[U-Boot] [PATCH] Add support to mkconfig for setting simple #defines in config.h

Scott Wood scottwood at freescale.com
Thu Aug 13 20:38:45 CEST 2009


Mike Frysinger wrote:
> On Thursday 13 August 2009 11:28:09 Scott Wood wrote:
>> Mike Frysinger wrote:
>>> On Friday 07 August 2009 15:22:15 Scott Wood wrote:
>>>> On Fri, Aug 07, 2009 at 09:03:03AM -0500, Kumar Gala wrote:
>>>>>  while [ $# -gt 0 ] ; do
>>>>>  	case "$1" in
>>>>>  	--) shift ; break ;;
>>>>>  	-a) shift ; APPEND=yes ;;
>>>>>  	-n) shift ; BOARD_NAME="${1%%_config}" ; shift ;;
>>>>> +	-D) shift ; DEFINES="#define ${1/=/\t}\n"${DEFINES} ; shift ;;
>>>> How about something like:
>>>>
>>>> -D) shift ; DEFINES="${1//_/ } ${DEFINES}"; shift ;;
>>> mkconfig's interpreter is /bin/sh, so if you want to use the string
>>> replace bashism, you'll have to change it to /bin/bash.  or use the
>>> POSIX: "#define ${1%%=*}\t${1#*=}\n${DEFINES}"
>> Do you have a non-bash version of my alternative (which handles multiple
>> symbols rather than one symbol/value pair per -D)?
> 
> i dont know what you mean by one symbol/value pair per -D.  i assume that the 
> -D option here has the exact same behavior as that of gcc.

That's what Kumar's patch has it do, but I'm not sure that it's the most 
useful thing to expose to the makefile.  Maybe we should use a different 
option letter (-d?) and implement the GCC-like symbol/value semantics 
later if we end up needing it.

Hopefully kconfig is merged before then. :-)

> do you have an example invocation of mkconfig to show what you mean ?

Consider a target such as BOARDNAME_66MHz_NANDBOOT_PCISLAVE.

It could then invoke "$(MKCONFIG) -D $@ -a configname arch cpu board".
mkconfig would break up the -D argument into:

#define CONFIG_BOARDNAME
#define CONFIG_66MHz
#define CONFIG_NANDBOOT
#define CONFIG_PCISLAVE

-Scott


More information about the U-Boot mailing list