[Bug 4878] msm machine update/cleanup

[bugzilla-daemon at amethyst.openembedded.net]

http://bugs.openembedded.net/show_bug.cgi?id=4878





--- Comment #4 from Lukas Gorris <lukas.gorris at gmx.de>  2008-11-28 20:16:35 ---
(In reply to comment #3)
> (In reply to comment #2)
> > (In reply to comment #1)
> > > (In reply to comment #0)
> > > 
> > > > This patch does the followng: Add machines htcvogue htcraphael htcdiamond and
> > > > their kernels. 
> > > 
> > > Why so many kernel recipes? We have machine overrides for a reason. Secondly,
> > > defconfigs should be put in OE, not copied out of the tree.
> > > 
> > 
> > 
> > I need to pass the buck to XorA here about the defconfigs. he did the initial
> > kernel recipe ;P 
> 
> In OE the defconfigs should be in OE so we can fix them without needing to
> apply patches, which sucks

ok you have a point there. so if you need oabi or so for some distro you can
change that from OE. will put them in OE then. I'm not looking forward updating
them every day though.

> 
> > I don't see the pros of using hardcoded OE recipe. these machines all change
> > quickly, and so do their defconfigs. likewise proposed to keep the level of
> > required maintainance as low as possible.
> 
> Right, putting the defconfig in OE actually eases the maintenance. The point
> about change is bogus, since the SRCREV of the kernel is locked down.
> 
> > Can we please just have the 3 recipes for now? It is only temporarily until we
> > have all the machines merged to one branch anyway.
> 
> No, we can't keep 3 different recipes, you can put them in one recipe really
> easy. Like I said, machine overrides are there for a reason.
> 

good if it makes you happy I try putting one instead.

> > > > Fix htckaiser and polaris, add new tunefile for msm machines
> > > > because the available one for arm11 doesn't work.
> > > 
> > > What exactly doesn't work? It works fine on the armv6 and armv7a machines I
> > > have here. If you get something like 'illegal' instruction, your kernel
> > > defconfig is broken, see my point above about keeping defconfigs in OE.
> > 
> > Correct. I get illegal instruction error in the images when building with the
> > compiler switches "-mfpu=vfp -mfloat-abi=softfp" set by
> > "org.openembedded.dev/conf/machine/include/tune-arm1136jf-s.inc"
> > 
> > our kernel configs are just great. 
> 
> If they were, you wouldn't have the problems you are talking about

uhm could you please respond to what I detailed above? you seriously think the
msm kernel hackers use no VFP for nothing?
> 
> > compare them with the defconfigs for the
> > halibut evaluation board and g1 used by the google kernel developers:
> > http://android.git.kernel.org/?p=kernel/msm.git;a=blob;f=arch/arm/configs/msm_defconfig;hb=refs/heads/android-msm-2.6.25
> > 
> > line 287 "# CONFIG_VFP is not set"
> > we use the same.
> 
> I wasn't talking about that.

great we put the defconfigs in OE. what about the tunefile? what must I do to
convince you it is what OE needs?

-- 
Configure bugmail: http://bugs.openembedded.net/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.