Paul Eggert wrote: > > $ ../configure CC=gcc CXX=g++ CPPFLAGS=-Wall 2>&1 | tee log1 > > Oh, that enables warnings (i.e., it looks for lint), but it is not > compiling with -DGCC_LINT.
It enables warnings, because in my experience '-Wall' is a good recipe across packages (except for -Wstringop-overflow :) ). I don't use -DGCC_LINT, because that's not part of the solution to me. https://lists.gnu.org/archive/html/bug-gnulib/2023-05/msg00120.html > In general we don't need to support that > combination, any more than we need to support lint checking with random > compilers. I disagree. For 30 years, "gcc -Wall" has been and is the best recipe for getting meaningful warnings across packages. It's not like a "random compiler". > I configure > diffutils with --enable-gcc-warnings, which does more checking than > -Wall does, and also defines GCC_LINT. Indeed, I don't use --enable-gcc-warnings, because - this option is not available across all packages, - in some packages (possibly coreutils?) it also turns on '-Werror', leading to build failures, which is not the outcome I want. > It's not worth changing diffseq.h back since the change is innocuous. > But I hope we don't need to do much more of this sort of makework. For now, gettext builds without warnings with "-Wall" and gcc 13. So, for now, I'm fine. But when gcc 14 is released, with new warning options in -Wall and possibly regressions, some new workarounds may be needed. Bruno
