[OE-core] SRC_URI computing order

[Richard Purdie]

On Sun, 2013-11-03 at 23:16 +0100, Andrea Adami wrote:
> On Sat, Nov 2, 2013 at 9:47 AM, Eric Bénard <eric at eukrea.com> wrote:
> > Hi Richard,
> >
> > Le Wed, 30 Oct 2013 15:15:12 +0000,
> > Richard Purdie <richard.purdie at linuxfoundation.org> a écrit :
> >
> >> On Tue, 2013-10-29 at 08:28 +0100, Eric Bénard wrote:
> >> > Hi Khem,
> >> >
> >> > Le Mon, 28 Oct 2013 20:45:21 -0700,
> >> > Khem Raj <raj.khem at gmail.com> a écrit :
> >> >
> >> > > On Mon, Oct 28, 2013 at 7:10 AM, Eric Bénard <eric at eukrea.com> wrote:
> >> > > > Hi Richard,
> >> > > >
> >> > > > I saw your patch fixing FILESPATH's and Kergoth's one fixing
> >> > > > PACKAGECONFIG processing order and I think I'm also facing an order
> >> > > > problem when SRC_URI is computed.
> >> > > >
> >> > > > So when building SRC_URI when two layers have bbappend which apply
> >> > > > patches : the SRC_URI seems to be built using an order I fail to
> >> > > > understand somewhere instead of priority or the overrides' order.
> >> > > >
> >> > > > The use case is a System on Module and its custom motherboard :
> >> > > > - meta-fsl-arm :
> >> > > > * linux-imx_xyz.bb :
> >> > > > SRC_URI = "patchgeneric1 ..."
> >> > > >
> >> > > > - meta-som-support :
> >> > > > * conf/machine/mysom.conf
> >> > > >
> >> > > > * linux-imx_xyz.bbappend :
> >> > > > SRC_URI_append_mysom = "patchsom1 patchsom2 ..."
> >> > > >
> >> > > > - meta-custommotherboard (SOM + Cunstom Motherboard) :
> >> > > > * conf/machine/myproduct.conf
> >> > > > MACHINEOVERRIDES_prepend = "mysom:"
> >> > > > include conf/machine/mysom.conf
> >> > > >
> >> > > > * linux-imx_xyz.bbappend :
> >> > > > SRC_URI_append_myproduct = "patchproduct1 patchproduct2 ..."
> >> > > >
> >> > > > in the end I get :
> >> > > > SRC_URI = "patchgeneric1 ... patchsoc1 ... patchproduct1 ...
> >> > > > patchsom1 ..."
> >> > > >
> >> > > > and of course as patchproduct* are supposed to apply on top of
> >> > > > patchsoc* the patch fail to apply.
> >> > > >
> >> > > > I didn't found a way to build SRC_URI in the order I would like (I
> >> > > > tested : changing MACHINEOVERRIDES 's order, changing layers' priority,
> >> > > > changing machine's name to see if that was an alphabetical order ...).
> >> > > >
> >> > > > In the end the only thing which worked was to add an (empty by default)
> >> > > > variable in som's SRC_URI and filling this variables from the
> >> > > > custommotherboard's bbappend.
> >> > > >
> >> > > > Is the behaviour I'm seeing expected or is there something wrong in my
> >> > > > setup ?
> >> > >
> >> > > what is your OVERRIDES order.
> >> > >
> >> > "${TARGET_OS}:${TRANSLATED_TARGET_ARCH}:build-${BUILD_OS}:pn-${PN}:${MACHINEOVERRIDES}:${DISTROOVERRIDES}:${CLASSOVERRIDE}:forcevariable"
> >> >
> >> > so it follows the MACHINEOVERRIDES order (and I tried both append and
> >> > prepend to hack MACHINEOVERRIDES without any behaviour change).
> >>
> >> I think what Khem is asking is what OVERRIDES expands to?
> >>
> >> You mean patchso* and not patchsoc* above, right? Or should patchsom1 be
> >> patchsoc2?
> >>
> > oops :
> > I expect  SRC_URI = "patchgeneric1 ... patchsom1 ... patchproduct1 ..."
> > and I get :
> > SRC_URI = "patchgeneric1 ... patchproduct1 ... patchsom1 ..."
> >
> >> Its hard to follow and it might be easier if you could share a
> >> simplified test case we could reproduce this with. I don't doubt there
> >> is an issue in there but we need a way to reproduce and debug this.
> >>
> > OK, I'm preparing a simple testcase to reproduce that with oe-core +
> > meta-fsl-arm + meta-som + meta-baseboard.
> >
> > Eric
> > _______________________________________________
> > Openembedded-core mailing list
> > Openembedded-core at lists.openembedded.org
> > http://lists.openembedded.org/mailman/listinfo/openembedded-core
> 
> 
> I have to report an undesiderate behavior:
> 
> the formfactor files in our .bbappend are not considered :/
> DEBUG: Searching for machconfig in paths:....
>   /oe/oe-core/meta/recipes-bsp/formfactor/formfactor-0.0/
>   /oe/oe-core/meta/recipes-bsp/formfactor/formfactor/
>   /oe/oe-core/meta/recipes-bsp/formfactor/files/
>   /oe/meta-handheld/recipes-bsp/formfactor/files/poodle
> 
> so we end up with the empty machconfig of
> /oe/oe-core/meta/recipes-bsp/formfactor/files/
> 
> Surely this didn't happen when we tested the recipe.

With which revision of OE-Core? Was this with the dora release tag,
current dora head or master?

Cheers,

Richard