Paul Eggert wrote: > > is it really worth making them > > pass on old platforms like this, which are using very old Unicode data > > files? Certainly, these tests should be the least of your concerns if > > you are using Linux 2.6. > > In that case perhaps we should just skip the tests on old platforms. > What would be a good way to do that?
Maybe change the expression (defined __GLIBC__ && __GLIBC__ == 2 && __GLIBC_MINOR__ >= 13 && __GLIBC_MINOR__ <= 15) to (defined __GLIBC__ && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 15) ? But before doing such a change, it would be good to confirm that glibc versions < 2.13 indeed have a problem. I cannot confirm it: Compiling m4-1.4.21 on - CentOS 5 (glibc 2.5), - CentOS 6 (glibc 2.12) does not show these test failures. In summary, I think we better leave things as they are. People are free to not run "make check" if they are building on systems from 2008. Bruno
