Eric DeCosta <edecosta@xxxxxxxxxxxxx> writes:
>> Ugly.
>>
>> One overrides backend with common, while the other one doesn't.
>> That asymmetry alone should stop us and wonder if there is something fishy
>> in the approach that can be improved. It makes it look like the word
>> "common" means something quite different between the -ipc and the -
>> settings world.
>>
>> I suspect that in both, you should not expose "unix" to this part of the
>> Makefile. Linux and macOS occasionally being similar in some places does
>> not have to be exposed here. INstead you can use backend "linux" and
>> "macos", whose C sources may include from a separate C source file whose
>> name may contain "unix". That would allow you to get rid of
>> FSMONITOR_DAEMON_COMMON in a cleaner way.
>
> Let me see if I am understanding you correctly. Are you suggesting something like:
Not really.
What I had in mind was an arrangement more like
$ for variant in linux macos; do grep include fsm-ipc-$variant.c; done
#include "fsm-ipc-common-unix.cinclude"
#include "fsm-ipc-common-unix.cinclude"
and COMPAT_OBJS knowning only about fsm-ipc-linux.o or
fsm-ipc-macos.o, depending on which platform you are building.
