The instrospection use flag in dependencies

[aCOSwt]

This discussion started on irc but I find that asking for help on irc channels is not very handy when one is a complete n00b in the matter.
So... if someone here can help me understanding.

1/ I get a package which actually installs code depending on gobject-introspection.
2/ This package also actually depends on other packages (gtk+, pango, atk, librsvg, webkit-gtk...) which also honour the introspection use flag.
3/ Of course the introspection use flag defaults to set for all these dependencies but devs rightly say that one should not assume that users use default settings.

My problem is : How can I know precisely (without reading thousands of lines of code) which of the dependencies actually need to be emerged USE=+introspection ?

Of course I have been suggested to test the package in a chroot environment with an empty world and consider what happens if I get a gtk use=-introspection for example...

But I find this just too crazy, well, I mean a suggestion equivalent to "hunting the haggies". I just cannot imagine testing that way the package waiting for something wrong to occur {for each dep in the deplist emerged -introspection}

There must be something somewhere I must be missing.
Which one ?
_________________

[krinn]

Why not just

[aCOSwt]

Thanks krinn.

Well... I should have started saying that I am trying to write the ebuild for the first package. So... portage wont refuse building it if I make it rdepend gtk+
My question in fact is : How can I know that I should write the rdepend gtk+[introspection] ?
gtk+ being an example among half a dozen...
_________________

[krinn]

well, i would say you are suppose to have access to the program API you will depend on, and that doc should tell you if a function depend on another one.
so if you need funcTT from gtk+, you should read somewhere in gtk api doc that funcTT is only enable if introspection is enable. In this case, you would then know you should depend on gtk+[introspection] to make sure gtk+ have the function enable.

If you're doing it blindly, no doc... I understand more now why someone told you to chroot in a clean env to test it.
You can just make your ebuild gtk+ and one day or another you will certainly get a bug report telling you it fail with unknown function/symbol functTT.

But you should just read the doc to find out what you need, and you don't have to care about reading all dependencies doc to find out the one you need. If X need Z need Y that need H that need K that need B to finally have gtk+ providing A. You should write your dependency on gtk+ need A, and while portage try to provide A it will complain for any missing Z, Y, H, K, B in order to have A enable. It's A ebuild task to care about its dependencies not your.

It's easier if you don't depend on a function/symbols but on a program from another ebuild, as you could just "equery f thefile" to find who provide it, and read its ebuild to see conditions to provide the program.

here's how i would find distccmon-gnome presence

[aCOSwt]

[mv]

[Onip]

I think that you have three methods:

• the one they suggest you on irc ("clean room" try and error)
  
• reading through documentation (or sources if the first one fails)
  
• write a "naive" ebuild and then wait for bugreports to come... this seems not too serious, but if you will be able to react and correct quickly then it may be enough
  

out of curiosity, which package are we talking about?

p.s. for what I understood from my readings when a package uses introspection then it probably needs it for everyone of its deps (at least glib based ones)
_________________
Linux Registered User n. 373835

Titus Lucretius Carus, De Rerum Natura - Tantum religio potuit suadere malorum

[aCOSwt]

@mv : Thank you for your support.

[krinn]