On Mon, Jul 17, 2006 at 11:41:24PM +0300, Jarkko Hietaniemi wrote: > Rafael Garcia-Suarez wrote: > > On 17/07/06, Jarkko Hietaniemi <[EMAIL PROTECTED]> wrote: > >>> Can it be that -std=c89 isn't that good an idea after all? > >>> This morning I disabled Cygwin, Now I should disable HP-UX 10.20. > >> The -std=c89 works for me in > >> * Debian 3.1 x86 - gcc 3.3.5 > > > > There are many glibc versions out there, and in many of them I've > > spotted anonymous unions, C++ comments, or other things that makes > > them incompatible with standard compilers (or even non-gcc compilers.) > > I'd rather disable this for linux completely. (Maybe that means it's > > not worth enabling it at all...) > > Oh, crap. Yes, please remove the -std=c89 sections from cflags.SH > and perlhack.pod, then.
FreeBSD system headers have never had a problem with pedantic ANSI. If -c89 is less strict, then I think that it should be safe indefinitely on FreeBSD. As Rafael says, glibc headers are another matter. lcc didn't work on Linux (for me) at one point, because of gcc-isms in a system header. (IIRC there are some evil gcc __things__ you can put in source code that gcc doesn't warn on, even in strict mode. I can sort of see why the header writers would want to do this, but it does become a bit of an arms race) Is it safe on all *BSDs? Nicholas Clark
