Add property to disable building all libs packages

[safern]

Contributes towards: dotnet/runtimelab#465

There are certain experiments in dotnet/runtimelab that don't need to build any libraries packages, we currently only offered a hook to disabled OOB packages.

cc: @jkotas

[ghost]

Tagging subscribers to this area: @safern, @ViktorHofer
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Contributes towards: dotnet/runtimelab#465

There are certain experiments in dotnet/runtimelab that don't need to build any libraries packages, we currently only offered a hook to disabled OOB packages.

cc: @jkotas

Author:	safern
Assignees:	-
Labels:	area-Infrastructure-libraries
Milestone:	-

[ViktorHofer]

Why would this project then even be invoked if it doesn't do anything?

[safern]

Why would this project then even be invoked if it doesn't do anything?

It is the simplest way to flip the switch, other than removing libs.packages from the default libs subsets. I would want people that create a new experiment to only have to change 1 flag, other than having to update yml files or remove stuff from default settings.

[ViktorHofer]

It is the simplest way to flip the switch, other than removing libs.packages from the default libs subsets

Removing libs.packages from the default subsets would be the right thing to avoid the unnecessary restore, evaluation and build of libraries-packages.proj. That should just be a one line change in Subsets.props.

[ViktorHofer]

Especially the restore can be significant because of the project restoring the to be harvested packages as well.

[safern]

That complicates things. There are experiments that want to build 1 OOB package, but not other packages. There are some experiments that want to build, framework packages but not OOB packages. There are experiments that want to build no libraries packages at all. These are just experiments, so I think the simplest for all kind of experiments the better, having both switches in the same place is in my opinion the easiest way to communicate, rather than having to change different files and things based on the needs of the experiment.

I was going to go with removing the libs.packages subset from the default subsets to begin with, but then that will, might change in the future where it might not be part of the default, subset, etc and it will be controlled by what the yml calls (which target it invokes), so then that complicates more and more.

[safern]

Especially the restore can be significant because of the project restoring the to be harvested packages as well.

We could disable the harvesting if we're not building packages at all.

[ViktorHofer]

With @Anipik's change #45564 you can specify IsPackable to indicate if the pack target should be invoked at all.

In regards to experiments:

• No packages should be built: set <IsPackable>false</IsPackable> in the libraries' Directory.Build.props file.
• Some packages should be built: same as above + set <IsPackable>true</IsPackable> in the leafs that should be built.
• All packages should be built: no change needed.

[safern]

Ok, then I'll wait for @Anipik's change to go in. In the meantime I'll just merge this in runtimelab in the experiments that are building extra packages and causing the publish issue to unblock the official build.

[ViktorHofer]

Sounds good 👍 thanks for fixing this so promptly