[oe] [meta-oe][v2] 00/22] systemd rework

[Andreas Müller]

On Fri, Feb 17, 2012 at 7:53 PM, Andreas Müller
<schnitzeltony at googlemail.com> wrote:
> On Thu, Feb 16, 2012 at 1:53 PM, Martin Jansa <martin.jansa at gmail.com> wrote:
>> On Thu, Feb 16, 2012 at 03:02:54AM +0100, Andreas Müller wrote:
>>> V1 -> V2:
>>> [2]  do not check configuration variables for cross and nativesdk
>>> [3]  rename systemd_package to systemd_generate_package_scripts
>>> [6]  parse unit files to determine the files to pack
>>> [7]  create warning if *-systemd package already rdepends
>>>      it's base package
>>> added in V2:
>>> [8]-[20] remove obsolet systemd code from recipes (these were used as
>>>      test cases)
>>> [21] dhcp-(relay) test case for multiple systemd packages
>>> [22] modified to start service
>>
>> Whole series looks good to me sofar.
>>
>> I'm not sure what's meant by this warning:
>> WARNING:
>> /OE/shr-core/meta-openembedded/meta-oe/recipes-connectivity/gateone/gateone_git.bb:
>> it is recommended to set SYSTEMD_PACKAGES as <existing-package>-systemd
>> WARNING:
>> /OE/shr-core/meta-openembedded/meta-oe/recipes-devtools/cloud9/cloud9_0.5.1.bb:
>> it is recommended to set SYSTEMD_PACKAGES as <existing-package>-systemd
>>
>> I guess we should
>> -SYSTEMD_PACKAGES = "${PN}"
>> +SYSTEMD_PACKAGES = "${PN}-systemd"
>> with this patchset too and add ${PN}-systemd to task which pulls gateone/cloud9.
>>
>> And I'm not sure if this change is intentional:
>> http://git.shr-project.org/git/?p=buildhistory.git;a=commitdiff;h=3d27d7267406bd0fb2ea2a48b108dbb8bc7c128f#patch18
>>
>> as xserver-nodm-init-systemd doesn't really need xserver-nodm-init, but
>> probably won't harm if xserver-nodm-init is used only to pull
>> xinit+xserver-common dependency.
>>
> The only 'clean' way out here I see is an additional variable
> overriding automatic rdepending. How about
>
> SYSTEMD_RDEPENDS: If this variable is set with a list of packages, its
> contents are used for redepending (systemd is still added). It should
> be possible to set packages outside of recipe and of course the
> contents should be checked during parse. What do you think?
>
How about this: If systemd.bbclass detects, that
RDEPENDS_${systemd-base-package} is not empty, it will only append
'systemd'. So the recipe can override the automatism easily and there
is no need for another variable.

Andreas