Try resolve before removing manifest when instantiating fails

[aplavin]

Whenever Pluto cannot instantiate the env, it removes the manifest completely:

Operation failed. Updating registries and trying again...
  exception = 
    Unsatisfiable requirements detected for package Statistics [10745b16]:
     Statistics [10745b16] log:
     ├─possible versions are: 1.11.0 - 1.11.1 or uninstalled
     └─restricted to versions 1.10.0 by an explicit requirement — no versions left
    Updating registry at `~/.julia/registries/General.toml`
Warning: Operation failed. Removing Manifest and trying again...
  exception = 
    Unsatisfiable requirements detected for package Statistics [10745b16]:
     Statistics [10745b16] log:
     ├─possible versions are: 1.11.0 - 1.11.1 or uninstalled
     └─restricted to versions 1.10.0 by an explicit requirement — no versions left

I believe the most common scenario when it happens is a Julia version update (like in this example).

This automatic behavior is much more destructive compared to the Julia REPL suggestion of running ]resolve. Removing manifest leads to up/downgrading of all dependencies, basically whatever the Pkg resolver wants to do from scratch.

Can Pluto at least try ]resolve first to keep the manifest as long as possible?
And potentially – ask the user for an explicit confirmation before performing the destructive action of dropping the manifest...

[fonsp]

Can you share a notebook file that we can test with?

[fonsp]

julia 1.11 notebook that fails on julia 1.10.jl.zip

[fonsp]

Pluto does do resolve when loading a notebook. You can see this in the Status tab (>0 sec time spent on resolve) and in the pkg logs.

But it looks like resolving a Julia 1.11 environment in 1.10 does not always work. My notebook above had the following issue:

JuliaLang/Pkg.jl#4058

[fonsp]

We already try resolve before removing the manifest. This particular issue was caused by JuliaLang/Pkg.jl#4058 which will be fixed in upcoming 1.10 and 1.11 releases