Two questions on path.glob-tree behavior

[gabrieldevillers]

I have encountered the following issue: https://stackoverflow.com/questions/79284424/impossible-to-link-against-target-defined-in-another-jamfile-relative-sources-m

Environment and version details

• Operating System+version: Windows 11
• Compiler+version: MSVC 19.40.33813
• Shell: Git Bash
• B2 Version: B2 Version 4.10. OS=NT. ; B2 4.10-git
• B2 Configuration: will provide it if needed

Brief problem description

1. rule path.glob-tree seems to return paths that are relative to working directory, not the jamfile containing it. Is this intended ? If yes this should be documented I think.
2. the workaround I describe in the SO thread using path-constant causes path.glob-tree to return rooted paths, so I had to write the rule my-relative-path to get relative paths again. Is this normal (I guess it is because path-constant return an absolute path) ?
   2.1) Is there a built-in alternative to my-relative-path ?

[grisumbras]

1. glob-tree is a project rule, so it has access to the current project. path.glob-tree doesn't have access to the current project, because it's not a project rule, and in fact it doesn't even know if there is a current project. So, the behaviour is kind of obvious if you think about it, but I agree that it's better to document such things.
2. Project rule rule glob-tree-ex ( paths + : wildcards + : excludes * ) is what you are looking for. It's not documented, and it definitely should be.

[gabrieldevillers]

Thank you very much, glob-tree-ex seems indeed to be what I need.