remove more LTO remove much of the optimizer; more to do remove most of C preprocessor (except hashtables and line map) remove most of C-family frontend (except c.opt) remove some asserts like in gcc 4.5 for Modula-3 compatibility remove runtime stuff (libgcc, unwind, gthr) extern "C" everything implement the tree-nested change (note that frontends can provide custom tree codes and gimplification, so we should be able to do this work in parse.c) In combination with parse.c change, this is enough to compile all of m3core and I think the "only" problem is that some functions aren't output. To be looked into soon, so we can really use 4.6 (and then on to 4.7.., and then on to C/C++ if 4.8 isn't out by then..) (detail)
work in progress -- add #ifdef __cplusplus extern "C" This is so setting breakpoints in m3cg on Darwin works, and seems reasonable to me anyway. (detail)
work in progress -- add #ifdef __cplusplus extern "C" This is so setting breakpoints in m3cg on Darwin works, and seems reasonable to me anyway. (detail)
work in progress -- add #ifdef __cplusplus extern "C" This is so setting breakpoints in m3cg on Darwin works, and seems reasonable to me anyway. (detail)
remove invalid checks from gcc 4.2/ARM_DARWIN, for now append -gcc46 for gcc 4.6, no need for lang-specs.h in my gcc 4.6 (Makefile.in trimmed), and nice can also just cd & make for my gcc 4.6 Makefile.in (detail)
work in progress -- add #ifdef __cplusplus extern "C" This is so setting breakpoints in m3cg on Darwin works, and seems reasonable to me anyway. (detail)
merge m3-parse.h into parse.c, except for the part that confuses gengtype (it doesn't understand #if and it sees multiple typedefs of same types) (detail)
work in progress -- add #ifdef __cplusplus extern "C" This is so setting breakpoints in m3cg on Darwin works, and seems reasonable to me anyway. (detail)
add comment that we modified this file (but we could move it to a frontend file) (detail)
fix extern C..dead file..will be deleting (detail)
rename away broken m3makefile the README even says the tests don't build due to FingerP no longer being exported (detail)
- fix a gcc 4.2 (ARM_DARWIN) compilation bug flag_tree_forwprop that we turn off, is not available there GCC46_ATTRIBUTE_UNUSED wasn't defined, fixed as part of next
- remove most of the #if GCC46 and #if !GCC46 -- leaving gcc 4.6 to be mostly like gcc 4.5, which is proving almost successful, with some attendant other diffs. (more coming, more todo)
This does (further) break compilation with cm3 -DGCC46, but that either was already broken, or at least didn't work anyway. (e.g. nested function support was incomplete) And what I have saved away is definitely progress, I'm just trying to break down the commits somewhat into smaller pieces, esp. e.g. this one that clearly has no impact on the current mainline (actually most of them won't, but it will be less clear with some of them) (detail)
work in progress -- add #ifdef __cplusplus extern "C" This is so setting breakpoints in m3cg on Darwin works, and seems reasonable to me anyway. (detail)
missed files that I had deleted and readded and then edited CVS doesn't allow that easily and I messed it up (detail)
change C++ comments to C comments (maybe we can/should go back to C..maybe) (detail)
trace the var names if option_trace_all instead of option_trace_all && m3gdb -- i.e. slightly more tracing, if tracing even asked for, and no change normally (detail)
remove -y flag, it interferes with our own (detail)
gcc 4.2: directly #define LANG_HOOKS_CALLGRAPH_EXPAND_FUNCTION tree_rest_of_compilation instead of via m3_expand_function (tree fndecl)
all versions: use default alias langhook, that returns -1, instead of our own that returns 0 not tested (detail)
![[fingerprint]](/static/31949961/images/16x16/fingerprint.gif)
