Bug 13188

I tried to modify the code but I still get error:

Log below:

| Running custom install script 'XXXXXXXXXXX/tmp/hosttools/sh -c touch $DESTDIR/usr'
| ln: failed to create symbolic link 'XXXXXXXXXX/tmp/work/cortexa17hf-neon-vfpv4-rk-linux-gnueabi/systemd/1_239-r0/image/sbin/init': File exists

No idea why the init file exists.

I can confirm it is a bug.
Althought I have not traced to the cause.
I had to add a line to remove the sbin/init file.

mv ${D}/${base_sbindir}/init

Then I had to mount the ext4 rootfs and manully create the link.
The current symbolic link is linked to a wrong location.

Can you explain in detail how to reproduce the error?
Lots of people are building systemd but perhaps this is a silent failure that we've been ignoring or it could be that you have a custom install script and have made an error. We need more info to proceed.

I have a local recipe that installs a systemd service to start my system at startup.
The bb file is something like below.
Hope it helps.

Thanks
Qingjun

SUMMARY = "XXX"
LICENSE = "CLOSED"
PV = "1.0"

SRC_URI = "... \
"

S = "${WORKDIR}"

inherit systemd
inherit bin_package
INSANE_SKIP_${PN} = "ldflags"
INHIBIT_PACKAGE_STRIP = "1"
INHIBIT_SYSROOT_STRIP = "1"
INHIBIT_PACKAGE_DEBUG_SPLIT = "1"

do_compile() {
}

do_install () {
	install -Dm 0644 ${S}/XXX.service ${D}/${systemd_unitdir}/system/XXX.service
}


SYSTEMD_SERVICE_${PN} = "XXX.service"
INSANE_SKIP_${PN} += "installed-vs-shipped"

Chen,

It appears an answer to the question has been given.  I have cleared NEEDINFO.  

Thanks,

Stephen

(In reply to comment #0)
> line 231:
> [ ! -e ${D}/init ] && ln -s ${rootlibexecdir}/systemd/systemd ${D}/init

This is not line 231 in yocto 2.6.1 systemd_239.bb.

Do you have local modifications or are you actually running some other yocto version?

wei,

Sorry for late reply. I guess the cause is that your ubuntu18.04 does not have /lib/systemd/systemd, causing ${D}/init being a dead link.
I think you are right, using '-L' could solve this problem. As an alternative, you can use 'ln -sf' instead 'ln -s'.
You can send out patch if you like, as you have already found the solution.
You can send patch to openembedded-core@lists.openembedded.org, for both thud and master branches.
If not convenient, I'll do it for you.

Best Regards,
Chen Qi

wei,

Do you still have this bug reproduced on your host?

I cannot reproduce this problem, nor can I find out the host environment of the reporter.

Qingjun(In reply to comment #9)
> I cannot reproduce this problem, nor can I find out the host environment of
> the reporter.

Sorry didn't have a chance to try new release of Yocto.
Will try to test against latest release in this week.

Regards

(In reply to comment #9)
> I cannot reproduce this problem, nor can I find out the host environment of
> the reporter.

Hello Chen,
I just tried systemd 2.41 on latest warrior-next branch.
The problem still exists

Below is the log. BTW I am running on a Ubuntu 18.04 X64.

ln: failed to create symbolic link '/home/weiqj/rk-yocto-bsp/libre-hub/tmp/work/cortexa17hf-neon-vfpv4-rk-linux-gnueabi/systemd/1_241+AUTOINC+c1f8ff8d0d-r0/image/sbin/init': $
WARNING: exit code 1 from a shell command.
ERROR: Function failed: do_install (log file is located at /home/weiqj/rk-yocto-bsp/libre-hub/tmp/work/cortexa17hf-neon-vfpv4-rk-linux-gnueabi/systemd/1_241+AUTOINC+c1f8ff8d0$

Still, I had to manually modify the script to make it work.

Hi Wei,

I'm wondering why other people don't have such problem. I also use ubuntu hosts.
Do you have some additional layer which might have systemd bbappend files?

What does your run.do_install file under ${WORKDIR}/temp look like?
What creates the ${D}/init link before this line? [ ! -e ${D}/init ] && xxx

Could you please show your run.do_install file?

Also, you mentioned ${D}/init, but your error log is about ${D}/sbin/init.

Anyway, I think showing your run.do_install will help.

Wei?

No reponse so I'm closing the defect. We'd be happy to have the defect re-opened if the needed info is provided.