From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arnd Bergmann Subject: Re: Merge headerfiles for m68k and m68knommu to arch/m68k/include/asm Date: Thu, 11 Dec 2008 17:48:18 +0100 Message-ID: <200812111748.20012.arnd@arndb.de> References: <20081208205814.GA22697@uranus.ravnborg.org> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Content-Disposition: inline Sender: linux-kbuild-owner@vger.kernel.org To: Andreas Schwab Cc: Geert Uytterhoeven , Sam Ravnborg , Roman Zippel , linux-m68k , Greg Ungerer , uclinux-dev , linux-kbuild List-Id: linux-m68k@vger.kernel.org On Thursday 11 December 2008, Andreas Schwab wrote: > Geert Uytterhoeven writes: > > > None of my m68k-linux compilers set __uClinux__, as expected. > > IIUC, userspace ABI is supposed to be completely identical. The header files are currently different for the definitions of __arch__swab32, EXEC_PAGESIZE, struct sigcontext, struct siginfo, struct pt_regs (coldfire only) and COMMAND_LINE_SIZE. I don't expect the last two to cause any trouble, but the others are likely to. __arch_swab32 should be protected with __mcfisaaplus__ (as it already is on m68knommu). pt_regs on m68knommu is already screwed by itself, because it depends on the CONFIG_COLDFIRE symbol. I don't think that it's possible to build a gdb for coldfire with the current headers, because CONFIG_* symbols are not visible in user space. The signal handling ABI is currently broken on m68knommu: commit 19dbaf6f6 "m68knommu: save reg a5 on context change", changes the layout of struct sigcontext in an incompatible way from older m68knommu and all m68k versions. Prior to that, Linux-2.5.64 changed m68knommu to use the generic siginfo instead of the m68k compatible one, thereby removing uid32 support from it. This breaks when you build a libc with uid32 against the m68k headers. Arnd <><