[OE-core] Why is "bitbake core-image-minimal" building gtk docs?

[Burton, Ross]

On 21 August 2013 09:40, Jack Mitchell <ml at communistcode.co.uk> wrote:

> Here is the patch which kicks it all off:
>
>
> http://lists.linuxtogo.org/pipermail/openembedded-core/2012-July/026077.html
>
> So, make of this what you will. I've cc'd Ross in to see if he has a
> specific reason for re-enabling gkt-doc in some packages. Maybe we
> should turn it into an image feature...?
>

tl;dr: it's a stub package, you're not building any gtk-doc, and you can't
turn it on if you wanted to.

There's some confusion in this thread, so I'll try and clear it up.

We need to build gtk-doc (or compatible) because it provides m4 autoconf
fragments that are needed at configure time by any package that uses
gtk-doc for documentation.  Previously we were patching recipes one by one
to disable the rules, but that's prone to failure.  We integrated
gtk-doc-stub (also used by gnome-os) which has the same external interfaces
but is entirely stubs, for example:

https://git.gnome.org/browse/gtk-doc-stub/tree/gtk-doc.m4

This way we can actually configure everything that uses gtk-doc without
needing the full gtk-doc package.

There is a gtk-doc class which adds a dependency on gtk-doc-stub-native and
then adds --disable-gtk-doc to EXTRA_OECONF.  This is why the explicit
--disable was removed from connman: it was redundant.

The other advantage of using a central gtk-doc class is that if someone did
want to build gtk-doc, they would just need to package up the real gtk-doc
and change the class.

Ross
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20130821/26383b80/attachment-0002.html>