On Fri, May 24, 2013 at 02:06:01PM -0500, Scott Wood wrote:
On 05/17/2013 07:34:25 AM, Tom Rini wrote:
On Fri, May 17, 2013 at 05:12:19PM +0800, ying.zhang@freescale.com wrote:
From: Ying Zhang b40530@freescale.com
There was some functionality will be used in the SPL. They had been excluded by ifndef CONFIG_SPL_BUILD. Now, put it into the SPL.
Signed-off-by: Ying Zhang b40530@freescale.com
Compared with the previous version, give up new symbol and
delete the line
ifndef CONFIG_SPL_BUILD in common/env_common.c
What the heck is going on? First, you seem to be changing a number of checks from !CONFIG_SPL_BUILD to !CONFIG_SPL_NAND_MINIMAL, and then stop defining CONFIG_SPL_NAND_MINIMAL always and only define it for CONFIG_SPL_BUILD. Next, powerpc uses -ffunction-sections/-fdata-sections/--gc-sections so outside of assembler files, we shouldn't need to be using CONFIG_SPL_BUILD to not build something that's a static function.
gc-sections does not work on anonymous strings. It also doesn't work to avoid code that doesn't compile due to some missing dependency (e.g. missing #define). I don't know off the top of my head if this code falls into one of these categories (I don't even know what code we're talking about from the context here, without digging back for the original patch).
I'm not sure why being a static function matters.
Yes, there's a gcc bug (afaict) about strings, and #define-related fail to builds don't work, just unreferenced functions. You do sometimes need/want to whack #ifndef's around static functions for when you have to not define function A because of one of the above fails to be collected, and then you end up with an unused function warning.