Re: Bug#639010: tablix2: FTBFs on hurd-i386
Svante Signell, le Wed 24 Aug 2011 08:11:20 +0200, a écrit :
> On Tue, 2011-08-23 at 19:30 +0200, Samuel Thibault wrote:
> > Svante Signell, le Tue 23 Aug 2011 19:24:52 +0200, a écrit :
> > > > Well, it's a GNU/Hurd bug, then. I have commited a fix to libpthread.
> > >
> > > Aha, I didn't know that the problem was due to Hurd.
> >
> > Well, since it's Hurd-specific, it must be a problem in Hurd...
> >
> > > Are you changing the system header files?
> >
> > The libpthread header, whose time.h inclusion is spurious.
>
> A question: Is there any logic written down somewhere which header files
> includes which?
"Simply" the posix standard.
Never overexpose names.
In this case, you're not supposed to include time.h since the user
didn't ask for it. So I've made libpthread include bits/types.h instead,
which only declares __foo symbols, which are reserved anyway so don't
hurt.
> Trying to find out why time.h was dragged by including
> stdlib.h was a mess.
Put #warning where the time() function was defined gave me the answer
immediatly.
> Is it really necessary to have this header files includes each other
> mess??
Unfortunately, yes, because <sys/types.h> needs to provide some pthread
types for some reason I can't remember now, and such thing.
Samuel
Reply to: