I found what the problem was. Some machine config file defined TARGET_ARCH =
"arm" and others did not. By default bitbake conf file sets TARGET_ARCH =
${TUNE_ARCH}. Although when evaluated TUNE_ARCH is set to "arm".
Using that diffsig tool bitbake gave different checksums because a change in
task dependencies since some machines TARGET_ARCH depended on TUNE_ARCH while
others didn't.
Thanks for your help Martin.
> -----Original Message-----
> From: Martin Jansa [mailto:martin.ja...@gmail.com]
> Sent: Wednesday, February 06, 2013 2:57 AM
> To: Cooper Jr., Franklin
> Cc: ross.bur...@intel.com; Openembedded-core@lists.openembedded.org
> Subject: Re: [OE-core] Multiple MACHINE building is broken in OE-core?
>
> On Wed, Feb 06, 2013 at 08:02:31AM +0000, Cooper Jr., Franklin wrote:
> > Hi Everyone,
> >
> > While performing denzil to danny work I ran into this same issue that Mike
> mentioned. I have 4 machines that are armv7a based and previously in denzil
> only very specific recipes that I deliberately made machine specific were
> being
> rebuilt. Now a significant amount of recipes are being unnecessarily rebuilt.
> I
> first noticed this when building Qt but this also occurs with something even
> simpler such as tslib. If I look at tslib's temp directory I end up seeing
> multiple
> logs for the majority of the tasks. For example I will see 4 log.do_compile
> logs.
> >
> > I ran the sstate-diff-machines.sh script that Martin mentioned for machines
> defined in meta-ti and I got the below output when using tslib as the
> -targets:
> >
> > list.M am3517-evm
> >
> > http://pastebin.com/zMH8PkmP
> >
> > list.M am335x-evm
> > http://pastebin.com/6jbP6guA
> >
> > The checksums differs for non-native recipes such as gettext, tslib and
> > zlib.
> >
> > If it matters I am using the 1.17.0 branch of bitbake since the master
> > branch
> throws parsing errors.
> >
> > Any suggestions on what could be the issue?
>
> Please pick something like tslib or gettext. Find first different checksum
> (do_configure is usually good for that).
>
> # Then to compare sigdata use something like:
> # $ ls sstate-diff/134934*/*/*/tslib/*do_configure*sigdata*
> # it should return 2 different sigdata with 2 different sstate # checksums,
> otherwise change ls expression
>
> # $ bitbake-diffsigs sstate-diff/134934*/*/*/tslib/*do_configure*sigdata*
>
> It will probably show some other recipe causing the difference, so just repeat
> ls, bitbake-diffsigs until you find the root cause.
>
> Cheers,
>
> --
> Martin 'JaMa' Jansa jabber: martin.ja...@gmail.com
_______________________________________________
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
