Skip to contentSkip to conversation

Add support PLATFORM_LDFLAGS to cmd_link_o_target

PATCH New Nobuhiro Iwamatsu 2 participants 7 messages Dec 25, 2010 – Dec 27, 2010 SuperH

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 Add support PLATFORM_LDFLAGS to cmd_link_o_target

Current cmd_link_o_target function in config.mk does not support the set of the endian.

Some architecture is bi-endian (e.g. mips and sh).
Therefore, there is case supporting big endian and little endian with one toolchain.
For example, when user builds target of big endian in host of little endian, they need set endian.

---
 config.mk |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
config.mk +1 -1
index c6d6f7b..bbcefbe 100644
@@ -262,7 +262,7 @@ $(obj)%.s:	%.c
 
 # If the list of objects to link is empty, just create an empty built-in.o
 cmd_link_o_target = $(if $(strip $1),\
-		      $(LD) -r -o $@ $1 ,\
+		      $(LD) $(ENDIANNESS) -r -o $@ $1 ,\
 		      rm -f $@; $(AR) rcs $@ )
 
 #########################################################################
Signature · 1 line
1.7.2.3
#1 lore raw

PATCH 1/2 Add support PLATFORM_LDFLAGS to cmd_link_o_target

Oh, sorry.
I mistake subject...
This is not for 2010.12..

Nobuhiro

2010/12/26 Nobuhiro Iwamatsu <[email protected]>:

Quoted 31 lines

Current cmd_link_o_target function in config.mk does not support the set of the endian.

Some architecture is bi-endian (e.g. mips and sh).
Therefore, there is case supporting big endian and little endian with one toolchain.
For example, when user builds target of big endian in host of little endian, they need set endian.

---
?config.mk | ? ?2 +-
?1 files changed, 1 insertions(+), 1 deletions(-)

config.mk +0 -0
index c6d6f7b..bbcefbe 100644
@@ -262,7 +262,7 @@ $(obj)%.s: ?%.c
 

?# If the list of objects to link is empty, just create an empty built-in.o ?cmd_link_o_target = $(if $(strip $1),\
- ? ? ? ? ? ? ? ? ? ? $(LD) -r -o $@ $1 ,\
+ ? ? ? ? ? ? ? ? ? ? $(LD) $(ENDIANNESS) -r -o $@ $1 ,\
? ? ? ? ? ? ? ? ? ? ?rm -f $@; $(AR) rcs $@ )

?######################################################################### --
1.7.2.3

Signature · 3 lines
Nobuhiro Iwamatsu
?? iwamatsu at {nigauri.org / debian.org}
?? GPG ID: 40AD1FA6
#2 lore raw

PATCH 2/2 sh: Add ENDIANNESS to config.mk of sh2

In reply to Nobuhiro Iwamatsu
---
 arch/sh/cpu/sh2/config.mk |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
arch/sh/cpu/sh2/config.mk +2 -1
index 52d5a0f..02ef981 100644
@@ -21,6 +21,7 @@
 # MA 02111-1307 USA
 #
 #
+ENDIANNESS = -EB
 PLATFORM_CPPFLAGS += -m3e -mb
 PLATFORM_RELFLAGS += -ffixed-r13
-PLATFORM_LDFLAGS += -EB
+PLATFORM_LDFLAGS += $(ENDIANNESS)
Signature · 1 line
1.7.2.3
#3 lore raw

PATCH 1/2 Add support PLATFORM_LDFLAGS to cmd_link_o_target

In reply to Nobuhiro Iwamatsu

config.mk $(obj)%.s: %.c · line 265

Quoted 13 lines · On Saturday, December 25, 2010 18:02:26 Nobuhiro Iwamatsu wrote:

Current cmd_link_o_target function in config.mk does not support the set of the endian.

Some architecture is bi-endian (e.g. mips and sh). Therefore, there is case supporting big endian and little endian with one toolchain.
For example, when user builds target of big endian in host of little endian, they need set endian.

# If the list of objects to link is empty, just create an empty built-in.o cmd_link_o_target = $(if $(strip $1),\

config.mk +1 -1
-		      $(LD) -r -o $@ $1 ,\
+		      $(LD) $(ENDIANNESS) -r -o $@ $1 ,\

i dont think we should start declaring random new variables with specific purposes. better to split the "u-boot final" LDFLAGS out into their own variable (LDFLAGS_u-boot) and keep LDFLAGS as a "these are the flags that need to be used with $(LD)".
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20101226/ddb419ca/attachment.pgp

#4 lore raw

PATCH 1/2 Add support PLATFORM_LDFLAGS to cmd_link_o_target

Hi, mike.

2010/12/27 Mike Frysinger <[email protected]>:

Quoted 10 lines · On Saturday, December 25, 2010 18:02:26 Nobuhiro Iwamatsu wrote:
Quoted 9 lines

Current cmd_link_o_target function in config.mk does not support the set of the endian.

Some architecture is bi-endian (e.g. mips and sh). Therefore, there is case supporting big endian and little endian with one toolchain.
For example, when user builds target of big endian in host of little endian, they need set endian.

Quoted 7 lines
Quoted 4 lines

?# If the list of objects to link is empty, just create an empty built-in.o ?cmd_link_o_target = $(if $(strip $1),\
- ? ? ? ? ? ? ? ? ? $(LD) -r -o $@ $1 ,\
+ ? ? ? ? ? ? ? ? ? $(LD) $(ENDIANNESS) -r -o $@ $1 ,\

i dont think we should start declaring random new variables with specific purposes.

I agree. But....

Quoted 3 lines

better to split the "u-boot final" LDFLAGS out into their own variable (LDFLAGS_u-boot) and keep LDFLAGS as a "these are the flags that need to be used with $(LD)".

cmd_link_o_target is not used in the last of u-boot (u-boot final) . But this is used in the middle of build.
For example, when we make libstubs.o in examples/standalone/Makefile.

: examples/standalone/Makefile
-----
89
90 all: $(obj).depend $(OBJS) $(LIB) $(SREC) $(BIN) $(ELF)
91
92 ######################################################################### 93 $(LIB): $(obj).depend $(LIBOBJS)
94 $(call cmd_link_o_target, $(LIBOBJS))
95
96 $(ELF):
-----

Therefore, I think that we have to add a new variable to this.

Best regards,
Nobuhiro

Signature · 1 line
Nobuhiro Iwamatsu
#5 lore raw

PATCH 1/2 Add support PLATFORM_LDFLAGS to cmd_link_o_target

Quoted 19 lines · On Sunday, December 26, 2010 23:17:39 Nobuhiro Iwamatsu wrote:

2010/12/27 Mike Frysinger <[email protected]>:

Quoted 10 lines · On Saturday, December 25, 2010 18:02:26 Nobuhiro Iwamatsu wrote:
Quoted 9 lines

Current cmd_link_o_target function in config.mk does not support the set of the endian.

Some architecture is bi-endian (e.g. mips and sh). Therefore, there is case supporting big endian and little endian with one toolchain.
For example, when user builds target of big endian in host of little endian, they need set endian.

Quoted 7 lines

config.mk $(obj)%.s: %.c · line 265

Quoted 4 lines

# If the list of objects to link is empty, just create an empty
built-in.o cmd_link_o_target = $(if $(strip $1),\

config.mk +1 -1
-                   $(LD) -r -o $@ $1 ,\
+                   $(LD) $(ENDIANNESS) -r -o $@ $1 ,\

i dont think we should start declaring random new variables with specific purposes.

Quoted 7 lines

I agree. But....

Quoted 3 lines

better to split the "u-boot final" LDFLAGS out into their own variable (LDFLAGS_u-boot) and keep LDFLAGS as a "these are the flags that need to be used with $(LD)".

cmd_link_o_target is not used in the last of u-boot (u-boot final) .

this is irrelevant to what i suggested

Quoted 17 lines

But this is used in the middle of build.
For example, when we make libstubs.o in examples/standalone/Makefile.

: examples/standalone/Makefile

-----
89
90 all: $(obj).depend $(OBJS) $(LIB) $(SREC) $(BIN) $(ELF)
91
92
#########################################################################
93 $(LIB): $(obj).depend $(LIBOBJS)
94 $(call cmd_link_o_target, $(LIBOBJS))
95
96 $(ELF):
-----

Quoted 1 line

Therefore, I think that we have to add a new variable to this.

no, we dont. do as i suggested:
- add $(LDFLAGS_$(@F)) to Makefile:GEN_UBOOT
- move -T/-B flags from Makefile:LDFLAGS to Makefile:LDFLAGS_u-boot
- move --gc-sections from LDFLAGS to LDFLAGS_u-boot in arch/*/config.mk
- add $(LDFLAGS) to cmd_link_o_target

your patch doesnt scale to address all problems
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20101226/0c790267/attachment.pgp

#6 lore raw

PATCH 1/2 Add support PLATFORM_LDFLAGS to cmd_link_o_target

2010/12/27 Mike Frysinger <[email protected]>:

Quoted 49 lines · On Sunday, December 26, 2010 23:17:39 Nobuhiro Iwamatsu wrote:
Quoted 19 lines

2010/12/27 Mike Frysinger <[email protected]>:

Quoted 10 lines · On Saturday, December 25, 2010 18:02:26 Nobuhiro Iwamatsu wrote:
Quoted 9 lines

Current cmd_link_o_target function in config.mk does not support the set of the endian.

Some architecture is bi-endian (e.g. mips and sh). Therefore, there is case supporting big endian and little endian with one toolchain.
For example, when user builds target of big endian in host of little endian, they need set endian.

Quoted 7 lines
Quoted 4 lines

?# If the list of objects to link is empty, just create an empty built-in.o cmd_link_o_target = $(if $(strip $1),\
- ? ? ? ? ? ? ? ? ? $(LD) -r -o $@ $1 ,\
+ ? ? ? ? ? ? ? ? ? $(LD) $(ENDIANNESS) -r -o $@ $1 ,\

i dont think we should start declaring random new variables with specific purposes.

Quoted 7 lines

I agree. But....

Quoted 3 lines

better to split the "u-boot final" LDFLAGS out into their own variable (LDFLAGS_u-boot) and keep LDFLAGS as a "these are the flags that need to be used with $(LD)".

cmd_link_o_target is not used in the last of u-boot ?(u-boot final) .

this is irrelevant to what i suggested

Quoted 17 lines

But this is used in the middle of build.
For example, when we make libstubs.o in examples/standalone/Makefile.

: examples/standalone/Makefile

-----
?89
?90 all: ? ?$(obj).depend $(OBJS) $(LIB) $(SREC) $(BIN) $(ELF) ?91
?92
#########################################################################
93 $(LIB): $(obj).depend $(LIBOBJS)
?94 ? ? $(call cmd_link_o_target, $(LIBOBJS))
?95
?96 $(ELF):
-----

Quoted 1 line

Therefore, I think that we have to add a new variable to this.

Quoted 7 lines

no, we dont. ?do as i suggested:
?- add $(LDFLAGS_$(@F)) to Makefile:GEN_UBOOT
?- move -T/-B flags from Makefile:LDFLAGS to Makefile:LDFLAGS_u-boot ?- move --gc-sections from LDFLAGS to LDFLAGS_u-boot in arch/*/config.mk ?- add $(LDFLAGS) to cmd_link_o_target

your patch doesnt scale to address all problems

OK, I understood your indication.
I will send new patch.

Thanks!

Nobuhiro

Signature · 1 line
Nobuhiro Iwamatsu