Paul Eggert wrote: > 0001-gettext-h-tests-don-t-assume-ENABLE_NLS.patch I disagree with this one. Many packages invoke 'textdomain (PACKAGE)' unconditionally; therefore the unit test ought to do the same thing. If !ENABLE_NLS, this is supposed to expand to a no-op.
If this produces a gcc warning, we need to deal with the gcc warning. But changing the unit test to do things differently than the packages do is wrong. Bruno
