> CWOPTS= -Wall -Wtraditional \ > -Wshadow -Wmissing-prototypes -Wstrict-prototypes
I apologize. It troubled me that I got no warnings from kernel source but several warnings from build tools. I rechecked my test setup and found that I had appended the additional warning options only to HOSTCFLAGS, not the real CFLAGS. Adding these options to the real CFLAGS made a huge difference indeed. Both -Wshadow and -Wmissing-prototypes gave hundreds, if not thousands, of warnings. I have not yet found that they show any real problem, so far they are just noise (like using the same function names locally as the userspace C library does, although libc is nether linked nor the headers included (-nostdinc)). But I believe there could be useful warnings too. So these warnings are worth investigating but not worth including the options in kernel tree IMHO - they produce too much noise that is not useful. Kernel coding style does not mandate using function prototypes for each and every local function in C code (and rightfully so), -Wmissing-prototypes would change that with questionable gain (-Wstrict-prototypes should be enough in practice). -Wshadow gives too many false positives about C and math libraries that are not in use. That said, I think it's worth to sometimes turn on more warning options, go over the huge output and see if there are any useful warnings. Plain -W is also worhth checking, and maybe -Wcast-align. I'll see if I can do it during the holidays. -- Meelis Roos ([EMAIL PROTECTED])