Is RDEPENDS = "foo" supported?

Is RDEPENDS = "foo" supported?

[Stanislav Brabec]

Is RDEPENDS = "foo" allowed in recipes? If yes, then it does not work.

I have been searching for a reason, why my image does not set keymap.
Reason: console-tools were missing. Replacing RDEPENDS by RDEPENDS_${PN}
in the keymaps recipe fixed it.

Does it mean, that this and another 622 recipes use invalid syntax of
RDEPENDS or is it a bug in the toolchain?

Looking at the staging-pkg runtime data with just a plain "RDEPENDS ="
into work/, mentioned dependencies are missing.


________________________________________________________________________
Stanislav Brabec
http://www.penguin.cz/~utx/zaurus

Re: Is RDEPENDS = "foo" supported?

[Phil Blundell]

On Mon, 2009-06-22 at 21:28 +0200, Stanislav Brabec wrote:
> Is RDEPENDS = "foo" allowed in recipes? If yes, then it does not work.

It is allowed, and in at least some situations it will even work, but in
the vast majority of cases it is not the right thing to do.  There are
two problems with it: both of them have been discussed here on the list
quite recently, but to recap:

1) ${RDEPENDS} will apply to all the output subpackages, i.e. ${PN}-dev,
${PN}-doc and various ${PN}-locale-* things.  In most cases, ${PN} is
the only one that you actually want to receive the dependency.

2) ${RDEPENDS} will be overridden by ${RDEPENDS_${PN}} if the latter is
set.  This is sometimes done by .bbclass files and the like, in which
case any dependencies you set in ${RDEPENDS} will be ineffective.

> I have been searching for a reason, why my image does not set keymap.
> Reason: console-tools were missing. Replacing RDEPENDS by RDEPENDS_${PN}
> in the keymaps recipe fixed it.

In this particular case, update-rc.d.bbclass sets RDEPENDS_${PN} and
hence (2) above will apply.

> Does it mean, that this and another 622 recipes use invalid syntax of
> RDEPENDS or is it a bug in the toolchain?

It's probably true that almost all of those 623 recipes should be using
RDEPENDS_${PN} or something similar instead of plain RDEPENDS.

p.

Re: Is RDEPENDS = "foo" supported?

[Denys Dmytriyenko]

On Mon, Jun 22, 2009 at 08:37:40PM +0100, Phil Blundell wrote:
> On Mon, 2009-06-22 at 21:28 +0200, Stanislav Brabec wrote:
> > Is RDEPENDS = "foo" allowed in recipes? If yes, then it does not work.
> 
> It is allowed, and in at least some situations it will even work, but in
> the vast majority of cases it is not the right thing to do.  There are
> two problems with it: both of them have been discussed here on the list
> quite recently, but to recap:

Here is the original thread, for reference:
http://thread.gmane.org/gmane.comp.handhelds.openembedded/24663

> 1) ${RDEPENDS} will apply to all the output subpackages, i.e. ${PN}-dev,
> ${PN}-doc and various ${PN}-locale-* things.  In most cases, ${PN} is
> the only one that you actually want to receive the dependency.
> 
> 2) ${RDEPENDS} will be overridden by ${RDEPENDS_${PN}} if the latter is
> set.  This is sometimes done by .bbclass files and the like, in which
> case any dependencies you set in ${RDEPENDS} will be ineffective.
> 
> > I have been searching for a reason, why my image does not set keymap.
> > Reason: console-tools were missing. Replacing RDEPENDS by RDEPENDS_${PN}
> > in the keymaps recipe fixed it.
> 
> In this particular case, update-rc.d.bbclass sets RDEPENDS_${PN} and
> hence (2) above will apply.
> 
> > Does it mean, that this and another 622 recipes use invalid syntax of
> > RDEPENDS or is it a bug in the toolchain?
> 
> It's probably true that almost all of those 623 recipes should be using
> RDEPENDS_${PN} or something similar instead of plain RDEPENDS.

Yeah, I had to fix some of my recipes in the overlay here...

-- 
Denys