 cuda.jl5.7 . But I need this to work on Colab! Happy to pay couple hours of consulting for anyone who can fix this problem.

Culprit may be LDPATH or runtime other than bundled w/ CUDA.jl getting loaded