David Michael, le Mon 08 Dec 2014 13:45:50 -0500, a écrit : > On Thu, Nov 27, 2014 at 8:19 AM, Justus Winter > <4win...@informatik.uni-hamburg.de> wrote: > > +#ifdef HURD_DEFAULT_PAYLOAD_TO_PORT > > +#if HURD_DEFAULT_PAYLOAD_TO_PORT > > +/* Any non-numeric value will fail this test. If 1 (or any number) is > > + given, do not inject the default translator function. */ > > +#undef HURD_DEFAULT_PAYLOAD_TO_PORT > > +#endif > > +#else > > + import <hurd/ports.h>; > > +#define HURD_DEFAULT_PAYLOAD_TO_PORT ports_payload_get_name > > +#endif > > My glibc builds are failing, and I think it's because of this > inclusion of <hurd/ports.h> in hurd_types.defs. > > Anything that includes <hurd/signal.h> now gets <hurd/ports.h> (via > the freshly output <hurd/msg.h>), which includes other files that > require <hurd/signal.h>, but they aren't getting any of its > definitions since _HURD_SIGNAL_H is already defined.
Ok, a classical inclusion loop issue. > Do you have any suggestions to work around this? I'd say just not include hurd/ports.h, and rather include it from the users of defs.h. Yes, that sucks a bit, but it seems to me that other solutions will bite use more. Samuel