On Fri, 2011-01-28 at 12:30 -0600, Scott Wood wrote:
diff --git a/config.mk b/config.mk index 5147c35..d7bb07f 100644 --- a/config.mk +++ b/config.mk @@ -260,8 +260,13 @@ $(obj)%.s: %.c #########################################################################
# If the list of objects to link is empty, just create an empty built-in.o +ifdef CONFIG_HAS_TPL +cmd_link_o_target = $(if $(strip $1),\
$(LD) -r -o $@ $1,\rm -f $@; $(AR) rcs $@ )+else cmd_link_o_target = $(if $(strip $1),\ $(LD) $(LDFLAGS) -r -o $@ $1,\ rm -f $@; $(AR) rcs $@ )
+endif
What's going on here?
For CONFIG_HAS_TPL, LDFLAGS has --gc-sections now, passing it to cmd_link_o_target here will fail in linking stage: " powerpc-none-linux-gnuspe-ld: gc-sections requires either an entry or an undefined symbol "
I think --gc-sections should go in LDFLAGS_u-boot instead.
LDFLAGS_u-boot has --gc-sections already, I did not change it. I only add --gc-sections to PLATFORM_LDFLAGS in arch/powerpc/config.mk under "ifdef CONFIG_HAS_TPL"
In any case, I don't think we want different behavior here based on whether we have TPL. Either LDFLAGS is used in partial linking, or it's not.
I don't understand why LDFLAGS was added here in patch http://lists.denx.de/pipermail/u-boot/2011-January/084705.html
It says "LDFLAGS sets necessary option by partial linking (use in cmd_link_o_target)." But without this changing, the partial linking worked well before. Please correct me if I am wrong.
So if someone can confirm LDFLAGS is not necessary to be added in cmd_link_o_target, I prefer not add it here.
Thanks.
Haiying