Hi, On Sat, Feb 19, 2011 at 01:35:09PM +0200, David Cohen wrote: > >> aha, now I get it, so shouldn't the real fix be including <linux/sched.h> > >> on <linux/wait.h>, I mean, it's <linuux/wait.h> who uses a symbol > >> defined in <linux/sched.h>, right ? > > That's a tricky situation. linux/sched.h includes indirectly > linux/completion.h which includes linux/wait.h. Ok, so the real problem is that there is circular dependency between <linux/sched.h> and <linux/wait.h> > By including sched.h in wait.h, the side effect is completion.h will > then include a blank wait.h file and trigger a compilation error every > time wait.h is included by any file. true, but the real problem is the circular dependency between those files. > > Surprisingly many other files still don't seem to be affected. But this > > is actually a better solution (to include sched.h in wait.h). > > It does not affect all files include wait.h because TASK_* macros are > used with #define statements only. So it has no effect unless some > file tries to use a macro which used TASK_*. It seems the usual on > kernel is to include both wait.h and sched.h when necessary. > IMO your patch is fine. I have to disagree. The fundamental problem is the circular dependency between those two files: sched.h uses wait_queue_head_t defined in wait.h wait.h uses TASK_* defined in sched.h So, IMO the real fix would be clear out the circular dependency. Maybe introducing <linux/task.h> to define those TASK_* symbols and include that on sched.h and wait.h Just dig a quick and dirty to try it out and works like a charm -- balbi -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
