[U-Boot] nand spl build with wrong CONFIG_SYS_TEXT_BASE

[Wolfgang Denk]

Dear Scott Wood,

In message <20101109143747.685f973f at udp111988uds.am.freescale.net> you wrote:
>
> I don't see how
> 
> #ifdef NAND_SPL
> #define CONFIG_SYS_TEXT_BASE xxx
> #else
> #define CONFIG_SYS_TEXT_BASE yyy
> #endif
> 
> is more of a maintenance problem than
> 
> #define CONFIG_SYS_TEXT_BASE_SPL xxx
> #define CONFIG_SYS_TEXT_BASE yyy
> 
> followed by multiple instances of
> 
> #ifdef NAND_SPL
> do something with CONFIG_SYS_TEXT_BASE_SPL
> #else
> do something with CONFIG_SYS_TEXT_BASE
> #endif

Assuming that your code really needs to know the start address of the
image, it should probably do something like this instead:

#ifdef NAND_SPL
#define CONFIG_SYS_TEXT_BASE xxx
#define CONFIG_SYS_MONITOR_BASE CONFIG_SYS_TEXT_BASE
#else
#define CONFIG_SYS_TEXT_BASE yyy
#define CONFIG_SYS_MONITOR_BASE CONFIG_SYS_TEXT_BASE
#endif

followed by multiple instances of

do something with CONFIG_SYS_MONITOR_BASE

CONFIG_SYS_TEXT_BASE is a symbos primarily intended for the linker
call, i. e. for Makefile context.  This is a pretty special situation,
that does not apply to "normal" CONFIG_* settings.

> And as I've pointed out, this gets to be an even bigger mess with
> larger SPLs that want to pull in bigger chunks of U-Boot based on
> CONFIG_ symbols.  Are we going to have to create an SPL version of each
> of them, too?  And a third version of each symbol, for a three-stage
> boot?

I don't see why that should be the case?  AFAICT so far
CONFIG_SYS_TEXT_BASE is one of a very small number of highly special
variable definitions that need to be exported to the Makefile context.

> Are you going to insist that we rip out all the other CONFIG_ symbols
> that are conditional on NAND_SPL (but happen to not be referred to in
> makefiles -- is that really the meaningful semantic difference here)?
> CONFIG_SYS_HRCW_HIGH versus CONFIG_SYS_HRCW_HIGH_SPL?
> CONFIG_DEFAULT_IMMR versus CONFIG_DEFAULT_IMMR_SPL?

Why should I?  None of these are needed anywhere in the Makefiles, are
they?

> Should we replace CONFIG_NS16550_MIN_FUNCTIONS with directly ifdeffing
> NAND_SPL in the serial code?

I think you misinterpret the situation.

> It is the same purpose.  The problem is treating separate, only
> somewhat related image builds as being the same build and requiring
> that they have the same config.  They have separate makefiles, separate
> linker scripts, separate everything except some shared code... and
> autoconf.mk.  Just because we got away with it until now does not make
> it the right thing to do.

Right - "separate makefiles, separate linker scripts, separate
everything" _and_ separate text base addresses.


Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"I didn't know it was impossible when I did it."