This is not a patch against GCC trunk. Please try again with a patch that
is diffed against current GCC SVN trunk and that builds cleanly (including
building with --enable-werror-always, starting from a native bootstrapped
build of current trunk as the compiler you use to compile your cross
compiler). For example, it includes the obsolete poisoned target macros
TARGET_VERSION and IRA_COVER_CLASSES, and the handle_option hook has not
been updated for the interface changes to avoid global state in such
hooks. Furthermore, the code has at least one old-style function
definition (notice_update_cc), which should have been caught by the
warnings in the build. Beyond anything caught by warnings, if the code
defines any target macros that are in the process of being converted to
hooks (default hook definitions in targhooks.c that use the macro, but no
code elsewhere in the compiler using the macro, only defining it) then
please convert the target to define the hooks directly and not the
obsolescent macros. FUNCTION_VALUE is one example of an obsolescent macro
still defined by some ports, but that new ports should not define.
GO_IF_LEGITIMATE_ADDRESS is another.
Please see the GNU Coding Standards regarding proper formatting of
diagnostic messages (don't start with an uppercase letter or end with
".").
The toplevel changes are incorrect. See my recent cleanups. There is no
such target as target-mudflap, and there is generic code to handle
target-libmudflap depending on target OS so the cases based on target
architecture shouldn't care about that. Furthermore, a case for
cr16-uclinux can never match, because config.sub maps that to
cr16-unknown-elf. If you want to do any toplevel changes in future
relating to a GDB port for uClinux (say) then you'll need to adjust
config.sub in config.git first.
Remove .file directives from your .S files unless you're really sure they
are needed and correct. See my 2008-07-02 patch.
The libstdc++-v3 changes also seem suspicious. Newlib-using ports
shouldn't need changes to crossconfig.m4 because that code isn't used for
them; see the
if test "x${with_newlib}" = "xyes"; then
conditional in configure.ac. And the comment "# SiTel" makes no sense to
me; comments should be full English sentences.
I'm still not satisfied with the forked version of the unwind code.
Please show a diff of what the minimum changes would be to the
target-independent unwind-dw2-* files for CR16 to work with those files
based on behavior in those files being conditioned as needed on target
macros defined by CR16; we should only use forked files if such changes to
condition the generic ones are unmaintainably messy.
--
Joseph S. Myers
jos...@codesourcery.com
