On Fri, Mar 26, 2010 at 10:50 AM, Andreas Gerdd <[email protected]> wrote: > I always wonder, if an errorless warningless kernel build process is possible? > Too many warning messages appearing during system builds.
Too many for what? > They always scare me, even though the system is working fine. > Are they really not important? Don't they affect the general stability > or even maybe the security? You have to eyeball each one carefully to answer that question. There are warnings that will *not* be eliminated because the code *is* correct and would have to be made *worse* to silence the compiler. That that would be Dumb, so no, you will never see a warningless build process unless you can magically make gcc less stupid. The fact that OpenBSD supports a wide range of platforms is part of this, as the code needs to be correct on both ILP32 and LP64, with both signed char and unsigned char, etc. There are a number of places where the 'obvious' change to eliminate a warning would break the code on other platforms. "Oops". Then there are the warnings from software that the project is not ready to be the primary maintainer for, like gcc, gdb, and xenocara. For those, eliminating the warnings would just mean that merging from the master source would be more difficult. If you're concerned about those, write up a patch and submit it to the upstream maintainer. I wouldn't hold out much more for some of them, but hey, we all need a windmill to tilt at sometimes... However, there are warnings that are associated with real errors or that can be fixed without compromising quality. Many of us try to keep our eyes open for them and track them down, but when you're 80% through a change it's unwise to let yourself be distracted. If you're interested, track them down and work out patches, then post them to the tech list, like Igor Zinovik has been doing with memory and fd leaks recently (Good stuff; I think most of them have already been committed to the tree). Philip Guenther
