On Thu, 10 Mar 2022 09:17:16 GMT, Julian Waters <jwat...@openjdk.org> wrote:
>> Some of the --without options are not properly handled and will crash when >> processed (For example, --without-version-string), in other cases the >> --without-* option will actually silently produce incorrect results instead >> of actually doing what --without-* implies (For example, >> --without-build-user and all the --with-vendor-* options). The most elegant >> way to solve this would simply be to handle such cases and display warnings >> when they're encountered (or if the option is critical to the build process, >> throwing an error) >> >> Even if it doesn't make sense to pass said option however, it should display >> a warning instead of letting configure exit with a confusing error when it's >> run > > Julian Waters has updated the pull request incrementally with one additional > commit since the last revision: > > Fix syntax errors make/autoconf/jdk-version.m4 line 507: > 505: # Set vendor version string if --without is not passed > 506: # Check not required if an empty value is passed, since > VENDOR_VERSION_STRING > 507: # would then be set to "" I'm not sure that works. In general, values like these are *supposed* to have a value. The default fallback is "N/A", which is not the empty string. I see no point in allowing overriding this to be empty. It just introduces cases where the build might fail later, for something that will hardly ever be tested. make/autoconf/jdk-version.m4 line 554: > 552: elif test "x$with_macosx_bundle_build_version" = xno; then > 553: # Interpret --without-* as empty string instead of the literal "no" > 554: MACOSX_BUNDLE_BUILD_VERSION= Same objection here. And even if it works right now, if we do allow this, it means that all future changes in the build system that involves MACOSX_BUNDLE_BUILD_VERSION needs to be verified that it works with the empty string. ------------- PR: https://git.openjdk.java.net/jdk/pull/7713