Closed Keno closed 6 months ago
This is kind of strange because the Base precompilation parses the TOML files and the precompilation happens after the TOML files are updated. So it is unclear to me how these packages gets found at all by Base package precompilation. Will investigate.
It seems the manifest is not pruned properly. After running the operations above there is still a deps.Enzyme
in the Manifest.toml even though nothing is referring to it.
Enzyme gets added as a dep to DiffEqBase here:
Must have something to do with that Enzyme being a weak dep. So deps_map
is wrong?
The manifest pruning issue seems to be a problem all the way back to 1.10... But even so, this still seems to indicate a regression in the Base precompilation code because I don't understand why it would ever "reach" the orphan Enzyme.
Attaching the problematic env here so I can easily look into that from home.
That only add the extension data so I don't see how that would do anything related to this.
This should not happen anymore after https://github.com/JuliaLang/Pkg.jl/pull/3864
In the following, the precompile after the second add should not have attempted to precompile
Enzyme
, because it got removed from the manifest. Indeed, and explicitprecompile
after theadd
succeeds. It looks like the precompile after theadd
may be issued with the wrong (outdated) list of packages.