On Monday 27 July 2009 21:36:48 Rob Landley wrote: > I don't understand the difference between ARCH_HAS_MMU and ARCH_USE_MMU. > (For a kernel, sure. For a C library, not so much.) As far as I can tell, > there isn't one, and ARCH_HAS_MMU has been obsoleted in favor of > ARCH_USE_MMU.
the HAS option is to work sanely with kconfig. arches declare whether their arch/cpu can possibly have a MMU thus preventing people from turning it on if their system doesnt actually support it (like on Blackfin or NIOS or many ARM parts). the USE option is completely up to the user as for whether they wish to build for a MMU setup > More to the point, since ARCH_USE_MMU is a dependent symbol on the one > that's selected, you can switch ARCH_USE_MMU off for any architectures, > including ones like x86_64 where ARCH_HAS_MMU is forced on. So I can treat > x86_64 as a nommu platform even though it never is, but when you actually > do this the result is a build break. Specifically: > > GEN include/bits/uClibc_config.h > make: *** No rule to make target `libc/sysdeps/linux/x86_64/crtreloc.c', > needed by `lib/crtreloc.o'. Stop. > > So the ability to do that isn't currently very useful (and the > straightforward way to implement it would be to have one symbol that's > freely selectable for all architectures anyway). so post a patch to fix the problem. there's no reason that we cant build a nommu version of uClibc for every arch. > Does anybody know anything about the future of this symbol? Is this > something that fell on the floor when Erik and Manuel and Glenn and > everybody wandered off, or is there an actual reason to have two different > symbols that seem to mean the same thing? i added it on purpose and there are no intentions of changing it in any way -mike
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ uClibc mailing list [email protected] http://lists.busybox.net/mailman/listinfo/uclibc
