You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Right now, lib_paths and libs are quite decoupled. It should be more deterministic to express which library belongs to which path. This would also be aligned with @vector-of-boollibman proposal at https://github.com/vector-of-bool/libman
The text was updated successfully, but these errors were encountered:
This kind of boils down to an impedance mismatch between the way Conan (and many tools) represent libraries versus how libman wants to consume them. A library has:
A name
One or more #include directories
One or more "linkables" (library files, as specified by this issue)
Public usage dependencies
Private link-time dependencies
When a package provides a single "library," Conan's representation is fairly sufficient with a few tweaks, given that the CppInfo object has includes, libs, and lib_paths. For packages which expose multiple linkables and/or multiple #include trees, the communication begins to break down.
In my repository, I'm working on a LibMan generator which looks into the UserInfo for a dependency in order to generate the metadata needed for the build system (otherwise tried to auto-generate it from CppInfo). It may be useful to experiment on a UserInfo solution might look like before moving it to the CppInfo object?
Right now, lib_paths and libs are quite decoupled. It should be more deterministic to express which library belongs to which path. This would also be aligned with @vector-of-bool
libman
proposal at https://github.com/vector-of-bool/libmanThe text was updated successfully, but these errors were encountered: