[OE-core] [PATCH] sstate: Improve handling of machine specific manifests

[Richard Purdie]

On Mon, 2012-10-22 at 13:08 +0200, Martin Jansa wrote:
> On Mon, Oct 22, 2012 at 11:58:28AM +0100, Richard Purdie wrote:
> > On Mon, 2012-10-22 at 12:44 +0200, Martin Jansa wrote:
> > > On Fri, Oct 19, 2012 at 03:48:55PM +0100, Richard Purdie wrote:
> > > > Now do_package isn't machine specific, we're only left with do_populate_sysroot as a
> > > > machine specific task. This change marks only the machine specific manifests as machine
> > > > specific, defaulting to PACKAGE_ARCH for everything else.
> > > > 
> > > > This means we do less work where there are multiple machines using the same
> > > > core package architecture and we can start to clean up the sstate duplicate files
> > > > whitelist.
> > > > 
> > > > Signed-off-by: Richard Purdie <richard.purdie at linuxfoundation.org>
> > > > ---
> > > > diff --git a/meta/classes/sstate.bbclass b/meta/classes/sstate.bbclass
> > > > index d2a120b..dee84bf 100644
> > > > --- a/meta/classes/sstate.bbclass
> > > > +++ b/meta/classes/sstate.bbclass
> > > > @@ -17,10 +17,7 @@ SSTATE_EXTRAPATH   = ""
> > > >  SSTATE_EXTRAPATHWILDCARD = ""
> > > >  SSTATE_PATHSPEC   = "${SSTATE_DIR}/${SSTATE_EXTRAPATHWILDCARD}*/${SSTATE_PKGSPEC}"
> > > >  
> > > > -# In theory we should be using:
> > > > -# SSTATE_DUPWHITELIST = "${DEPLOY_DIR_IMAGE}/ ${DEPLOY_DIR}/licenses/ ${DEPLOY_DIR_IPK}/all/ ${DEPLOY_DIR_RPM}/all ${DEPLOY_DIR_DEB}/all/ ${TMPDIR}/pkgdata/all${TARGET_VENDOR}-${TARGET_OS}"
> > > > -# However until do_package is not machine specific, we'll have to make do with all of deploy/pkgdata.
> > > > -SSTATE_DUPWHITELIST = "${DEPLOY_DIR}/ ${TMPDIR}/pkgdata/"
> > > > +SSTATE_DUPWHITELIST = "${DEPLOY_DIR_IMAGE}/ ${DEPLOY_DIR}/licenses/"
> > > 
> > > Looks like warnings are back :/
> > > 
> > > WARNING: The recipe attr is trying to install files into a shared area when those files already exist. Those files are:
> > >    /OE/jansa-test/shr-core/tmp-eglibc/deploy/ipk/armv7a-vfp-neon/attr-locale-de_2.4.46-r4_armv7a-vfp-neon.ipk
> > >    /OE/jansa-test/shr-core/tmp-eglibc/deploy/ipk/armv7a-vfp-neon/attr-dbg_2.4.46-r4_armv7a-vfp-neon.ipk
> > >    /OE/jansa-test/shr-core/tmp-eglibc/deploy/ipk/armv7a-vfp-neon/attr-locale-sv_2.4.46-r4_armv7a-vfp-neon.ipk
> > > ...
> > > 
> > > and new warnings from pkgdata
> > > WARNING: The recipe bison is trying to install files into a shared area when those files already exist. Those files are:
> > >    /OE/jansa-test/shr-core/tmp-eglibc/pkgdata/armv7a-vfp-neon-oe-linux-gnueabi/bison
> > >    /OE/jansa-test/shr-core/tmp-eglibc/pkgdata/armv7a-vfp-neon-oe-linux-gnueabi/runtime/bison-locale-nl.packaged
> > >    /OE/jansa-test/shr-core/tmp-eglibc/pkgdata/armv7a-vfp-neon-oe-linux-gnueabi/runtime/bison-dbg.packaged
> > >    /OE/jansa-test/shr-core/tmp-eglibc/pkgdata/armv7a-vfp-neon-oe-linux-gnueabi/runtime/bison-doc
> > >    /OE/jansa-test/shr-core/tmp-eglibc/pkgdata/armv7a-vfp-neon-oe-linux-gnueabi/runtime/bison-locale-th.packaged
> > > ...
> > 
> > The question is why as they shouldn't be, these changes were meant to
> > fix this properly. Initially I wondered if this was another
> > manifestation of https://bugzilla.yoctoproject.org/show_bug.cgi?id=3219
> > but I'm not so sure.
> 
> Probably not as this happens on builder with 2 machines using the same
> tune and the same CCARGS.
> 
> > Can you figure out which two recipes are trying to install these sets of
> > files?
> 
> I'll try to compare them with scripts/sstate-diff.sh again to see if
> checksums are the same between those 2 machines, but those warnings are
> shown already when building 1 machine.
>  
> > Or perhaps this is a one off transition issue I didn't see here when
> > testing this? Does a build from a clean tmp do this?
> 
> Yes I've removed tmp-eglibc before starting this build (kept only
> sstate-cache) and it's building first machine.

If the warnings are showing up even after building one machine, there is
something very strange going on. Did it run the setscene do_package task
for these recipes?

Its as if it installed from sstate and then decided to overwrite the
files. Something isn't making sense though :/. Firstly, the sstate
should have been invalidated due to the package class and variable
changes. Secondly, if it had installed the files, it should be
uninstalled them before installing a second set. So I'm confused as to
what is going on here...

Cheers,

Richard