| Message ID | 1272380840-1132-1-git-send-email-froydnj@codesourcery.com |
|---|---|
| State | New |
| Headers | show |
On 04/27/2010 05:07 PM, Nathan Froyd wrote: > To restore buildability of the *-user configs, move calls to set-vpath > out from under ifdefs, and conditionally provide a path for set-vpath. Uhm, out of curiosity what make is this? Lots of patches have been committed to *-user in the meanwhile, so it looks like a problem in your make. Paolo
On Tue, Apr 27, 2010 at 06:10:50PM +0200, Paolo Bonzini wrote: > On 04/27/2010 05:07 PM, Nathan Froyd wrote: >> To restore buildability of the *-user configs, move calls to set-vpath >> out from under ifdefs, and conditionally provide a path for set-vpath. > > Uhm, out of curiosity what make is this? Lots of patches have been > committed to *-user in the meanwhile, so it looks like a problem in your > make. This happens to me with three different versions of make: - 3.80, no patches applied. - 3.81, no patches applied. - 3.81 from Ubuntu Hardy (I haven't looked to see what patches are applied). I realize that other people have been humming along with *-user builds, but perhaps the distro that they use patches make appropriately? (Certainly the set-vpath patch indicates that some distro(s), somewhere, apply some set of patches to make...?) -Nathan
>>> To restore buildability of the *-user configs, move calls to set-vpath >>> out from under ifdefs, and conditionally provide a path for set-vpath. >> >> Uhm, out of curiosity what make is this? Lots of patches have been >> committed to *-user in the meanwhile, so it looks like a problem in your >> make. > > This happens to me with three different versions of make: > > - 3.80, no patches applied. > - 3.81, no patches applied. > - 3.81 from Ubuntu Hardy (I haven't looked to see what patches are applied). Interesting. BTW I am using Fedora. Given this, Acked-by: Paolo Bonzini <pbonzini@redhat.com> Paolo
diff --git a/Makefile.target b/Makefile.target index 65beed5..e1c0f4a 100644 --- a/Makefile.target +++ b/Makefile.target @@ -79,9 +79,11 @@ signal.o: QEMU_CFLAGS += $(HELPER_CFLAGS) ######################################################### # Linux user emulator target -ifdef CONFIG_LINUX_USER +# This call (sans `if') ought to be under the ifdef below, but that +# triggers bugs in make. +$(call set-vpath,$(if $CONFIG_LINUX_USER,$(SRC_PATH)/linux-user:$(SRC_PATH)/linux-user/$(TARGET_ABI_DIR))) -$(call set-vpath, $(SRC_PATH)/linux-user:$(SRC_PATH)/linux-user/$(TARGET_ABI_DIR)) +ifdef CONFIG_LINUX_USER QEMU_CFLAGS+=-I$(SRC_PATH)/linux-user -I$(SRC_PATH)/linux-user/$(TARGET_ABI_DIR) obj-y = main.o syscall.o strace.o mmap.o signal.o thunk.o \ @@ -111,9 +113,9 @@ endif #CONFIG_LINUX_USER ######################################################### # Darwin user emulator target -ifdef CONFIG_DARWIN_USER +$(call set-vpath,$(if $CONFIG_DARWIN_USER,$(SRC_PATH)/darwin-user)) -$(call set-vpath, $(SRC_PATH)/darwin-user) +ifdef CONFIG_DARWIN_USER QEMU_CFLAGS+=-I$(SRC_PATH)/darwin-user -I$(SRC_PATH)/darwin-user/$(TARGET_ARCH) @@ -138,9 +140,9 @@ endif #CONFIG_DARWIN_USER ######################################################### # BSD user emulator target -ifdef CONFIG_BSD_USER +$(call set-vpath,$(if $CONFIG_BSD_USER,$(SRC_PATH)/bsd-user)) -$(call set-vpath, $(SRC_PATH)/bsd-user) +ifdef CONFIG_BSD_USER QEMU_CFLAGS+=-I$(SRC_PATH)/bsd-user -I$(SRC_PATH)/bsd-user/$(TARGET_ARCH)
When I try to build a linux-user target, I get: Makefile:84: *** missing `endif'. Stop. As best as I can figure out, what happens is that when we call eval from set-vpath, make is checking that there are no dangling conditionals at the end of the input. But in this case make is doing the wrong thing: the input is just what gets fed to eval, not the wider Makefile from which eval was called. To restore buildability of the *-user configs, move calls to set-vpath out from under ifdefs, and conditionally provide a path for set-vpath. Signed-off-by: Nathan Froyd <froydnj@codesourcery.com> --- Makefile.target | 14 ++++++++------ 1 files changed, 8 insertions(+), 6 deletions(-)