| Message ID | 1310071467-9017-1-git-send-email-computersforpeace@gmail.com |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, Jul 7, 2011 at 16:44, Brian Norris wrote: > When using "make CROSS=mipsel-linux-", I get the following errors: > > /bin/sh: /home/norris/git/mtd-utils/mipsel-linux/include/version.h.tmp: No such file or directory > make: *** [/home/norris/git/mtd-utils/mipsel-linux/include/version.h.tmp] Error 1 > > So instead of "building" our version.h within $(BUILDIR), we should > just build it in the main source directory, as it isn't architecture- > specific or anything. Hopefully that doesn't "break the rules" of our > cross-compilation setup... i dont think it's just a matter of "diff build dir for diff toolchains, so common version.h is ok". it's a matter of "building out of tree should build out of tree". i think the fix to your problem is: $(BUILDDIR)/include/version.h.tmp + $(Q)mkdir -p $(dir $@) $(Q)echo '#define VERSION "$(VERSION)"' > $@ -mike
On Thu, Jul 7, 2011 at 1:48 PM, Mike Frysinger <vapier.adi@gmail.com> wrote: > i dont think it's just a matter of "diff build dir for diff > toolchains, so common version.h is ok". it's a matter of "building > out of tree should build out of tree". OK, but what is the point of in- vs. out-of-tree? So that we can have a readonly source directory where the mtd-utils code stays? Or so that we keep object files away from code? Or some other reason? > i think the fix to your problem is: > $(BUILDDIR)/include/version.h.tmp > + $(Q)mkdir -p $(dir $@) > $(Q)echo '#define VERSION "$(VERSION)"' > $@ Except that gives this error: CC ftl_format.o CC lib/libmtd.o In file included from lib/libmtd.c:40:0: ./include/common.h:28:21: fatal error: version.h: No such file or directory compilation terminated. make: *** [/home/norris/git/mtd-utils/mipsel-linux/lib/libmtd.o] Error 1 Really, the build system is setup to look for headers in-tree, not out-of-tree. I'm not sure a good way to fix this other than to generate that header in-tree...any better suggestions? Brian
On Thu, Jul 7, 2011 at 17:09, Brian Norris wrote: > On Thu, Jul 7, 2011 at 1:48 PM, Mike Frysinger wrote: >> i dont think it's just a matter of "diff build dir for diff >> toolchains, so common version.h is ok". it's a matter of "building >> out of tree should build out of tree". > > OK, but what is the point of in- vs. out-of-tree? So that we can have > a readonly source directory where the mtd-utils code stays? this one imo, but maybe Artem can speak to the origins (which predate my contributions) >> i think the fix to your problem is: >> $(BUILDDIR)/include/version.h.tmp >> + $(Q)mkdir -p $(dir $@) >> $(Q)echo '#define VERSION "$(VERSION)"' > $@ > > Except that gives this error: > CC ftl_format.o > CC lib/libmtd.o > In file included from lib/libmtd.c:40:0: > ./include/common.h:28:21: fatal error: version.h: No such file or directory > compilation terminated. > make: *** [/home/norris/git/mtd-utils/mipsel-linux/lib/libmtd.o] Error 1 > > Really, the build system is setup to look for headers in-tree, not > out-of-tree. I'm not sure a good way to fix this other than to > generate that header in-tree...any better suggestions? top level Makefile: -CPPFLAGS += -I./include -I./ubi-utils/include $(ZLIBCPPFLAGS) $(LZOCPPFLAGS) +CPPFLAGS += -I./include -I$(BUILDDIR)/include -I./ubi-utils/include $(ZLIBCPPFLAGS) $(LZOCPPFLAGS) -mike
On Thu, Jul 7, 2011 at 2:12 PM, Mike Frysinger <vapier.adi@gmail.com> wrote: > top level Makefile: > -CPPFLAGS += -I./include -I./ubi-utils/include $(ZLIBCPPFLAGS) $(LZOCPPFLAGS) > +CPPFLAGS += -I./include -I$(BUILDDIR)/include -I./ubi-utils/include > $(ZLIBCPPFLAGS) $(LZOCPPFLAGS) I still get the same error. The rule for: CC lib/libmtd.o doesn't include $(CPPFLAGS). Is this intentional? See: %: %.o $(LDDEPS) $(call BECHO,LD) $(Q)$(CC) $(CFLAGS) $(LDFLAGS) $(LDFLAGS_$(notdir $@)) -g -o $@ $^ $(LDLIBS) $(LDLIBS_$(notdir $@)) Anyway, it's hard to trade lines about Makefile fixes, since one fix will pop up another problem in hacked-together build systems like mtd-utils...can you reproduce my bug completely, Mike? Or shall I continue trial and error until we find something good enough for both of us? :) Brian
On Thu, Jul 7, 2011 at 17:36, Brian Norris wrote: > On Thu, Jul 7, 2011 at 2:12 PM, Mike Frysinger wrote: >> top level Makefile: >> -CPPFLAGS += -I./include -I./ubi-utils/include $(ZLIBCPPFLAGS) $(LZOCPPFLAGS) >> +CPPFLAGS += -I./include -I$(BUILDDIR)/include -I./ubi-utils/include >> $(ZLIBCPPFLAGS) $(LZOCPPFLAGS) > > I still get the same error. > > The rule for: > CC lib/libmtd.o > doesn't include $(CPPFLAGS). Is this intentional? See: > %: %.o $(LDDEPS) > $(call BECHO,LD) > $(Q)$(CC) $(CFLAGS) $(LDFLAGS) $(LDFLAGS_$(notdir $@)) -g -o > $@ $^ $(LDLIBS) $(LDLIBS_$(notdir $@)) that's a link, not a compile, so CPPFLAGS shouldnt be needed as it should only be used on *.o files. that also shows "BECHO,LD" but your quoted output is "CC ...", and the CC rule does use CPPFLAGS. > Anyway, it's hard to trade lines about Makefile fixes, since one fix > will pop up another problem in hacked-together build systems like > mtd-utils...can you reproduce my bug completely, Mike? Or shall I > continue trial and error until we find something good enough for both > of us? :) np ... i'll look at it -mike
On Thu, 2011-07-07 at 17:12 -0400, Mike Frysinger wrote: > On Thu, Jul 7, 2011 at 17:09, Brian Norris wrote: > > On Thu, Jul 7, 2011 at 1:48 PM, Mike Frysinger wrote: > >> i dont think it's just a matter of "diff build dir for diff > >> toolchains, so common version.h is ok". it's a matter of "building > >> out of tree should build out of tree". > > > > OK, but what is the point of in- vs. out-of-tree? So that we can have > > a readonly source directory where the mtd-utils code stays? > > this one imo, but maybe Artem can speak to the origins (which predate > my contributions) Sorry, I do not know about the origins, I came into MTD area when most of the things have been already implemented and worked.
May I get I push for these?
On Tue, 2011-07-19 at 11:17 -0700, Brian Norris wrote:
> May I get I push for these?
Done, sorry and thanks for reminding!
diff --git a/Makefile b/Makefile index afbe201..7c20ea5 100644 --- a/Makefile +++ b/Makefile @@ -38,7 +38,7 @@ TARGETS = $(BINS) TARGETS += lib/libmtd.a TARGETS += ubi-utils/libubi.a -OBJDEPS = $(BUILDDIR)/include/version.h +OBJDEPS = include/version.h include common.mk @@ -53,7 +53,7 @@ endif find $(BUILDDIR)/ -xdev \ '(' -name '*.[ao]' -o -name '.*.c.dep' ')' \ -exec rm -f {} + - rm -f $(BUILDDIR)/include/version.h + rm -f include/version.h $(MAKE) -C $(TESTS) clean install:: ${BINS} ${SCRIPTS} @@ -68,9 +68,9 @@ tests:: cscope: cscope -bR -$(BUILDDIR)/include/version.h: $(BUILDDIR)/include/version.h.tmp +include/version.h: include/version.h.tmp $(Q)cmp -s $@ $@.tmp && rm -f $@.tmp || mv $@.tmp $@ -$(BUILDDIR)/include/version.h.tmp: +include/version.h.tmp: $(Q)echo '#define VERSION "$(VERSION)"' > $@ #
When using "make CROSS=mipsel-linux-", I get the following errors: /bin/sh: /home/norris/git/mtd-utils/mipsel-linux/include/version.h.tmp: No such file or directory make: *** [/home/norris/git/mtd-utils/mipsel-linux/include/version.h.tmp] Error 1 So instead of "building" our version.h within $(BUILDIR), we should just build it in the main source directory, as it isn't architecture- specific or anything. Hopefully that doesn't "break the rules" of our cross-compilation setup... Signed-off-by: Brian Norris <computersforpeace@gmail.com> --- Makefile | 8 ++++---- 1 files changed, 4 insertions(+), 4 deletions(-)