------- Comment #7 from rob1weld at aol dot com 2009-01-08 17:53 ------- I skirted the i386 file by adding this to the line 189 of the Makefile:
# Added - "make profiledbootstrap" is slow and we don't want to break the build # i386.c - causes "ISO C90 forbids variable length array 'vec'" i386.o-warn = -Wno-error That saves a lot of work on that 'generated file' but is not the correct fix. Building the gcc directory generates few messages compared to the number of faults thus far versus it's large size. The few that are being show are related to the "gcc/gcov.gcda", "gcc/errors.gcda" and "gcc/gcov-dump.gcda" which makes sense that we do not have those ".gcda" files. I don't know that we need to worry about making profiled versions of "gcc/genchecksum.c" and other locally used generator executables. The ada directory works very well. I could not see a single ".gcda" file miss. Once we've finished gnatbind and get on to the 'cp' directory we loose: ../../gcc_trunk/gcc/cp/cp-lang.c:132: note: file /usr/share/src/gcc_build/gcc/cp/cp-lang.gcda not found, execution counts estimated ../../gcc_trunk/gcc/cp/call.c:7312: note: file /usr/share/src/gcc_build/gcc/cp/call.gcda not found, execution counts estimated ... (so far) Every(?) file in 'cp' is unprofiled. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38753
