[OE-core] some variables not expanded corectly before BUILDCFG_VARS are shown

[Martin Jansa]

On Sat, Dec 01, 2012 at 02:34:37PM +0000, Richard Purdie wrote:
> On Fri, 2012-11-30 at 17:40 +0100, Martin Jansa wrote:
> > Hi,
> > 
> > when testing tune files I've noticed that TUNE_FEATURES are always shown
> > the same, even when build uses different value set from local.conf:
> > 
> > I've added DEFAULTTUNE to BUILDCFG_VARS to be sure:
> > BUILDCFG_VARS = "BB_VERSION BUILD_SYS NATIVELSBSTRING TARGET_SYS MACHINE
> > DISTRO DISTRO_VERSION TUNE_FEATURES DEFAULTTUNE TARGET_FPU"
> > 
> > I was changing DEFAULTTUNE with machine override in local.conf or .inc
> > file required from local.conf.
> > 
> > $ grep DEFAULTTUNE conf/local.conf 
> > DEFAULTTUNE_om-gta02 = "arm920t"
> > DEFAULTTUNE_nokia900 = "cortexa8t-neon"
> > DEFAULTTUNE_tuna = "cortexa9t-neon"
> > 
> > But the output at the beginning of build log was still the same. 
> > 1st is without any change to DEFAULTTUNE (so correctly armv4t and armv7a-neon)
> > 2nd is with DEFAULTTUNE set in conf/local.conf with machine override
> > 3rd is with default DEFAULTTUNE changed in tune-* file
> > 
> > 2nd and 3rd correctly uses cortexa9t-neon/cortexa8t-neon/arm920t tune, but only 3rd shows it:
> > nokia900/20121130041857.log:DEFAULTTUNE       = "armv7a-neon"
> > nokia900/20121130041857.log:TUNE_FEATURES     = "armv7a vfp neon"
> > nokia900/20121130101904.log:DEFAULTTUNE       = "armv7a-neon"
> > nokia900/20121130101904.log:TUNE_FEATURES     = "armv7a vfp neon"
> > nokia900/20121130135815.log:DEFAULTTUNE       = "cortexa8t-neon"
> > nokia900/20121130135815.log:TUNE_FEATURES     = "armv7a vfp thumb neon cortexa8"
> > om-gta02/20121130042112.log:DEFAULTTUNE       = "armv4t"
> > om-gta02/20121130042112.log:TUNE_FEATURES     = "armv4 thumb"
> > om-gta02/20121130102310.log:DEFAULTTUNE       = "armv4t"
> > om-gta02/20121130102310.log:TUNE_FEATURES     = "armv4 thumb"
> > om-gta02/20121130140126.log:DEFAULTTUNE       = "arm920t"
> > om-gta02/20121130140126.log:TUNE_FEATURES     = "armv4 thumb arm920t"
> > tuna/20121130041547.log:DEFAULTTUNE       = "armv7a-neon"
> > tuna/20121130041547.log:TUNE_FEATURES     = "armv7a vfp neon"
> > tuna/20121130101500.log:DEFAULTTUNE       = "armv7a-neon"
> > tuna/20121130101500.log:TUNE_FEATURES     = "armv7a vfp neon"
> > tuna/20121130135506.log:DEFAULTTUNE       = "cortexa8t-neon"
> > tuna/20121130135506.log:TUNE_FEATURES     = "armv7a vfp thumb neon cortexa8"
> > 
> > What's even more strange is that "thumb" override get applied in different order in 3rd build:
> > $ bitbake-diffsigs  tmp-eglibc/sstate-diff/13542[78]*/nokia900/cor*/libmad*/*do_con*
> > basehash changed from 8ad5d9c1803fed09a89e626c07581cb1 to 7e316bb3c389fe3cda6652ac1476a883
> > Variable EXTRA_OECONF value changed from 
> > -enable-speed --enable-shared --enable-fpm=arm --disable-aso --enable-fpm=default 
> > to 
> > -enable-speed --enable-shared --disable-aso --enable-fpm=default --enable-fpm=arm
> > # because libmad has:
> > # The ASO's don't take any account of thumb...
> > EXTRA_OECONF_append_thumb = " --disable-aso --enable-fpm=default"
> > EXTRA_OECONF_append_arm = " --enable-fpm=arm"
> > But both 2srd and 3rd got --mtune=cortexa8.
> > 
> > buildhistory.bbclass seems to read the same TUNE_PKGARCH (armv4t), 
> > all other builds are in arm920tt-oe-linux-gnueabi:
> > $ find ~/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi
> > /OE/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi
> > /OE/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/bblayers
> > /OE/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/bblayers/1.0-r0
> > /OE/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/bblayers/1.0-r0/bblayers-1.0
> > /OE/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/bblayers/1.0-r0/temp
> > /OE/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/bblayers/1.0-r0/temp/run.buildhistory_commit.4401
> > /OE/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/bblayers/1.0-r0/temp/run.buildhistory_commit.6566
> > /OE/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/bblayers/1.0-r0/temp/run.buildhistory_commit.11843
> > 
> > Any idea what is causing that?
> 
> Was this with or without the recent bitbake patch "data_smart: Improve
> get_hash to account for overrides and key expansion" ? if without,
> please see if that helps...

It was without this patch, only current master, but with this patch it
looks the same in log (I haven't checked if thumb/arm overrides are
still applied in different order, but output of 
bitbake -e | grep OVERRIDES is the same for all 3 cases also with or 
without this patch).

Cheers,

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa at gmail.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20121201/1995574c/attachment-0002.sig>