[yocto-builds] buildbot failure in The Yocto Autobuilder on nightly-refkit

Bystricky, Juro juro.bystricky at intel.com
Wed Oct 25 13:47:33 PDT 2017


Yes, mea culpa. I modified util-linux-ptest so it covers a lot more cases now (all available).
I pulled in some more stuff via RDEPENDS, but only for util-linux-ptest. This was typically based on test
complaints that something is missing/not found ...

https://bugzilla.yoctoproject.org/show_bug.cgi?id=10953

Juro

________________________________________
From: Patrick Ohly [patrick.ohly at intel.com]
Sent: Wednesday, October 25, 2017 10:38 AM
To: Saul Wold; Ylinen, Mikko; Bystricky, Juro
Cc: yocto-builds at yoctoproject.org
Subject: Re: [yocto-builds] buildbot failure in The Yocto Autobuilder on nightly-refkit

On Wed, 2017-10-25 at 10:01 -0700, Saul Wold wrote:
> On Wed, 2017-10-25 at 07:18 +0300, Ylinen, Mikko wrote:
> >
> >
> > On Wed, Oct 25, 2017 at 12:14 AM, Saul Wold <sgw at linux.intel.com>
> > wrote:
> > > Patrick, Mikko:
> > >
> > > Is there an issue with btrfs?
> > >
> > > ERROR: The following unsupported recipes are required for the
> > > build:
> > >   btrfs-tools at core (would be supported in workspacelayer)
> > >   lzo at core (would be supported in workspacelayer)
> > >   which at core (would be supported in workspacelayer)
> > >
> > > Seems btrfs-tools triggered this issue.
> > >
> >
> > These are new dependencies to something and they'd need to be added
> > to the 'supported recipes' (or cleaned out by reworking the
> > dependencies).
> >
> > Perhaps autobuilder should make these warnings instead of errors.
> >
>
> I believe that this error comes from the supportrecipes.bbclass in
> the refkit code, not something the autobuilder can affect.

There's a configure option to turn the error into a warning:
SUPPORTED_RECIPES_CHECK = warn

This might be more appropriate for the AB.

> But a little more investigating found that Juro's patch to util-linux
> to fix up util-linux's ptest setup added btrfs-tools, which also
> caused lzo and which to be pulled in.
>
> The dependency line from the log:
>   fwupd -> acl -> attr -> autoconf-native -> ca-certificates-native
> ->
> cve-check-tool-native -> glib-2 -> *btrfs-tools -> coreutils -> dbus
> ->
> dbus-test -> e2fsprogs -> gnome-desktop-testing -> kbd -> libcap ->
> libcgroup -> libpam -> *lzo
>
> was a little confusing since it showed glib-2 initially, but when I
> ran
> it locally I got a more reasonable output:
>
> refkit-image-minimal -> acl -> attr -> *btrfs-tools -> coreutils ->
> dbus -> dbus-test -> e2fsprogs -> glib-2 -> gnome-desktop-testing ->
> kbd -> libcap -> libcgroup -> libpam -> shadow -> systemd -> systemd-
> compat-units -> shared-mime-info -> *which
>
> Still not correct as attr does not depend on btrfs-tools, it was
> really
> util-linux.

Internally there's no difference between direct and indirect
dependencies and therefore supported-recipes.bbclass cannot decide
whether a dependency should be shown (because it is actually listed in
the recipe) or skipped (because it gets inherited).

In practice, looking at the longest paths first usually works better.
It would be a worthwhile enhancement of the class to implement that
heuristic there.

--
Best Regards

Patrick Ohly
Senior Software Engineer

Intel GmbH
Open Source Technology Center
Usenerstr. 5a                           Phone: +49-228-2493652
53129 Bonn
Germany


More information about the yocto-builds mailing list