Jan Hubicka <[email protected]> writes: >> + >> #endif /* AUTO_PROFILE_H */ >> diff --git a/gcc/c/Make-lang.in b/gcc/c/Make-lang.in >> index f09fc99467b..95a7dad8350 100644 >> --- a/gcc/c/Make-lang.in >> +++ b/gcc/c/Make-lang.in >> @@ -102,7 +102,7 @@ create_fdas_for_cc1: ../stage1-gcc/cc1$(exeext) >> ../prev-gcc/$(PERF_DATA) >> echo $$perf_path; \ >> if [ -f $$perf_path ]; then \ >> profile_name=cc1_$$component_in_prev.fda; \ >> - $(CREATE_GCOV) -binary ../stage1-gcc/cc1$(exeext) -gcov >> $$profile_name -profile $$perf_path -gcov_version 2 || exit 1; \ >> + $(CREATE_GCOV) -binary ../stage1-gcc/cc1$(exeext) -gcov >> $$profile_name -profile $$perf_path -gcov_version 3 || exit 1; \ > > While we want to bump up gcov version to represent 32bit discriminators, > the gcov file format is intended to be extensible. Since you are just > adding new section I think everything should just work if create_gcov is > updated to add it to version 2. (i.e. older GCC versions should ignore > it). Does that work?
It would surprise me. The gcc autofdo often ignores the lengths of headers and generally isn't very robust and can easily be crashed. -Andi
