Re: Unexpected soft update inconsistency

[Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>]

:On Sun, Jun 11, 2006 at 05:47:31PM -0700, Matthew Dillon wrote:
:>     libc_r would then use the new entry points and would not mess with
:>     the non-blocking flag on the FP at all.
:
:While libc_r is the most typical source of non-blocking descriptors, it
:is not that uncommon. I prefer to fix the stdio code to properly deal
:with the situation.
:
:I just haven't had the time to polish it enough (yet).
:
:Joerg

    Ok.  We have a three-pronged approach.  I just committed the kernel side
    of things and I'll 'fix' libc_r tomorrow.  Joerg will continue to work
    on fixes to STDIO.

    Basically libc_r will only have to set O_NONBLOCK for accept() and
    connect() and that's it.  read/write/pread/pwrite will be converted
    to __pread and __pwrite calls, readv/writev/preadv/pwritev will be
    converted to __preadv/__pwritev, and all the socket message calls
    (recv*(), send*()) will use the new MSG_ flags.

					-Matt
					Matthew Dillon 
					<dillon@xxxxxxxxxxxxx>