On Sun, May 19, 2002 at 04:19:35PM +0100, Mark Brown wrote: > One thing I have noticed with some of the patches for Hurd that I've > seen is that porters sometimes take advantage of glibc features that > aren't widely avaliable. This obviously makes it more difficult to use > the patches - upstream may well care about portability beyond > glibc-based operating systems. I'd suggest encouraging porters to check > that functions they use are widely avaliable (often the man pages have a > fairly good overview of which standards things conform to).
Sometimes it is a necessity, particularly with realpath(). However, this should always be guarded by proper autoconf checks etc. If there is a portable solution, it is preferred over the more system specific solutions. There are several people in Debian GNU/Hurd who have a copy of POSIX and experience in writing portable programs and can help out on such issues, and everybody who is working on porting issues should feel free to ask there what the best approach is. (BTW, It's better to not trust the Linux man pages too much. The "conforming to" information is useful, but the man pages in general are not reliable). Sometimes, we also do it to match the code base we are working on. If people use #ifdef __linux__ for glibc features, they should rot in hell for doing that, but well, a quick fix is to replace it with __GLIBC__, which is much better than the defined(__linux__)||defined(__GNU__). Of course, the real fix would be to use autoconf to test for the features needed (for example). Sometimes, the upstream author might disagree with how a patch is written, this is normal, esp if the patch was written up-front without consultation. Then the patch can either be a guide that shows what the Hurd requirements are (as a working example), or it can be resubmitted with upstream's author input in mind. Actually, I expect that every patch I submit gets scrutinized and reviewed. If people just apply my patch "blindly", I get nervous... Thanks, Marcus -- `Rhubarb is no Egyptian god.' Debian http://www.debian.org [EMAIL PROTECTED] Marcus Brinkmann GNU http://www.gnu.org [EMAIL PROTECTED] [EMAIL PROTECTED] http://www.marcus-brinkmann.de -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
