Fix deduplication of -enable-experimental-feature

[Dreksh]

When using a Swift package that contains multiple experimental features (i.e. https://github.com/sersoft-gmbh/swift-smtp ), the error that is returned shows treats one of the features as an input file.

This is due to the generated parameters within the *.swiftmodule-0.params file that is used to set the build arguments, where -enable-experimental-feature was deduplicated. The deduplication happens during bzl_selects.to_starlark as it operates on sets, instead of lists.

Changes:

• Introduced bzl_selects.unwrap_list to operate on lists instead of sets
• Changed the copts to rely on unwrap_list, as it shouldn't be treated as a set.

[cgrindel]

Thank you for the contribution! Please see my comment asking whether we can handle this inside to_starlark(). Also, could we add https://github.com/sersoft-gmbh/swift-smtp to one of the examples so that we reproduce the issue and confirm that it stays fixed?

[cgrindel]

Is there any way to detect this in bzl_selects.to_starlark()? It is not obvious to me why copts would get this special handling.

[Dreksh]

The reason is that we add a string repeatedly:

rules_swift_package_manager/swiftpkg/internal/swiftpkg_build_files.bzl

Line 103 in d671d9e

copts.append("-enable-experimental-feature")

to_starlark relies on using sets within its processing, which will deduplicate the string that's added.

I'm not familiar with the repository, so I decided to add a new function to reduce the chances that I'll affect other places that use to_starlark.

Possible Alternatives:

• Add a parameter like dedup = True, and only perform deduplication after the list has been created, so that if dedup = False was passed, it will return the list without deduping.
• Move adding -enable-experimental-feature after running to_starlark, so that it skips the deduplication. This allows us to reuse to_starlark as it is now. Experimental features will need to be placed in a different list, so it doesn't affect the other copts from unsafe flags.

[Dreksh]

I tried doing -enable-experimental-feature after running to_starlark, but didn't realise that it returns a structure instead of a list. So this method won't work, as to_starlark will have copts in an internal structure. (I did not dig into whether it'll be easy to append values into that list).

I'll give adding the dedup parameter a try (maybe renamed to preserve_list?)

[cgrindel]

@Dreksh Were you going to pick this up again? If not, I am inclined to close this. We can always open a new PR when/if you want to continue.

[cgrindel]

Alternatively, we could convert it to a Draft PR.