Re: XEmacs' pathname canonicalization versus Red Hat Linux 7's realpath()

Ulrich> Alexandre Oliva <aoliva(a)redhat.com&gt; writes: Ulrich> I guess this can be done. I've looked around a bit and found that Ulrich> other systems have a limit of 20. So I bumped it. I think 20 is too low. I worked in an environment years ago we had a real problem with the limit of 20, that required reengineering the way the project was organized. There can be good reasons for using symlinks, for example imagine many trees of sparsely populated symlinks pointing at other such trees. I recommend at least bumping the limit to 32, which is XEmacs' hardcoded limit. I believe the best thing is to not to have such an arbitrary limit, but to implement "tortoise-hare". Ulrich, would you accept a change to libc's realpath that would eliminate the limit, with no performance hit unless the hop count goes beyond, say, 16? Ulrich> But this does not excuse you from cleaning up your mess. The GNU way is to not have arbitrary limits.