Make project compatible with FetchContent

[LecrisUT]

There are a few issues that should be resolved to make importing find_package and FetchContent equivalent:

• alias library: fdict -> fdict::fdict
• namespace targets, options, etc. In order to avoid nameclashes
• Add options: FDICT_INSTALL, FDICT_SHARED_LIBS to control how fdict is being built
• Set and propagate the variables in fdictConfig.cmake using return(PROPAGATE) or set(PARENT_SCOPE)

[zerothi]

Perhaps you can clarify some details I have been having about cmake, in the Config.cmake files generated, should one always specify set({pkg}_FOUND TRUE) or not? For instance you might end up in situations where a configuration file creates successfully finds the target it is supposed too, but it doesn't return _FOUND=true, the cmake documentation regarding this is very poor, and I cannot seem to figure out conventions in other packages?

Lastly, I have been focusing on getting the cmake stuff in. And was planning on getting the sub-project part done soonish, so I'll get my hands dirty again and will likely ask you some questions (here) if you don't mind? :)

[LecrisUT]

should one always specify set({pkg}_FOUND TRUE) or not?

In principle one does not have to do it. Normally you set <Package>_FOUND=FALSE when you specify that the import fails. I need to do more testing to confirm why <Package>_FOUND is not set to TRUE.

And was planning on getting the sub-project part done soonish, so I'll get my hands dirty again and will likely ask you some questions (here) if you don't mind? :)

Sure thing, feel free to make a PR and I can review and give some tips there.

[zerothi]

should one always specify set({pkg}_FOUND TRUE) or not?

In principle one does not have to do it. Normally you set <Package>_FOUND=FALSE when you specify that the import fails. I need to do more testing to confirm why <Package>_FOUND is not set to TRUE.

That information would be great, I have just encountered problems where the targets were found, but the _FOUND was not set (hence false), and then a Warning is issued for the users. This is then wrong... My solution was to never trust _FOUND but always check for the target...

[LecrisUT]

That information would be great, I have just encountered problems where the targets were found, but the _FOUND was not set (hence false), and then a Warning is issued for the users.

Note that FetchContent does not do an install() + find_package(), and in that case you need to synchronize the variables set in find_package() (hence the issue raised here)

[zerothi]

Might I ask what you are planning to use fdict for?

[LecrisUT]

Might I ask what you are planning to use fdict for?

I don't know if it's a well-known design pattern, but I'm using this to add a context object/field so it is easier to write extensions in octopus. I.e. add an arbitrarily typed dictionary which is exactly what this package is doing

[zerothi]

I believe this is now fixed. Please re-open if necessary!

[zerothi]

Well, actually a small question @LecrisUT.

I removed FDICT_INSTALL because I ended up in problems where the targets where not exported correctly if I couldn't install them... So what do you do when install is hidden behind FDICT_INSTALL when you want it to be used as a subproject? I couldn't get it to work!

[LecrisUT]

I removed FDICT_INSTALL because I ended up in problems where the targets where not exported correctly if I couldn't install them

Not sure I follow that. With export you mean the FetchContent consuming project would not be able to install correctly? Do you have a version to look at?

[zerothi]

I mean something like this:

if(FDICT_INSTALL)
  install(... )
  export(... NAMESPACE ...)
endif()

In these cases the target will not be exported, which is kind of what I need? No?

If I add it as a subdirectory, it complained about the target not defined. And I would rather use the export functionality to expose the namespace target, rather than manually defining it (seems redundant!)?

[LecrisUT]

In these cases the target will not be exported, which is kind of what I need? No?

Not really, if it's only used as a static library for a binary having just the binary installed should be fine. I remember there being some weirdness when a shared library target links to a static and it's then exported. I don't remember what was going on there though.

export() only exports targets for build-dir, and install(EXPORT) only at install step.

If I add it as a subdirectory, it complained about the target not defined. And I would rather use the export functionality to expose the namespace target, rather than manually defining it (seems redundant!)?

For add_subdirectory you should use add_library(ALIAS). In order to use export you need to find_package and that's quite a complicated procedure.

Is it complaining when trying to install a fdict_user project?