Skip to content

Conversation

@rayegun
Copy link
Contributor

@rayegun rayegun commented Sep 22, 2022

Attempt to fix JuliaRegistries/General#68204

Might also need to be a BuildDependency in addition to Dependency

@stemann
Copy link
Contributor

stemann commented Sep 25, 2022

When this Torch JLL is registered (JuliaRegistries/General#68204), it might break current Torch.jl usages, as the C wrapper ("libtch") is not included - as there is no compat entry for Torch_jll in the currently registered Torch.jl.

@giordano
Copy link
Member

@gbaraldi I suspect JuliaPackaging/BinaryBuilderBase.jl#271 broke something:

ERROR: LoadError: TypeError: in keyword argument platforms, expected Vector{<:AbstractPlatform}, got a value of type Vector{Any}
Stacktrace:
 [1] top-level scope
   @ /agent/_work/1/s/T/Torch/build_tarballs.jl:247
in expression starting at /agent/_work/1/s/T/Torch/build_tarballs.jl:247

@gbaraldi
Copy link
Contributor

gbaraldi commented Sep 25, 2022

I will take a look.

Edit: It was that, it made it type unstable, so for empty arrays, which openblas_platforms is, it got Platforms[] but returned Any[]

@giordano giordano closed this Sep 25, 2022
@giordano giordano reopened this Sep 25, 2022
@giordano giordano merged commit dc06b99 into JuliaPackaging:master Sep 25, 2022
@giordano
Copy link
Member

When this Torch JLL is registered (JuliaRegistries/General#68204), it might break current Torch.jl usages, as the C wrapper ("libtch") is not included - as there is no compat entry for Torch_jll in the currently registered Torch.jl.

@stemann that should be fixed in the registry then (and in the package as well)

@giordano
Copy link
Member

@Wimmerer maybe next time don't use "fix <PR in the General registry>", this closed JuliaRegistries/General#68204 😉

@rayegun
Copy link
Contributor Author

rayegun commented Sep 25, 2022

Wait what, I figured that would have privileges preventing it.

Yeah sorry about this PR in general, I think I made it right before sleeping then forgot 🥲

# Dependency("TensorRT_jll"; platforms = cuda_platforms), # Building with TensorRT is not supported: https://github.com/pytorch/pytorch/issues/60228
Dependency("XNNPACK_jll"; compat = "0.0.20210622"),
BuildDependency(PackageSpec("protoc_jll", Base.UUID("c7845625-083e-5bbe-8504-b32d602b7110"), v"3.13.0")),
Dependency(PackageSpec("protoc_jll", Base.UUID("c7845625-083e-5bbe-8504-b32d602b7110")); compat="3.13.0"),
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

JuliaRegistries/General#68946 sounds like protoc breaks the ABI across minor versions. What a mess

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants