On Wed, Aug 10, 2011 at 16:04:15 +0300, Avi Kivity wrote: > On 08/10/2011 01:04 PM, Jiri Denemark wrote: > > This patch fixes build when any of the include paths from QEMU_CFLAGS > > contains a header file with similar name to a header file in qemu > > sources. I hit it with error.h included by qapi/qapi-types-core.h. GCC > > decided to use /usr/include/alsa/error.h instead of qemu's error.h. > > > > Signed-off-by: Jiri Denemark<jdene...@redhat.com> > > --- > > rules.mak | 8 ++++---- > > 1 files changed, 4 insertions(+), 4 deletions(-) > > > > diff --git a/rules.mak b/rules.mak > > index 612ae37..1a2622c 100644 > > --- a/rules.mak > > +++ b/rules.mak > > @@ -15,21 +15,21 @@ MAKEFLAGS += -rR > > QEMU_DGFLAGS += -MMD -MP -MT $@ -MF $(*D)/$(*F).d > > > > %.o: %.c > > - $(call quiet-command,$(CC) $(QEMU_CFLAGS) $(QEMU_INCLUDES) > > $(QEMU_DGFLAGS) $(CFLAGS) -c -o $@ $<," CC $(TARGET_DIR)$@") > > + $(call quiet-command,$(CC) $(QEMU_INCLUDES) $(QEMU_CFLAGS) > > $(QEMU_DGFLAGS) $(CFLAGS) -c -o $@ $<," CC $(TARGET_DIR)$@") > > > > > > Shouldn't we use -isystem instead of -I for system headers?
That would be ideal but unfortunately QEMU_CFLAGS also contains output of pkg-config --cflags which uses -I for header paths. We would need to convert them to -isystem. And is -isystem even portable to other compilers (in case we care about that)? IMHO just moving paths to qemu sources first is easier and the result is the same. Jirka
