On Tue, Jan 5, 2021 at 5:27 AM Christian Biesinger <cbiesin...@google.com> wrote: > > On Fri, Jan 1, 2021 at 1:07 AM H.J. Lu via Gdb-patches > <gdb-patc...@sourceware.org> wrote: > > > > On Thu, Dec 31, 2020 at 3:50 PM Joseph Myers <jos...@codesourcery.com> > > wrote: > > > > > > On Sat, 19 Dec 2020, H.J. Lu via Gcc-patches wrote: > > > > > > > Work around what appears to be a GNU make bug handling MAKEFLAGS > > > > values defined in terms of make variables, as is the case for CC and > > > > friends when we are called from the top level Makefile. > > > > > > This description, and the comment in Makefile.am repeating it, is rather > > > unhelpful as it provides no way for a reader to know what the supposed bug > > > is. Reviewers need to be able to work out whether the proposed workaround > > > is correct or the right approach for working around the bug. Maintainers > > > in future need to be able to tell what the bug is. So the comment needs > > > to explain what the bug is and give a reference to a report for the bug in > > > the GNU make bug tracker, so that subsequent maintainers can look at that > > > bug to tell if the workaround is still needed at all. > > > > > > > I just copied the same workaround from other directories in GCC. > > But could you explain under which circumstances the bug happens? >
To rebuild a subdirectory with different CFLAGS/CXXFLAGS without regenerating new Makefiles, like bootstrapping GCC and doing PGO build in binutils/GDB, we pass new CFLAGS/CXXFLAGS from toplevel Makefile to the subdirectory. Without this workaround, the old CFLAGS/CXXFLAGS are used in the subdirectory. The same workaround is used in subdirectories for bootstrapping GCC. My patch extends it to GDB for PGO build. -- H.J.
