Clarify validity requirements in docs of PackageFilesInfo/PackageFilegroupInfo

[nnmm]

Is it invalid (meaning under no circumstances a good idea, assumed by consuming rules not to happen) to have

• A PackageFilegroupInfo that contains the same destination file twice? I assume yes, c.f. Make duplicate output paths an error instead of a warning #682
• A PackageFilesInfo that contains directory Files?
• A PackageFilesInfo that contains symlink Files?
• One or two PackageFilesInfo in the same PackageFilegroupInfo containing both a file a/b and a file a/b/c, where a/b is a regular or symlink File?
• One or two PackageFilesInfo in the same PackageFilegroupInfo containing both a file a/b and a file a/b/c, where a/b is a directory File?

Imo it would be helpful to mention that in the docs of the respective provider. That way, these providers can become a more standardized interface between producing and consuming rules.

[aiuto]

Do you intend to create new rules which would consume these providers?

I'm having trouble understanding the questions abouta/b and a/b/c mixing where a/b is a regular file. Do you have a working example which can do that? If the source package a/b exists, so there can be a target a/b/c, then it is almost always incorrect in Bazel to have the package a produce the plain or symlink target a/b.

[nnmm]

@aiuto Yes, I do have a custom rule consuming these providers.

By "containing a file a/b", I mean an entry in dest_src_map with key a/b. So this has nothing to do with targets' names/packages, only with the destination path.

[aiuto]

We're talking about the same thing. It is very difficult to be clear without a concrete example. The srcs of a pkg_ rule are targets themselves, even if they are plain files. While it is possible to create something like that, it is an unsupported edge case in Bazel. To put it a different way. If //a/b:c is a valid target in your project, and you have also defined the target //a:b, then you are doing something wrong at the core Bazel level. I'm not going to spend any thought time about the expected behavior when we hit those situations.

…

On Tue, Oct 17, 2023, 7:20 AM Nikolai Morin ***@***.***> wrote: @aiuto <https://github.com/aiuto> Yes, I do have a custom rule consuming these providers. By "containing a file a/b", I mean an entry in dest_src_map with *key* a/b. So this has nothing to do with targets' names/packages, only with the destination path. — Reply to this email directly, view it on GitHub <#746 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAXHHHAYCP55L6WFF23JDD3X7ZSZHAVCNFSM6AAAAAA4C4ZJEOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONRWGIYDSOBSGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>