[OE-core] First run postinst broken

Andreas Müller schnitzeltony at googlemail.com
Sun Sep 20 18:56:40 UTC 2015


On Sun, Sep 20, 2015 at 12:30 PM, Andreas Müller
<schnitzeltony at googlemail.com> wrote:
> Hi,
>
> on images build with current master the postinsts for first run on
> target are broken. Looking into
>
> 'meta/recipes-devtools/run-postinsts/run-postinsts/run-postinsts.service'
>
> I would expect
>
> '/etc/*-postinsts'
>
> in the image created for packets creating
>
>  '<package>.postinst returned 1, marking as unpacked only,
> configuration required on target.'
>
> in log.do_rootfs. I grepped around but could not find the place
> creating the files '/etc/*-postinsts'. Where are these files supposed
> to be created?
>
> Meanwhile I'll start a build with working configuration to find out
> where they are created...
>
This is interesting: On a system with working
target-first-run-postinst I have also no files/dirs matching
'/etc/*-postinsts'. After first start 'systemct status rub-postinsts'
reports

sh-4.3# systemctl status run-postinsts
��run-postinsts.service - Run pending postinsts
   Loaded: loaded (/lib/systemd/system/run-postinsts.service; enabled;
vendor preset: enabled)
   Active: inactive (dead)
Condition: start condition failed at Sat 2000-01-01 00:02:10 UTC; 15
years 8 months ago

Jan 01 00:02:10 varsomimx6q systemd[1]: Started Run pending postinsts.

Hoppla: systemd says that start condition is not met but start the
service at the same time?!?

The image with 'working' postinst has systemd 219 the 'nonworking'
225. Looking in systemd's commits/News did not give hints but I start
to think that run-postinsts.service was started accidently by a bug in
systemd which is fixed in systemd 225.

Still have some difficulties in understanding the role of
packet-manager and interception scripts. I don't think it is that easy
to remove

ConditionPathExistsGlob=#SYSCONFDIR#/*-postinsts

from 'meta/recipes-devtools/run-postinsts/run-postinsts/run-postinsts.service'

Packagemanager / interception scripts experts welcome..

Andreas



More information about the Openembedded-core mailing list