[patch] fix omap compile with randconfig

This fixes:

arch/arm/mach-omap2/built-in.o: In function `omap_reserve':
/data/l/linux-n900/arch/arm/mach-omap2/common.c:36: undefined
  reference to `omap_fb_reserve_memblock'

Signed-off-by: Pavel Machek <redacted>

diff --git a/arch/arm/mach-omap2/common.c b/arch/arm/mach-omap2/common.c
index 9beecde..857907f 100644
--- a/arch/arm/mach-omap2/common.c
+++ b/arch/arm/mach-omap2/common.c

@@ -33,5 +33,7 @@ void __init omap_reserve(void)
 	omap_dsp_reserve_sdram_memblock();
 	omap_secure_ram_reserve_memblock();
 	omap_barrier_reserve_memblock();
+#ifdef CONFIG_OMAP2_VRFB
 	omap_fb_reserve_memblock();
+#endif
 }

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

On Sun, Jan 12, 2014 at 5:00 PM, Pavel Machek [off-list ref] wrote:

What is this patch against? omap_fb_reserve_memblock() doesn't exist
in v3.13-rc8, nor in -next.

Adding an ifdef here is likely the wrong thing anyway, there should be
a stub provided for !CONFIG_OMAP2_VRFB in the header file instead.


-Olof

On Mon 2014-01-13 10:02:58, Olof Johansson wrote:

On Sun, Jan 12, 2014 at 5:00 PM, Pavel Machek [off-list ref] wrote:

quoted

This fixes:

arch/arm/mach-omap2/built-in.o: In function `omap_reserve':
/data/l/linux-n900/arch/arm/mach-omap2/common.c:36: undefined
  reference to `omap_fb_reserve_memblock'

Signed-off-by: Pavel Machek <redacted>

diff --git a/arch/arm/mach-omap2/common.c b/arch/arm/mach-omap2/common.c
index 9beecde..857907f 100644
--- a/arch/arm/mach-omap2/common.c
+++ b/arch/arm/mach-omap2/common.c

@@ -33,5 +33,7 @@ void __init omap_reserve(void)
        omap_dsp_reserve_sdram_memblock();
        omap_secure_ram_reserve_memblock();
        omap_barrier_reserve_memblock();
+#ifdef CONFIG_OMAP2_VRFB
        omap_fb_reserve_memblock();
+#endif
 }

What is this patch against? omap_fb_reserve_memblock() doesn't exist
in v3.13-rc8, nor in -next.

It is actually against -n900 tree, sorry about that. But this means
that this regression likely will not hit -next :-).

Adding an ifdef here is likely the wrong thing anyway, there should be
a stub provided for !CONFIG_OMAP2_VRFB in the header file instead.

Agreed.

I found few more compile problems, but those are neither easy nor I
expect them to happen commonly. In particular BIG_ENDIAN=y breaks
flexcan, MMU=n break some core code, and CONFIG_ARM_KPROBES_TEST=y
just does not work with my compiler here. Probably not worth trying to
reproduce on vanilla tree?

Thanks,
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html