On Wed, Mar 15, 2017 at 06:18:35PM +0800, Stefan Hajnoczi wrote: > On Fri, Mar 10, 2017 at 02:49:22PM -0600, Eric Blake wrote: > > And here's where I'm stuck: the makefiles are broken. Touching > > scripts/tracetool/format/h.py does NOT cause tracetool to be re-run by a > > mere 'make'; I've had to resort to 'make -B block/trace.h-timestamp' to > > get things to rebuild. And this is in spite of the fact that h.py > > _should_ be getting listed in $(tracetool-y) by trace/Makefile.objs, and > > $(tracetool-y) is listed as a dependency of %/trace.h-timestamp in the > > top-level Makefile. I would appreciate anyone with advice or an idea on > > how to patch Makefile to get the dependency working without me having to > > manually kick it. > > Also CCing Daniel Berrange. He recently touched the tracing Makefiles > and may have ideas.
I've been looking at this and I'm damned if I understand what's broken. All the required dependancies look to be expressed in the Makefile %/trace.h: %/trace.h-timestamp %/trace.h-timestamp: $(SRC_PATH)/%/trace-events $(tracetool-y) and $(tracetool-y) expands to the list of source files tracetool-y = $(SRC_PATH)/scripts/tracetool.py tracetool-y += $(shell find $(SRC_PATH)/scripts/tracetool -name "*.py") If I do 'touch hw/net/trace-events', then hw/net/trace.h gets rebuilt, but if do 'touch scripts/tracetool.py' it doesn't get built. So somehow make seems to be dropping the $(tracetool-y) deps despite being listed against the %/trace-h-timestamp file, and despite earlier deps on trace-events being honoured Regards, Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://entangle-photo.org -o- http://search.cpan.org/~danberr/ :|
