On Thu, Nov 9, 2017 at 11:18 PM, Christopher Li <sparse@xxxxxxxxxxx> wrote: > On Fri, Nov 10, 2017 at 5:26 AM, Luc Van Oostenryck > <luc.vanoostenryck@xxxxxxxxx> wrote: >>> The selfcheck on sparse-llvm.sc is independent of the CFLAGS changes. >> >> On the contrary, it's very much at the core of the way the >> different CFLAGS variable are used. > > I mean the bug is triggerable even before the CFLAGS patch applied. > >> >> Everywhere you will use something like: >> <sometarget>.o: <some CFLAGS variant> += <something> >> you will also need to add <sometarget>.sc >> It's the way you solve the problem and it may seems to you >> independent of the CFLAGS changes and yet in my version of >> the CFLAGS changes this bug was solved automatically by how >> the CFLAGS were used (I discovered the bug while writting the >> patch; you know this "ohoh, this can't possibly be correct"). > > I see. You avoid using target specific variables. That might be > a good idea. I was looking for the smallest fix to so that > don't impact too much on the other Makefile changes. > > I don't like to have CFLAGS += for every thing though. > It only works if the options always append from the tail. > If it is order sensitive, need to insert in the middle, then > you will need to have sub variable to group from any way, > like your CPPFLAGS. In the big makefile cleanup series, I've done slightly differently but yes, no more target specific vars, only <target>_CFLAGS and using 'cflags' for flags private to sparse. Aesthetically, it's not what I found the most pleasing but this does well the job (SPARSE_CFLAGS, _CFLAGS, C_FLAGS, ... could have been used too). -- Luc -- To unsubscribe from this list: send the line "unsubscribe linux-sparse" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
