Patchwork Use __DATE__ and __TIME__ instead of date(1)

login
register
about
Submitter Patrick Georgi
Date 2010-09-24 22:31:05
Message ID <4C9D26A9.1000709@georgi-clan.de>
Download mbox | patch
Permalink /patch/1973/
State Rejected
Headers show

Comments

Patrick Georgi - 2010-09-24 22:31:05
Hi,

attached patch uses the __DATE__ and __TIME__ macros instead of date(1)
calls. User visible results are:
- different date format (no timezone included)
- different time stamps in a single build
- less calls to tools from make (and it also slightly helps ccache do
its magic)

Signed-off-by: Patrick Georgi <patrick.georgi@coresystems.de>
Peter Stuge - 2010-09-24 22:50:27
Patrick Georgi wrote:
> attached patch uses the __DATE__ and __TIME__ macros instead of date(1)
> calls. User visible results are:
> - different date format (no timezone included)

Ok.


> - different time stamps in a single build

Hm, but there's only one build.h, right?


//Peter
Patrick Georgi - 2010-09-24 22:53:18
Am 25.09.2010 00:50, schrieb Peter Stuge:
>> - different time stamps in a single build
> Hm, but there's only one build.h, right?
It's evaluated on compile time, not when build.h is generated.
Peter Stuge - 2010-09-24 22:56:06
Patrick Georgi wrote:
> >> - different time stamps in a single build
> > Hm, but there's only one build.h, right?
> It's evaluated on compile time, not when build.h is generated.

DOH! Yes of course. :) Is it too ugly to add a compile step to
expand the macro into build.h ?


//Peter
Patrick Georgi - 2010-09-24 22:58:01
Am 25.09.2010 00:56, schrieb Peter Stuge:
> DOH! Yes of course. :) Is it too ugly to add a compile step to
> expand the macro into build.h ?
While keeping all the macros that are in build.h alive? Yes.

I guess, the real question is how many times these values are used in
the first place. As far as I can see, it's only for the banners of
romstage and ramstage. Not exactly crucial.


Patrick
Peter Stuge - 2010-09-24 23:03:36
Patrick Georgi wrote:
> Am 25.09.2010 00:56, schrieb Peter Stuge:
> > DOH! Yes of course. :) Is it too ugly to add a compile step to
> > expand the macro into build.h ?
> While keeping all the macros that are in build.h alive? Yes.

And it is the same as the previous date command I guess..

Too ugly even if done right at the beginning?

--8<-- build_time.c
int main() {
printf("#define TIME \"%s\"\n#define DATE \"%s\"\n",__TIME__,__DATE__);
return 0;
}
-->8--

build.h: build_time
	build_time > build.ht
	echo ... >> build.ht


> I guess, the real question is how many times these values are used in
> the first place. As far as I can see, it's only for the banners of
> romstage and ramstage. Not exactly crucial.

No, but is there a point in having the timestamp be the same? Maybe
there's actually a point in having them be different!


//Peter
Frieder Ferlemann - 2010-09-25 13:53:07
Hi,

Am 25.09.2010 00:50, schrieb Peter Stuge:
> Patrick Georgi wrote:
>> attached patch uses the __DATE__ and __TIME__ macros instead of date(1)
>> calls. User visible results are:
>> - different date format (no timezone included)
> 
> Ok.

Maybe use a date format which is closer to ISO 8601?

> date +"%F %T"
2010-09-25 15:39:41

> date +"%FT%T"
2010-09-25T15:39:43

( http://www.uic.edu/depts/accc/software/isodates/datefmt.html#Wanted )

Greetings,
Frieder
Patrick Georgi - 2010-09-25 14:16:42
Am 25.09.2010 15:53, schrieb Frieder Ferlemann:
> Maybe use a date format which is closer to ISO 8601?
How do I specify that in "__DATE__" which this thread and patch is about?


Patrick
Frieder Ferlemann - 2010-09-25 14:45:50
Hi Patrick,

Am 25.09.2010 16:16, schrieb Patrick Georgi:
> Am 25.09.2010 15:53, schrieb Frieder Ferlemann:
>> Maybe use a date format which is closer to ISO 8601?
> How do I specify that in "__DATE__" which this thread and patch is about?

Sorry that I kind of highjacked the mail thread
to request a specific output format instead of
concentrating on the implementation.

And sorry I cannot think on how to get an ISO 8601
format derived from "__DATE__" or "__TIME__":

Unfortunately C99 (according to  WG14/N1256 ISO/IEC 9899:TC3, section 6.10.8)
specifies "__DATE__" as
... [a character string literal of the form "Mmm dd yyyy", where] ...
but would also allows for an [implementation-deļ¬ned valid date] ...

Greetings,
Frieder

Patch

Index: Makefile

===================================================================
--- Makefile	(Revision 5836)

+++ Makefile	(Arbeitskopie)

@@ -319,12 +319,12 @@ 

 	printf "#define __BUILD_H\n\n" >> $(obj)/build.ht
 	printf "#define COREBOOT_VERSION \"$(KERNELVERSION)-r$(shell if [ -d $(top)/.svn -a -f "`which svnversion`" ]; then svnversion $(top); else if [ -d $(top)/.git -a -f "`which git`" ]; then git --git-dir=/$(top)/.git log|grep git-svn-id|cut -f 2 -d@|cut -f 1 -d' '|sort -g|tail -1; fi; fi)\"\n" >> $(obj)/build.ht
 	printf "#define COREBOOT_EXTRA_VERSION \"$(COREBOOT_EXTRA_VERSION)\"\n" >> $(obj)/build.ht
-	printf "#define COREBOOT_BUILD \"`LANG= date`\"\n" >> $(obj)/build.ht
+	printf "#define COREBOOT_BUILD __DATE__ \" \" __TIME__\n" >> $(obj)/build.ht
 	printf "\n" >> $(obj)/build.ht
 	printf "#define COREBOOT_COMPILER \"$(shell LANG= $(CC) --version | head -n1)\"\n" >> $(obj)/build.ht
 	printf "#define COREBOOT_ASSEMBLER \"$(shell LANG= $(AS) --version | head -n1)\"\n" >> $(obj)/build.ht
 	printf "#define COREBOOT_LINKER \"$(shell LANG= $(LD) --version | head -n1)\"\n" >> $(obj)/build.ht
-	printf "#define COREBOOT_COMPILE_TIME \"`LANG= date +%T`\"\n" >> $(obj)/build.ht
+	printf "#define COREBOOT_COMPILE_TIME __TIME__\n" >> $(obj)/build.ht
 	printf "#define COREBOOT_COMPILE_BY \"$(subst \,@,$(shell PATH=$$PATH:/usr/ucb whoami))\"\n" >> $(obj)/build.ht
 	printf "#define COREBOOT_COMPILE_HOST \"$(shell hostname -s 2>/dev/null)\"\n" >> $(obj)/build.ht
 	printf "#define COREBOOT_COMPILE_DOMAIN \"$(shell test `uname -s` = "Linux" && dnsdomainname || domainname 2>/dev/null)\"\n" >> $(obj)/build.ht