On 08.05.2018 16:06, Philippe Mathieu-Daudé wrote: > On 05/08/2018 10:50 AM, Thomas Huth wrote: >> On 08.05.2018 15:45, Cornelia Huck wrote: >>> On Tue, 8 May 2018 15:38:03 +0200 >>> Thomas Huth <[email protected]> wrote: >>> >>>> On 08.05.2018 15:23, Cornelia Huck wrote: [...] >>>>> Hm, I'm wondering why we do the typedef in cpu-qom.h, while other >>>>> architectures do it in their cpu.h. >>>> >>>> See: >>>> >>>> commit ef2974cc270d51959ce90df6b4d4d41635d7a603 >>>> Author: David Hildenbrand <[email protected]> >>>> Date: Wed Sep 13 15:24:02 2017 +0200 >>>> >>>> target/s390x: move some s390x typedefs to cpu-qom.h >>>> >>>> This allows us to drop inclusion of cpu_models.h in cpu-qom.h, and >>>> prepares for using cpu-qom.h as a s390 specific version of typedefs.h >>>> >>>> Signed-off-by: David Hildenbrand <[email protected]> >>>> Message-Id: <[email protected]> >>>> Reviewed-by: Thomas Huth <[email protected]> >>>> Signed-off-by: Cornelia Huck <[email protected]> >>>> >>>> Thomas >>> >>> Gargh, this is all very confusing... >> >> If you'd ask me, I'd say we should get rid of the typedefs and do it the >> Linux kernel way and enforce using "struct xyz" everywhere, then you >> also do not have this problem with typedefs.h anymore ... but well, so >> far it seems as I'm still part of a minority with this opinion here. > > Maybe not getting rid of the typedefs, but I agree with removing typedefs.h.
But you need a way to do forward declarations ... thus you need something like typedefs.h as long as you use typedefs. Otherwise older versions of GCC will choke on multiple "typedef struct XyZ XyZ" statements. Thomas
