[PATCH][v3] Argument order in compiler options

[Patrick Georgi]

Hi,

For building coreboot v3 on cygwin, these changes (among others that I 
still clean up) are needed.
Otherwise, it doesn't find the symbols in the libraries (ld arguments 
are parsed right-to-left).

Regards,
Patrick Georgi

Signed-Off-By: Patrick Georgi

Index: util/kconfig/lxdialog/Makefile
===================================================================
--- util/kconfig/lxdialog/Makefile	(revision 687)
+++ util/kconfig/lxdialog/Makefile	(working copy)
@@ -24,8 +24,8 @@

$(obj)/util/kconfig/lxdialog/lxdialog: $(obj)/dochecklxdialog $(patsubst %,$(obj)/util/kconfig/lxdialog/%,$(lxdialog-objs))
    $(Q)printf "  HOSTCC  $(subst $(shell pwd)/,,$(@))\n"
-	$(Q)$(HOSTCC) $(HOST_LOADLIBES) \
-		$(patsubst %,$(obj)/util/kconfig/lxdialog/%,$(lxdialog-objs)) -o $@
+	$(Q)$(HOSTCC) \
+		$(patsubst %,$(obj)/util/kconfig/lxdialog/%,$(lxdialog-objs)) -o $@ $(HOST_LOADLIBES)

$(obj)/util/kconfig/lxdialog/%.o: $(src)/util/kconfig/lxdialog/%.c
    $(Q)mkdir -p $(obj)/util/kconfig/lxdialog/
Index: util/kconfig/Makefile
===================================================================
--- util/kconfig/Makefile	(revision 687)
+++ util/kconfig/Makefile	(working copy)
@@ -102,11 +102,11 @@

$(obj)/util/kconfig/mconf: $(patsubst %,$(obj)/util/kconfig/%,$(mconf-objects))
    $(Q)printf "  HOSTCC  $(subst $(shell pwd)/,,$(@))\n"
-	$(Q)$(HOSTCC) $(CURSESLIBS) $(INTLLIBS) -o $@ $^
+	$(Q)$(HOSTCC) -o $@ $^ $(CURSESLIBS) $(INTLLIBS)

$(obj)/util/kconfig/conf: $(patsubst %,$(obj)/util/kconfig/%,$(conf-objects))
    $(Q)printf "  HOSTCC  $(subst $(shell pwd)/,,$(@))\n"
-	$(Q)$(HOSTCC) $(CURSESLIBS) -o $@ $^
+	$(Q)$(HOSTCC) -o $@ $^ $(CURSESLIBS)

$(obj)/util/kconfig/qconf: $(patsubst %,$(obj)/util/kconfig/%,$(qconf-objects))
    $(Q)printf "  HOSTCXX $(subst $(shell pwd)/,,$(@))\n"

[Segher Boessenkool]

> For building coreboot v3 on cygwin, these changes (among others that I 
> still clean up) are needed.
> Otherwise, it doesn't find the symbols in the libraries (ld arguments 
> are parsed right-to-left).

No they're not.  Object files are linked left to right; object files
in archives are only included if any of their symbols are required.
Because of that, in general, libraries should be last on the command
line.

Patch looks good though.

Segher

[Peter Stuge]

On Tue, Jun 03, 2008 at 04:30:42PM +0200, Patrick Georgi wrote:
> Signed-Off-By: Patrick Georgi

With the below question answered, this is:

Acked-by: Peter Stuge
> Index: util/kconfig/lxdialog/Makefile
>
> --- util/kconfig/lxdialog/Makefile	(revision 687)
> +++ util/kconfig/lxdialog/Makefile	(working copy)
> @@ -24,8 +24,8 @@
>
> $(obj)/util/kconfig/lxdialog/lxdialog: $(obj)/dochecklxdialog $(patsubst %,$(obj)/util/kconfig/lxdialog/%,$(lxdialog-objs))
>    $(Q)printf "  HOSTCC  $(subst $(shell pwd)/,,$(@))\n"
> $(Q)$(HOSTCC) $(HOST_LOADLIBES) \
> $(patsubst %,$(obj)/util/kconfig/lxdialog/%,$(lxdialog-objs)) -o $@
> $(Q)$(HOSTCC) \
> $(patsubst %,$(obj)/util/kconfig/lxdialog/%,$(lxdialog-objs)) -o $@ $(HOST_LOADLIBES)

HOST_LOADLIB*E*S

Is that last E a typo?
> $(obj)/util/kconfig/lxdialog/%.o: $(src)/util/kconfig/lxdialog/%.c
>    $(Q)mkdir -p $(obj)/util/kconfig/lxdialog/
> Index: util/kconfig/Makefile
> ===================================================================
> --- util/kconfig/Makefile	(revision 687)
> +++ util/kconfig/Makefile	(working copy)
> @@ -102,11 +102,11 @@
>
> $(obj)/util/kconfig/mconf: $(patsubst %,$(obj)/util/kconfig/%,$(mconf-objects))
>    $(Q)printf "  HOSTCC  $(subst $(shell pwd)/,,$(@))\n"
> $(Q)$(HOSTCC) $(CURSESLIBS) $(INTLLIBS) -o $@ $^
> $(Q)$(HOSTCC) -o $@ $^ $(CURSESLIBS) $(INTLLIBS)
>
> $(obj)/util/kconfig/conf: $(patsubst %,$(obj)/util/kconfig/%,$(conf-objects))
>    $(Q)printf "  HOSTCC  $(subst $(shell pwd)/,,$(@))\n"
> $(Q)$(HOSTCC) $(CURSESLIBS) -o $@ $^
> $(Q)$(HOSTCC) -o $@ $^ $(CURSESLIBS)
>
> $(obj)/util/kconfig/qconf: $(patsubst %,$(obj)/util/kconfig/%,$(qconf-objects))
>    $(Q)printf "  HOSTCXX $(subst $(shell pwd)/,,$(@))\n"

[Patrick Georgi]

Peter Stuge schrieb:
> With the below question answered, this is:
>
> Acked-by: Peter Stuge

Thanks, r690.
> HOST_LOADLIB*E*S
>
> Is that last E a typo?

I think so (and see nothing that indicates otherwise) - I opted for the 
minimal patch, but I suppose fixing it (if only to avoid this question 
the next time a change affects that variable) is more sensible.

Thanks,
Patrick