On Mon, Oct 27, 2014 at 10:32 PM, Andrew MacLeod <amacl...@redhat.com> wrote: > No so much preamble for this one since cgraph.h is already representative of > only things in cgraph.c. So this is mostly shuffling of include files. > > One short 'interesting' patch, and a couple of longer boring mechanical > ones. > > symbol_table::initialize() was moved to cgraph.c. The use of TDI_cgraph to > initialize the dump file was the only reason dumpfile.h was a > pre-requirement for cgraph,h, thus the dependency was removed. > > ipa-inline.h, ipa-prop.h, ipa-reference.h, and ipa-utils.h all included > cgraph.h and one or 2 other files, so I simply flattened them and copied > their include list to all places that included them. lto-streamer.h > included cgraph.h and a bunch of stuff, so I simply removed cgraph.h from it > and pushed it through to all the other places that included lto-streamer and > into the eventual source file . I then removed all duplicate includes in > all those affected files. > > At this point, no .h file includes cgraph.h any more. > > gcc-plugin.h has a few of cgraph.h's includes added to help avoid some > plugin breakage, but depending on their uses, there are some ordering issues > that could be beyond my control. > > Bootstraps with no new regressions on x86_64-unknown-linux-gnu, and is > currently running through a full set of targets from config-list.mk. > Assuming no issues, OK for mainline?
Ok. Thanks, Richard. > Andrew >
