uname -m returns "armv71" on Ubuntu 9.04 with i.MX51. So, OS_TEST becomes armv71. But http://mxr.mozilla.org/mozilla-central/source/configure.in#7629, 7632 AC_SUBST(BUILD_CTYPES) 7633 if test "$OS_TEST" != "arm"; then 7634 BUILD_CTYPES=1 7635 AC_DEFINE(BUILD_CTYPES) 7636 fi So we don't compare just "arm" for OS_TEST whether ARM or not.
I think this would probably be better as an explicit whitelist of architectures we support. What happens when someone compiles for Sparc or some other weirdo arch?
In theory, we do support sparc, and all the other weird arches - libffi is supposed to handle them fine (as long as they have gcc), which is why I went with a blacklist. And if it's broken on a certain arch, having people file bugs is useful, too - in the powerpc case, for example, just pulling in an upstream patch fixed it. If you think we should go with a whitelist, though, I'm fine with it. |ls| in js/ctypes/libffi/src indicates that it has at least some degree of support for: alpha arm cris frv ia64 m32r m68k mips pa powerpc s390 sh sh64 sparc x86
(It does support arm, by the way, I just turned it off for now because it's slightly broken. It won't be that hard to fix.)
Comment on attachment 403172 [details] [diff] [review] patch v1 Ok, that's a pretty big list. I guess we might as well do it this way and if someone finds it's broken they can either fix it or add it to the blacklist. I note your list contains arm, do we have a followup on fixing the ctypes arm support?
Checked in on trunk (http://hg.mozilla.org/mozilla-central/rev/f95f9da863bb). Will land on branch as well. (In reply to comment #5) > I note your list contains arm, do we have a followup on fixing the ctypes arm > support? Not yet! Will file.
Landed on 1.9.2 per bug 513783.