[sfwnv-discuss] stdcxx4-4.2.1, madvise and posix_madvise

[Ben Taylor]

On Fri, Dec 18, 2009 at 3:13 PM, Ben Taylor <bentaylor.solx86 at gmail.com> 
wrote:
> I see from build 127,
>
> rw/config/stdcxx_amd64.h:#define _RWSTD_NO_MADVISE
> rw/config/stdcxx_i386.h:#define _RWSTD_NO_MADVISE
>
> I'm guessing that the madvise defined in /usr/include/sys/mman.h
> doesn't follow the correct prototype, ?since there is a new
> posix_madvise function defined.
>
> Are there any thoughts on whether there's an advantage to making
> stdcxx4 use the posix_madvise call?

Answering my own question, I would surmise the answer is probably: none.

I find this interesting comment in:

stdcxx-4.2.1/src/memattr.cpp

#  ifdef _RWSTD_OS_SUNOS

        char dummy = '\0';

        // on Solaris use mincore() instead of madvise() since
        // the latter is unreliable
        if (-1 == mincore (next, 1, &dummy))

            const int err = errno;
            errno = errno_save;

            if (ENOMEM == err)
                return next == page ? -1 : DIST (next, addr);
        }

Any comments on the preference of mincore over madvise?

Least I don't have to wedge a posix_madvise into S10's libc now...