[Bug 1607] __RTEMS_SIZEOF_VOID_P__ flawed design

[bugzilla-daemon at rtems.org]

https://www.rtems.org/bugzilla/show_bug.cgi?id=1607

--- Comment #26 from Joel Sherrill <joel.sherrill at oarcorp.com> 2010-07-16 14:15:31 CDT ---
(In reply to comment #25)
> (In reply to comment #13)
> > (In reply to comment #12)
> > > Let me see if I can sum up the primary issues:
> > > 
> > > + can't multilib the autoconf output so the use of generated sizeof can't
> > >   go in cpukit
> > Wrong.
> > 
> > You MUST NOT put your ___RTEMS_SIZEOF_VOID_P__ into an installed headers.
> <snip>
> > - autoconf detected values are configure-time detected constants.
> > - They are valid only for the CFLAGS having been used at configure-time
> > 
> > => They are WRONG and INVALID for all other compiler setups (CFLAGS,
> > multilibs).
> > 
> > => The autoconf detected constants may only be used inside of the build-tree
> > and must not be explictily exported from the build-tree (== it is wrong to
> > install them).
> > 
> > i.e. there is nothing wrong in using autoconf's SIZEOF_VOIDP during
> > compilation, except that you must not export the results in headers.
> > 
> What are some non-installed headers in which we might be able to place the
> macro?  Also, is percpu.h an installed header, and if so does that negate any
> approach we might take with autoconf?

I don't know if there is an example.  If we did this, I suppose we would
need a per_cpu_asm.h which is in libcpu and gets the sizeof with some 
autoconf generated file that is available while building libbsp and libcpu
BUT not installed to ${prefix}

-- 
Configure bugmail: https://www.rtems.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching all bug changes.