On Thu, May 15, 2014 at 12:20:06PM +0400, Yury Gribov wrote:
> On 05/15/2014 12:05 PM, Konstantin Serebryany wrote:
> >No. We have to support too many build systems and hence do not want
> >any configure step.
> >All configuration has to be done in the sources.
>
> Yeah, I see your point. But filling code with magic constants isn't
> very nice either.
>
> We could make a separate CMake check like GetTargetStructSize or
> something (somewhere in projects/compiler-rt/make/platform/*.mk).
Also, I'd suggest to have a look at glibc or libgomp how a directory system
with easily overridable target specific headers is way better than a maze of
#ifdefs all around the source.
I mean glibc sysdeps/ stuff or libgomp config/ stuff (the latter is far more
limited of course).
So, individual structures/constants/defines (or related groups thereof)
could be split into separate headers, and have target specific set of
-I arguments added by the Makefile/CMake, so that you can have some
defaults, some per-CPU defaults, some OS defaults and some per-CPU OS
defaults. Say, you could have a header file with layout of kernel stat
structure, one for Linux, another one for Darwin, another one for Windows,
and then if say on ppc64-linux it is different, it could supply it's own
file that would override the generic Linux one.
Jakub
