Calling CUDA.set_runtime_version!() with float parameter makes CUDA.jl unusable.

[jlk9]

When running a script using CUDA.jl, I called the function CUDA.set_runtime_version!(11.4), mistakenly supplying the argument 11.4 as a float instead of a string. Afterwords, when I tried to run any file that uses CUDA.jl (or add CUDA.jl on an interactive session of Julia), the following error message pops up:

---

ERROR: LoadError: MethodError: no method matching tryparse(::Type{VersionNumber}, ::Float64)
Closest candidates are:
tryparse(::Type{VersionNumber}, !Matched::AbstractString) at version.jl:122
Stacktrace:
[1] cuda_toolkit_tag()
@ Main ~/.julia/packages/CUDA_Runtime_jll/ZrwuA/.pkg/platform_augmentation.jl:29
[2] augment_platform!(platform::Platform)
@ Main ~/.julia/packages/CUDA_Runtime_jll/ZrwuA/.pkg/platform_augmentation.jl:137
[3] top-level scope
@ ~/.julia/packages/CUDA_Runtime_jll/ZrwuA/.pkg/select_artifacts.jl:11
[4] include(fname::String)
@ Base.MainInclude ./client.jl:476
[5] top-level scope
@ none:5
in expression starting at /home1/08955/jlk3675/.julia/packages/CUDA_Runtime_jll/ZrwuA/.pkg/select_artifacts.jl:11
ERROR: failed process: Process(`/home1/08955/jlk3675/julia-1.8.5/bin/julia -Cnative -J/home1/08955/jlk3675/julia-1.8.5/lib/julia/sys.so -g1 -O0 --color=no --history-file=no --startup-file=no --compiled-modules=yes --project=/home1/08955/jlk3675/.julia/environments/v1.8/Project.toml --eval 'append!(empty!(Base.DEPOT_PATH), ["/home1/08955/jlk3675/.julia", "/home1/08955/jlk3675/julia-1.8.5/local/share/julia", "/home1/08955/jlk3675/julia-1.8.5/share/julia"])
append!(empty!(Base.DL_LOAD_PATH), String[])

cd("/home1/08955/jlk3675/.julia/packages/CUDA_Runtime_jll/ZrwuA/.pkg")
include("/home1/08955/jlk3675/.julia/packages/CUDA_Runtime_jll/ZrwuA/.pkg/select_artifacts.jl")
' --startup-file=no x86_64-linux-gnu-libgfortran5-cxx11-libstdcxx30-julia_version+1.8.5`, ProcessExited(1)) [1]

Stacktrace:
[1] pipeline_error
@ ./process.jl:565 [inlined]
[2] read(cmd::Cmd)
@ Base ./process.jl:449
[3] collect_artifacts(pkg_root::String; platform::Base.BinaryPlatforms.Platform)
@ Pkg.Operations ~/julia-1.8.5/share/julia/stdlib/v1.8/Pkg/src/Operations.jl:630
[4] download_artifacts(env::Pkg.Types.EnvCache; platform::Base.BinaryPlatforms.Platform, julia_version::VersionNumber, verbose::Bool, io::Base.TTY)
@ Pkg.Operations ~/julia-1.8.5/share/julia/stdlib/v1.8/Pkg/src/Operations.jl:662
[5] add(ctx::Pkg.Types.Context, pkgs::Vector{Pkg.Types.PackageSpec}, new_git::Set{Base.UUID}; preserve::Pkg.Types.PreserveLevel, platform::Base.BinaryPlatforms.Platform)
@ Pkg.Operations ~/julia-1.8.5/share/julia/stdlib/v1.8/Pkg/src/Operations.jl:1282
[6] add(ctx::Pkg.Types.Context, pkgs::Vector{Pkg.Types.PackageSpec}; preserve::Pkg.Types.PreserveLevel, platform::Base.BinaryPlatforms.Platform, kwargs::Base.Pairs{Symbol, Base.TTY, Tuple{Symbol}, NamedTuple{(:io,), Tuple{Base.TTY}}})
@ Pkg.API ~/julia-1.8.5/share/julia/stdlib/v1.8/Pkg/src/API.jl:275
[7] add(pkgs::Vector{Pkg.Types.PackageSpec}; io::Base.TTY, kwargs::Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}})
@ Pkg.API ~/julia-1.8.5/share/julia/stdlib/v1.8/Pkg/src/API.jl:156
[8] add(pkgs::Vector{Pkg.Types.PackageSpec})
@ Pkg.API ~/julia-1.8.5/share/julia/stdlib/v1.8/Pkg/src/API.jl:145
[9] #add#27
@ ~/julia-1.8.5/share/julia/stdlib/v1.8/Pkg/src/API.jl:144 [inlined]
[10] add
@ ~/julia-1.8.5/share/julia/stdlib/v1.8/Pkg/src/API.jl:144 [inlined]
[11] try_prompt_pkg_add(pkgs::Vector{Symbol})
@ Pkg.REPLMode ~/julia-1.8.5/share/julia/stdlib/v1.8/Pkg/src/REPLMode/REPLMode.jl:717
[12] #invokelatest#2
@ ./essentials.jl:729 [inlined]
[13] invokelatest
@ ./essentials.jl:726 [inlined]
[14] check_for_missing_packages_and_run_hooks(ast::Any)
@ REPL ~/julia-1.8.5/share/julia/stdlib/v1.8/REPL/src/REPL.jl:176
[15] eval_user_input(ast::Any, backend::REPL.REPLBackend)
@ REPL ~/julia-1.8.5/share/julia/stdlib/v1.8/REPL/src/REPL.jl:146
[16] repl_backend_loop(backend::REPL.REPLBackend)
@ REPL ~/julia-1.8.5/share/julia/stdlib/v1.8/REPL/src/REPL.jl:247
[17] start_repl_backend(backend::REPL.REPLBackend, consumer::Any)
@ REPL ~/julia-1.8.5/share/julia/stdlib/v1.8/REPL/src/REPL.jl:232
[18] run_repl(repl::REPL.AbstractREPL, consumer::Any; backend_on_current_task::Bool)
@ REPL ~/julia-1.8.5/share/julia/stdlib/v1.8/REPL/src/REPL.jl:369
[19] run_repl(repl::REPL.AbstractREPL, consumer::Any)
@ REPL ~/julia-1.8.5/share/julia/stdlib/v1.8/REPL/src/REPL.jl:355
[20] (::Base.var"#967#969"{Bool, Bool, Bool})(REPL::Module)
@ Base ./client.jl:419
[21] #invokelatest#2
@ ./essentials.jl:729 [inlined]
[22] invokelatest
@ ./essentials.jl:726 [inlined]
[23] run_main_repl(interactive::Bool, quiet::Bool, banner::Bool, history_file::Bool, color_set::Bool)
@ Base ./client.jl:404
[24] exec_options(opts::Base.JLOptions)
@ Base ./client.jl:318
[25] _start()
@ Base ./client.jl:522

---

It appears calling CUDA.set_runtime_version!(11.4) changed the dictionary in Base.get_preferences(Base.UUID("76a88914-d11a-5bdc-97e0-2f5a05c973a2")) so that the key "version" now points to the float 11.4 instead of a string, causing this error. Removing and re-adding CUDA.jl did not fix it, in fact the error message now appears when I try to add CUDA.jl back so I cannot add the package. Manually changing the Base.get_preferences value to a string in an interactive Julia session does not work either.

To reproduce
Write a file called file.jl with these lines:

using CUDA
CUDA.set_runtime_version!(11.4)

Then run it using: julia file.jl
Now try to use the CUDA.jl package in another file or interactive session, and optionally try removing and adding back CUDA.jl

Version
I downloaded the most recent version of CUDA.jl when this error occurred. I do not know the exact specifics of its version and its dependencies because I tried removing and adding back CUDA.jl and can no longer add it. I was not using any other packages.

Expected behavior

I would expect calling CUDA.set_runtime_version!() with the wrong argument data type to return an error, and not permanently change a dict in Julia Base / CUDA_Runtime_jll that makes it incapable to subsequently use or add CUDA.jl.

Version info

Details on Julia:

Julia Version 1.8.5
Commit 17cfb8e65ea (2023-01-08 06:45 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 128 × AMD EPYC 7763 64-Core Processor
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-13.0.1 (ORCJIT, znver3)
  Threads: 1 on 128 virtual cores
Environment:
  LD_LIBRARY_PATH = 
  LD_PRELOAD = /opt/apps/xalt/xalt/lib64/libxalt_init.so

Details on CUDA:

I cannot post the output of CUDA.versioninfo() since I cannot add or use CUDA.jl.

Additional context

Add any other context about the problem here.

[jlk9]

Update: I was able to fix the incorrect version value datatype with the following:

using Preferences
CUDA_Runtime_jll_uuid = Base.UUID("76a88914-d11a-5bdc-97e0-2f5a05c973a2")
Preferences.delete_preferences!(CUDA_Runtime_jll_uuid, "version"; force=true)

This eliminated the faulty dictionary type, allowing me to add CUDA.jl back without throwing the above error.

[maleadt]

5c51766 & JuliaPackaging/Yggdrasil#6523