Re: 2.2.9 release status

[karlheg@bittersweet.inetarena.com (Karl M. Hegbloom)]

>>>>> "Adam" == Adam Di Carlo <adam@onshore.com> writes:

    >> Let me get two other things done this evening before you release 2.2.9
    >> please?
    >> 
    >> 1) remove sun4c images (very simple)
    >> 2) add tftpboot-noultra image (fairly simple)

    Adam> No problemo.  I don't even know if the critical busybox problems can
    Adam> be fixed today.

 If Erik or anyone else doesn't do it before I'm finished with the
 "/vmlinuz" symlink and "lilo.conf" thing, I'll try and see if I can
 figure out what's wrong and fix it.  I believe that things that are
 trying to write to the syslog are blocking. `cardmgr' is holding them
 back; when it's killed, they unblock and proceed.

 I would also like to add floppy disk detection, so that
 `choose_medium()' doesn't offer fd1 when it doesn't exist.  My
 initial experiments show that:

 Open of /dev/fd0 succeeds right away if there is a floppy in the
 drive.  It returns a file handle, and an HDIOGETGEO ioctl on that
 returns the geometry, therefore it exists.  If there is _not_ a
 floppy in the drive, the `open()' returns right away, but fails with
 "device not configured", and does not return a filehandle.  That's
 ok, the drive exists, but there's not a disk in the drive.

 Open of /dev/fd1 hangs for about 20 seconds when there is no second
 floppy drive, but it returns a filehandle, then the HDIOGETGEO fails
 with no such device or somesuch.  With some kernel configurations
 (Erik?  Will you help me on this?) a bunch of floppy status data is
 dumped to the console at this point.  If the open was wrapped with an
 `alarm()' or something, and we assume that if it cannot `open()' the
 device in 1 or 2 seconds then there is not a second floppy drive, it
 might work... but we don't want a bunch of kernel messages messing up
 the interface.