TODO2

FIXED -fix OABI compatibility mode for the kernel.
	when starting a fresh build, the kernel configuration stored in
	local/dns323/linux- is processed by buildroot, and CONFIG_OABI_COMPAT
	is disabled!

DONE -compile for speed.
	The default buildroot configuration is now -O2.
	As a result, available flash space is not enough anymore and some
	apps (ushare and transmission) were removed from the base system and
	added as a package. Also, ext2/3/4 are now kernel modules.
	Some packages redefine CFLAGS in their Makefile, thus buildroot configured
	default is lost (as it appears first in the compiler options list)
	This has to be done package by package.
	DONE -Samba needs it
	DONE -Kernel .config (don't compile for space)

DONE -URL in ipkg web page.
	Provide the source URL, so user can click it and see the
	package	homepage

DONE -Add ffp in ffp web package manager title

FIXED -Fix User services web page when there are no services available

DONT -Fix the /sbin/rc* shortcuts. They live in /Alt-F after a reboot.
	Remove before installing new ones. Is harmless.

HACKED -The start/stop services button should wait a little before returning, as
	some services takes some time to settle (transmission, e.g.)
	added a one second delay... too much for most services,
	too small for others (transmission)

FIXED -Remount /Alt-F with inotify during installing, etc, ipkg
	-packages_ipkg_proc.cgi 
	-create instead a ipkg wrapper? yes... ipkg calls ipkg-cl
	-but not for web page, sloow.

FIXED -Fix package_ipkg.cgi
	-is slooow
	-pkg names with a '-' are wrongly parsed (- is also the field separator)

-FIXED Fix package_ipkg_proc.cgi.
	-when installing ipkg, output is generated that makes a error page appear.

FIXED fix "configfs" to handle the new packages files

-CUPS:
	DONE create user/group cups:lpadmin
	DONE fix /etc/mdev.conf to make /dev/lp* belong to lp:lp
	add conf file to cups package

FIXED -status.cgi, handle comments and cups additions in /etc/printcap

DONE -Add a don't cache passwd in the login/again page, as is done in the
	System Utilities->Change password

FIXED -After changing the host IP in the host setup page, it will not change

DONE -Sort the packages by name in ipkg web manager

DONE -Add a configure button for each inetd service
	(dropped when splited inetd from network services)

FIXED -lpstat: Connection refused in web status page, when cups is not running

DONE -change name of ipkg to Alt-F in package manager web pages

DONE -add some feedback when downloading ipkg packages

DONE -show error status if ipkg fails for some reason

FIXED -host.cgi/host_proc.cgi: there is a bug when changing workgroup. The
	new workgroup becomes "oldwk" which is the name of a hidden
	input form... it seems to only hapens the *first* time it is changed
	
DONE -ipkg: web manager only supports one feed!

DIAGNOSED -nfs not working: MTU must be the same on both hosts! or else the msg
	received packet spanning multiple descriptors
	It was using udp as transport medium, which cause that problem in Gbps networks

FIXED -web login does not work if date is incorrect and timezone has been set
	If the box time is ahead of the real time, the web cookie expiration
	date is always expired... better to remove the cookie expiration date

DONE -The "eject" script should have a status output, indicating if the device is
	or is not ejected.
	This would be used by the "Disk Maintenance" web page to change the
	"Eject" to a "Load" button.
	-now has the "-s" option

DONE -"Disk Maintenance" web page should have a StandbyNow/WakeupNow button,
	(and removing the power status label?)

DONE -create tmpfs with more space; some packages are too big to fit the
	default 30MB while downloading.
	-now uses 64MB, (assume that disk swap is active...)

FIXED -cups uses a huge amount of memory in /tmp/cache.
	See how to make cups use the disk.
	/var/cache is not configurable, so put /var/cache on disk

FIXED -when installing some packages /etc/passwd et all are changed.
	If the user does not save settings in  flash afterwards, those
	users/groups are lost.
	They should be in /alt-F/etc/passwd et all, so when installing Alt-F
	they should reappear, but NO, the box settings take effect, as
	loadsave-setting from-alt-f is run when /Alt-F is mounted.
	-- Remember, the identity is in the box flash, not on the disk. --
	-In all web pages the user is warned to save settings, if needed to

DONE -foomatic-filters: beh should be removed from the package, as it is perl based

DONE -Add start short/long smart disk test to disk utils

DONE -merge /etc/tune.conf and /etc/hdsleep.conf into /etc/misc.conf

CHECKED -check --spare-devices usage in diskpart and diskwizard with raid1, raid5

DONE -do a mdadm --scan at diskpart/diskwiz end to regenerate /etc/mdadm.comf

DONE -set disk sleep timeout in rcS, as settings are not yet loaded at start
	when hotplugging disks.

FIXED -when RAID is recovering/syncing, sysctrl is continuously reporting
	a disk wakeup.

DONE -sysctrl: add mail sent on warning/critical temperature

DONE -sysctrl: add restore (start telnetd, clear settings?) option through box back button

DONE -partition erase dd output appears on output, direct it to /dev/null

FIXED -portmap does not work in B4:
	# portmap 
	Can't modify application's text section; use the GCC option -fPIE for
	position-independent executables.

DONE -the login through the serial line should have no password!!! For recovery reasons.

DONE -diskpart: add the possibility of Save/Load the partition table to/from the
	users desktop PC.
	for now, only on the box volatile /tmp

DONE -firmware web page:
	DONE -update flash time from 90 sec to 100 sec
	-change clear/flash settings. 

DONE -diskpart web page: don't allow changing an extended partition!
	implemented so that one can change it, one can't is create one

DONE -dnsmasq: time server: offer no advertise, advertise local, advertise server-name

FIXED -fix /var/xx links. At least /var/spool is wrong, should point to disk

DONE -make dnsmasq leases file in /tmp/ instead of /var/lib/misc/ ?
	reason: this makes the disks spinup, when aufs is active.
	no problem even on reboot, as dnsmasq verifies an IP is assigned before leasing it

DONE -firmare2: add a tryIt button, that only kexec the kernel/initramfs

FIXED -new user setup: when entered directly, e.g. though firstboot, the proposed user id
	is 99, not 1000!

HALF-DONE -investigate the RAID incremental problem when using dlink formated raid.
	/usr/sbin/hot.sh does not deliberately mount RAID

FIXED -the sysctrl "left/right disk (sdb) wakeup" (only, not standby) message continue to appear
	while the raid is: State : clean, degraded, recovering, spare rebuilding 
	degraded=1, sync_action=recover, array_state=clean

DONE -diskpart/diskwizard: all services are first stopped -- should they be restarted
	afterwards? Not all, partitions have changed, some needed fs may not even exists.
	But some services, those not dependind on fs, syslog and inetd e.g.
	should be restarted.

DONE -remote filesys: the mountpoint, tipically /mnt/xxx, is volatile. It needs to be
	persistent, so one can "mount -a" on startup (rcS). Should rcS parse
	/etc/fstab and create the mountpoints?
	After starting services, rcS can do a "mount -t cifs/nfs"

DONE -raid page: can't destroy linear/raid0.	raid0/linear has to have some operations available

HALF DONE -eject: think of a possibility of stopping arrays.
	currently, eject (through hot.sh) removes the partitions of the disk to be
	ejected from a raid array that contains it. This means that it will undergo
	a lenghly rebuild when latter loaded (eject -r)
	For some operations, however, the array should be stoped (disk part, e.g)
	- eject now supports a "-a" option to "eject" all disks

DONE -nfs: create a nfs-client init script, for mounting remote filesystems, without
	the need to start the nfs server just for this.

DONE -disk details such as name, family, capacity, SN, etc, should be collected by hot.sh
	and put at /etc/bay:
	left_dev=sda
	left_cap="320 GB"
	left_fam="SAMSUNG SpinPoint T166 series"
	left_mod="SAMSUNG HD321KJ"
	left_sn="S0MQJ1QP506705"

FIXED -Partitioning: /usr/bin/eject: .: line 132: can't open '/etc/bay' 
	/etc/bay, inetd.conf, rsyncd.conf, passwd !! (at least) disappears
	-seems to be related with having Alt-F aufs mounted during the partitioning
	-it was a typo, "rm" instead of a "mv" in loadsave_settings

DONE -create a directory creation/edition web page:
	-create the directory, assign it to a group, and set permissions
	-show all dirs to modify it? No, browse for it!

DONE -offer the possibility of recovering the original firmware settings from mtdblock1 when
	flashing a vendor's firmware and in the settings web page
	-only on the Firmware Updater

FIXED -login: when the login page opens, if one ignore it and goto other page, it is accepted!!!
	NO /etc/web-secret, no /tmp/cookie, do logout and instead of supplying a password
	or submiting empty password, just navigate to a page that *should* need
	authentication. check_cookie() seems to fail, as you can go to any page!

FIXED -uhdcpc update script keeps appending Allow directives to /etc/httpd.conf

FIXED -httpd services->network->inetd->httpd: currently only the first Allowed host appears to
	be edited.
	only one host/net allowed, use a #!# comment to distinguish from local net,
	which also #!# commented

DONE -perms: add a apply recursively to all subdirs, and apply owner:group recursively to all files

FIXED -rcatd start: Cannot change to /var/lib/atjobs: Permission denied
	# l /var/lib
	drwxrwx--T    2 root     root         4096 Nov 20 05:50 atjobs
	drwxrwx--T    2 root     root         4096 Nov 20 05:50 atspool


HALF FIXED spaces in fstab: patch applied to uclibc, now fix all scripts and web page...
	DONE: nfs*.cgi and smb*.cgi
	TODO: mountpoints still get spaces replaced by underscores.
	      But first check it possible!!!

FIXED -samba credentials: the username must be the the windows one, no commas if space, do as in:
	username=Manuela Lima
	The problem is if one user has both windows and unix accounts... how to do it?

DONE -backup:
	DONE -add users to backup group at user creation time? yes. Make backup group visible? no
	DONE -the backup script should be run by the user who ownes the origin directory.
	 For nfs this makes a difference, if the original files don't have permissions
	 for others or group and nfs is root_squashed
	 A nice side effect is that the backup files also belong to the same user.
	DONE: hot.sh: create a link /Backup to point to the backup dir
	FIXED: -there are problems with spaces in fstab! 040 or \040 or \ does not seems to work
		(but filesystems mounted by label still replace spaces in labels/mp with underscore)

	-an example of cifs mount: see the above "samba credentials" entry
	 mount -t cifs -o rw,nounix,noserverino,iocharset=utf8,uid=mlima,gid=users,credentials=/etc/samba/credentials.mlima \
		"//mono/Users/Manuela Lima" /mnt/smb

FIXED package/config/mconf-qconf are binaries in SVN

FIXED vsftp*.cgi assumes capital YES/NO in vsftpd.conf, anon_root not quoted

FIXED browse_dir doesn't transverse directories with spaces in its name

HALF DONE, see below
	-remote filesys: at rcE time, they must be unmounted after killing processes
	but before stopping the services -- at least nfs! Keep reading

rcnfs-client stop: unmount first filesystem! keep reading

DONE rcE: services are stopped first to unmount filesystem, as they can be using the filesystems.
	problem is that one of those services is nfs*; when latter one tries to unmount
	nfs filesystem, nfs is already dead.
	So, stopping nfs should be conditional on the existence of nfs mounted filesystems,
	 or should it unmount filesystems? 
	there ar now nfs-client, that umounts mounted nfs filesys
	rcE calls "rcall stop" that stops nfs-client and then nfs, then "eject -a"
	
FIXMED -mkreloaded.sh tries to access ../wiki!

FIXED -eject -a: does not return error when a filesystem can't be unmounted

DONE -usersgroups should detect that home diretories are  missing, even if there
	exists users
	
DONE -The DHCP udhcpc script /usr/share/udhcpc/default.script should do more
	configuration files and daemons signaling updates.
	It is already done in rcS, use it.
	FIXED -The script, at first boot, does not has a hostname.
	DONE -The host setup web page in DHCP mode also.

FIXED -fix diskmaint_proc.cgi:
	-if changing filesystem type from/to ext to vfat/ntfs or
	 vice-versa, the partition type must be changed too
	-also, fstab must be flushed, as labels/partition might have changed

CHANGED -The disk maker name does not appear in Disc Utilities (but appears in Status)

DONE -ipkg frontend, aufs.sh, packages_ipkg_proc.cgi and hot.sh needs to be
	fixed. Each one of them mounts/create /Alt-F in its own way!
	-aufs.sh install should be called by ipkg -install
	-aufs.sh inotify/reval be called before/afters ipkg calls ipkg-cl
	-aufs.sh mount/umount be called by hot.sh
	-ipkg -install be called by packages_ipkg_proc.cgi install
	-aufs.sh be called by packages_ipkg_proc.cgi
	
UNCOVERED -big MTU (>3712) generate lots of "page allocation failure", which impair the
	performance
	mem alloc is done by slab; slab uses 4096 bytes pages; one frame seems to use
	MTU+384 bytes; with a greater MTU, slab has to allocate 2 memory-contiguos pages
	from the "buddy" allocator. At power up, that is not problem, but after memory
	being fragmented it can be.
	the network driver allocated space for ~130 frames when "ifconf eth0 mtu NN"
	is performed. See network driver memory usage by alternatimg "ifconfig ... MTU NN" 
	with "cat /proc/slabinfo | grep size-"

DONE -diskpart: needs feedback on lenghly operations! (as diskwizard does)

FIXED -spaces in filenames/paths are problematic for NFS. Use octal for spaces? '\040'

FIXED -diskpart: the free space is only calculated when one clicks another size input field.
	-recalculate on mouseout

FIXED status page: smbstatus is badly parsed when share names have spaces on it
	status.cgi show wrongs in "Remotely Mounted Filesystems" CIFS

FIXED create a spindown package, based on spindown.sh
	-sysctrl does the same as the spindown.sh script

FIXED -transmission: rcscript not OK! same problem as minidlna, runas user:group

FIXED ipkg on uinstall all packages, or "aufs -u"
	The usual problem, aufs mount resource in use/busy
	should do a "rcall stop" first!

DON'T -put hardware board level in status web page.
	done and undone. Its now in alt-f.log

DONE -the longterm kernels are in another download directory,
	http://www.kernel.org/pub/linux/kernel/v2.6/longterm/v2.6.35/patch-2.6.35.xx.bz2

DONE -disk wizard: warn users that the created partitioning is not dlink compatible, or flash Alt-F
DONE -disk partitioner: warn users to create a dlink compatible partitioning, or flash Alt-F
DONE -disk filesystem: warn users that ext4 is not dlkink compatible, or flash Alt-F

DONE -diskmain: add mount options input field, and save value, based on fs-uuid, in misc.conf

-backup: i noticed that when you do a manual 
	backup from the webinterface, you can find only the start of the 
	process in the log, not when it is completed or if it has removed a 
	old backup. Furthermore i find this in the log: 
	Jan 27 02:00:01 DNS-323 cron.err crond[500]: USER root pid 447 cmd / 
	usr/bin/backup 0 

FIXED -disk wizard: Still has one weak point: "eject -a" does not returns in/sucess,
	so the wizard continues even if some filesystems were not unmounted!
	fix eject -a!

DONE -add ipkg as a dependency of all packages, to disallow ipkg removing itself when there are
	still other packages installed (and unmanagable after that)

DONE -add <pkg>.prerm to stop daemons before removing the package (on all packages who need it)
	DONE: avahi, netatalk, transmission, cups, automatic, minidlna, ushare, mt-daapd,
		fuppes, mediatomb

FIXED -"ipkg -clean" does not run the package uninstall script!
	Issues: created users and files are not removed.

DONE -add the possibility of seeings all possible logs to the System->utilities->Logs

DONE -add yet another security measure to funplug:
	at start it checks for an alt-f.fail file, and if it exists remove the funplug
	execute permissions and stops, ignoring the ONCE settings.
	If the alt-f.fail does not exists, create it and continue with the boot.
	When Alt-F itself boots, it removes the alt-f.fail file, if it exists.

DONE -add a filter input to the logs page in System->utilities->Logs

-smb: add allowed users input field

-fun_plug: change nobody:nobody to numeric dlink ids.
	-change fun_plug ownership to nobody, or set write permissions

-offer possibility of uninstalling Alt-F, by removing the fun_plug and alt-f, Alt-F directories.
	or at least change its ownership to nobody.
	The problem is that they are owned by root, and can't be deleted using samba.
	does ftp works?

DONE -temperature in C/F

-crypto web passwd (use busybox cryptpw -- Dwight Hubbard suggestion)

-stop using the domain reported by DHCP, as not all DHCP servers allow customizing it

DONE -host setup web page: don't allow/warn for more than +/-7500 bytes for MTU in the host web page

-dir_proc.cgi has some issues with spaces in paths in the parent directory of the current
	directory

-networking: add a test button that tests internet connectivity
	-ping the gateway and nslookup a well know host

-diskpart: when creating a new raid array, when one already exists, does not takes
	into account the already allocated minor. As result, one of the two raid
	is not assembled.
	FIX: examine prefered minor on already existing raid partitions
	NO: use mdadm --examine --scan -v

-when in DHCP mode, supply -h <hostname> to udhcpc,
	is builtin in the udhcp code, to fix need to edit sources... not for now

-ntpd: /etc/ntpd.drif should be in /var/lib/ntp, not in flash, as it is likely to change
	during operation; the disadvantage is that it is lost after a reboot.
	However, when a package is installed and Alt-F aufs is working,
	the drift file is keept in disk surviving reboots.
	On the other side his has the disadvantage of waking-up the disk when
	the drift file needs to be written -- use the "driftfile driftfile [
	minutes  [ tolerance ] ]" configuration options (see ntp-4.2.4p5/html/miscopt.html)

-iconv.h is removed from the stagging dir by buildroot! it says something
	like "removing iconv.h as locale is not defined".
	It comes from uClibc
	$BLDDIR/toolchain_build_arm/uClibc-0.9.30.3/Makefile.in
	as UCLIBC_HAS_XLOCALE is not defined, iconv.h is removed, even if
	it is not from uClibc.
	Note that this happens on the *second* initial build, not the first!
	(this is do a "make -O=...", followed for a second one)
	No solution? Patch this? hmmm
	Created a uClibc-0.9.30.3-iconv-h.patch, put in toolchain/uClibc lets see...

-minidlna: iconv.h strikes again!, see above

-add -o owner:group option to "mkpkg.sh" so ipkg-build can use it to
	create the package files with another ownership

-login expires after 1 hour. Restart the countdown on each new check_cookie,
	sendind a new cookie, valid for more 10mins

-create a applyfix script to be executed by hot.sh. If a file named Alt-F-*.fix is
	found in the root directory of any mounted filesystem, apply it as a patch.
	The script shall also be directly invoked.

-Fix the pkg-config and host-pkgconfig mess
	Read the package/pkg-config.mk comments
	Merge pkg-config and host-pkgconfig, with an option to create
	only on the host

-Change Alt-F directory name to .Alt-F, to make it "invisible" to most users
	-or use the bind technique explained in aufs page?:

		hide=/tmp/.aufs.hide
		mkdir $hide
		mount -n --bind $hide /Alt-F

		When you unmount the aufs, delete/modify the branch by
		remount, or you want to show the hidden branches again,
		unmount the bound /tmp/.aufs.hide.

		umount -n /Alt-F
	-just call it .Alt-F

-long standing issue: sntp/ntpd will issue the following error with the first built:
	# sntp pt.pool.ntp.org
	sntp: can't resolve symbol '__aeabi_d2iz'
	
	this happens either with the stripped or non stripped binary.
	one must remove the binaries from the rootfs and source and rebuild
	the package. After that it works fine... build problem?

	-this also happens with package nano.

	The missing symbols are defined in /lib/libgcc_s.so.
	Why are they not used?! ld (/lib/ld-uClibc-0.9.30.3.so) problem?

	The problem is not the app, it seems to be in libm. It happens
	after removing the "project_build_arm/dns323/root" directory and
	doing a make again. The second make don't make libraries in /lib
	correctly.

	******** FOR A RELEASE, ALWAYS DO A FRESH BUILD! ***************

-read aufsutils carefully, to see if they are needed. Author says yes.
	Alt-F has already used them, but currently don't

-partition erase on usb disk gave a "not space on disk" dd error!?

-diskpart: enable Save/Load the partition table to/from the users desktop PC. See above

-check if dnsmasq.cgi fully qualifies the network, or if it restricts to class C
	
-add a suggest button in the disk partitioner.
	Users can't use the wizard if they buy a new disk!
	-OR in the disk wizzard add checkboxes to each disk. This would enable applying the
	wizard to only one disk, as when the user buys a new disk

DONE -status web page: use Dwight Hubbard technique

FIXED -diskutil.cgi: sometimes the page wakeup disks; other, hitting StandbyNow does not work.

FIXED -nfs/nfs_client: the initscript does not remove module lockd

FIXED -libgcrypt depends on libgpg-error, but libgpg-error is not built first.
	problem with hyphen `-'in libgpg-error name?
	<pkg>_DEPENDENCIES seems to only work on buildroot autotargets. For "normal" makefile
	fragments, the dependency must be explicit in the <pkg> rule

DONE -setup host, add "Enable IPV6" checkbutton. If yes, modprobe ipv6, and set USE_IPV6=yes in
	/etc/misc.conf. Avahi, when started, will load ipv6 modules.
	Will other ipv6-enabled apps do the same? Yes. At least "curl -6 <host>",
	or forked-daapd does.

-create avahi configure page, at least with use-ipv6=yes/no.

-mkpkg.sh: when creating a package, and a directory is in the list of files, the *contents*
	of the directory goes to the tar file, which is not always the desired behavior, as
	the directory, eventually empty, must exist for the daemon to start, but the
	directory contents might belong to another package.
	E.g.: /etc/dbus-1/system.d/avahi-dbus.conf belongs to avahi, but
	/etc/dbus-1/system.d must exist for dbus to start (fixed in the initscript)

DONE -cups: remove the root:password from the url, firefox is always asking for confirmation.

FIXED -sysutils, start all: seems to send data back to browser

FIXED -sysutils, view logs: reset select box on (re)load, when using the back button.

-dnsmasq: supports dbus (at compile time) and ipv6 (at compile and runtime?).
	-see if it deserves enabling ipv6 (or if it is automaticaly enabled at compile time)
	-dbus does not seems to have any advantage.

-mediatomb seems to use libdvdnav and libdvdread to scan and serve dvd iso files:
	http://juliensimon.blogspot.com/2008/12/howto-compiling-mediatomb.html

	These two libraries are used to browse DVD file structures: mediatomb
	needs them to read and stream DVD ISO images.

	You do need to build and install libdvdread before libdvdnav, or
	the latter won't link.

	ftp://ftp6.mplayerhq.hu/MPlayer/releases/dvdnav/libdvdread-4.1.3.tar.bz2
	ftp://ftp6.mplayerhq.hu/MPlayer/releases/dvdnav/libdvdnav-4.1.3.tar.bz2

-minidlna: apply patch for sansung? No, there are lots of patches, can't follow them.
	http://sourceforge.net/tracker/?func=detail&aid=3058976&group_id=243163&atid=1121518

HACKED -avahi-daemon initscript: if avahi is already running, "start" fails.
	has to check avahi-status before starting it.

FIXED -forked-daapd: when running the experimental feed on 0.1B6, there is a linking problem:
	forked-daapd: can't resolve symbol 'in6addr_any
	so the *full* initial patch is needed, retrieve it from svn.

-aufs: stop overlaying the root directory, it degrades performance!
	Without any packages installed, I got 15MB/s read transfer speeds.
	Removing all packages, no aufs in /mnt, I got 20MB/s!
	Study what directories to overlay.

-vsftpd: add support for ipv6. vsftpd.conf:
	# listens on IPv4 sockets. This directive cannot be used in conjunction
	# with the listen_ipv6 directive.
	listen=YES
	#
	# This directive enables listening on IPv6 sockets. To listen on IPv4 and IPv6
	# sockets, you must run two copies of vsftpd with two configuration files.
	# Make sure, that one of the listen options is commented !!
	#listen_ipv6=YES

	Commenting? Really? Then, listen=yes and listen_ipv6=no is not OK? 
	Anyhow, can't start 2 servers on the same port, one with listen=yes
	and the other with listen_ipv6=yes

	This seems to indicate that for ipv6 a standalone vsftpd server has to be launched,
	or will inetd do the job?

-nfs: when starting nfs, the following error shows up in syslog:
		daemon.notice rpc.statd[3062]: Unable to read state from
		'/var/lib/nfs/state': errno 5 (Input/output error)
	/var/lib/nfs/state is created by the nfs initscript. If the initscript does not creates
	it, the error does not show up. But when nfs is stopped, the file appears.
	/var/lib/nfs/state should be created by statd and updated by sm-notify,
	see sm-notify and statd manpage.
	The problem seems to be related to "rpc.statd --no-notify".
	Occurs, at least, when no /Alt-F exists.

-nfs/ipv6: when starting nfs with ipv6, at least, syslog:
	svc: failed to register lockdv1 RPC service (errno 97)
	seems to mean that nfs has not ipv6 support, the errno 97 seems to mean
	"address family not supported". To build with ipv6 support the TI-RPC library is needed.
	http://sourceforge.net/projects/libtirpc/ on a x86 the library has 164KB.
	Also, rpcbind must replace portmap for ipv6 to work,
	http://nfsv4.bullopensource.org/doc/tirpc_rpcbind.php

DONE -ipv6/host setup: disable it by blacklisting it.

DONE -debian: add an install/run debian using debootstrap-static

FIXED -diskutils.cgi not working. No smart status/tests, spindown...

hot.sh: usblp is removed when a usb pen is unpluged.
	The kernel is generating the event! why? hot.sh is just doing its work.

FIXED sysutils: Hitting "clear print queues" sends an empty file to the browser.

diskmaint_proc, common.sh (fs_progress), hot_aux. Has to fix fsck to use a standard progress
	log file, instead of duplicating stdout and stderr.
	res=$(fsck -p -C4 /dev/... 2>&1  4<> outfile)
	but resize2fs/mke2fs also produce logs
	working on it and getting bored...

FIXED status.cgi: /tmp/smt_sda is created in /tmp for smartctl, and should not be needed anymore.

FIXED samba/backup: fix the available setting in /etc/samba/smb.conf
	currently those settings are changed by the web interface when the Users/Backup directory
	is created. If one fails to save settings, there is no other possibility of changing it
	Made all shares available in the samba configure page. This way users can manually
	enable them

-web server: every connection by a client browser, even for simple pages, seems to not be
	properly closed, being left in a TIME_WAIT state.

init.d/common: restart() has to have a timeout, as well as stop()

forked: when changing config file using web page, the page does a "rcforked restart"
	this deploy avahi/dbus stopping/restarting, with an error message appearing

remove all packages using the web page: when a package to be removed is running, and other
	depends on it, a message appears saying that the service can't be stopped at it
	needed for another service that is running; the service is not stopped and the
	package is removed... create a -force?

DONE	REVERTED, /var/cache is now again on tempfs (using swap)
	/var/cache can't be on disk, it avoids spindow! samba, e.g., stores data there,
	even if not being accessed, as does nfs.
	
	The problem is cups, that also uses it when a new printer is added, and all ppd
	files as parsed. The amount of data is huge, several tens of MB. Using only
	hplip, 11MB are used, with both hplip and foomatic 47MB are used!
	This was the only reason to move /var/cache out of tempfs, but has to be reversed.
	This data is temporary, is only needed when a printer is added or modified.

	Warn the user to reboot after a new printer is added? No!
	Detect a new cups printer (difficult -- needs to keep state) and stop cups,
	clean cache and restart cups? Who will do the printcap check?
	Should status.cgi detect the situation and take the necessary actions?
	end-up creating a log cleanup service

FIXED Reboot: sys_utils_proc.cgi, when called from firmware2_proc.cgi, does nothing and goes to
	sys_utils.cgi, that downloads a error message: "Usage httpd:..."

FIXED Log files: the filter field appears with a non-filled string, "all", "downloading", or other.

DONE RAID: disable RAID5 creation if not flashed
	Already done, in a different way (asking for user confirmation)

DONE Filesystems: disable ext4 creation if not flashed
	Already done, in a different way (asking for user confirmation)

DONE -cleanup: set to not enabled on boot, it might avoid disk spindow?

FIXED -loadsave-settings: really disable boot-disabled services.

FIXED -disk-utils: if no misc.conf settings, the spindown entry is disabled,
	even is power saving is also disabled

-rewrite INSTALL: login, host, time, wizard, user, net->services-smb

-FAQ: can't login. telnet, "adjtime -adjust", or "echo NONE-O > /etc/TZ"
-FAQ: swat, cups: ask credentials: root, pass
-better first time wizard!

FIXED -load setings: is it working?
	wasn't

DONE -create openssh package (not only with sftp, as for the base system)

openssh: fix conflicts with dropbear. Currently scp and ssh from dropbear (links) are overwritten
	create own keys or use/convert dropbear ones?
	create host key
	try to ssh an external host from within the box
	try to use sshd
	

FIXED -raid shrink: is being shrink too much, when mounting error:
	EXT4-fs (md0): bad geometry: block count 884736 exceeds size of device (870892 blocks)

DONE -debian password: use the same as /etc/websecret

DONE -use "checking" instead of "cleaning" in all web pages related to fsck.

FIXED -mtu: see if the dhcp stanza in /etc/network/interfaces accepts a mtu
	it appears as a IF_MTU variable to udhcpc

DONE -backup: do a "cp -a" for the first backup, instead of using rsync.
	as a matter of fact "cp -a", "tar" and "rsync" memory usage increseases too much
	on deeper and large directories. So used "cpio" for the first backup. BUT cpio
	cant use filer bigger than 4GB, so for those do a "cp -a". 

FIXED -samba mount: error when writing to fstab, parsing url

FIXED -debian: if installed in a raid, kexec fails:
	mount: mounting /dev/md0 on /mnt/md0 failed: No such file or directory 
	Stopping sysctrl: OK. 
	Executing Debian now... 
	Cannot open `/mnt/md0/vmlinuz': (null) 
	Cannot open `/mnt/md0/vmlinuz': No such file or directory 
	Failed. 

FIXED login/new user: accept more chars for passwd
	accepted all ASCII chars in the range [!-~]

FIXED partitioner: impossible to create greater then 1GB partitions?
	reason: awk integer printf "%d" limited range, used "%.0f"
	The Wizard is reported to create partitions bigger then 1TB

fix all *-config and *.pc in staging dir, as they usualy have prefix=/usr, making 
	compilation to use the system header files.
	Test on a computer without -devel packages

	grep ^prefix= /home/jcard/build-pkgs/build_arm/staging_dir/usr/bin/*-config
	grep ^prefix= /home/jcard/build-pkgs/build_arm/staging_dir/usr/lib/pkgconfig/*.pc 

FIXED Debian 'alt-f' command only works from within a serial connection;
	Now the 'alt-f' command just changes Debian /etc/default/kexec to use Alt-F 
	kernel/initramfs and issues a reboot.
	The 'debian' Alt-F command sets Debian /etc/default/kexec to use
	Debian kernel/initramfs, so a normal 'reboot' command within Debian reboots to Debian. 

FIXED debian kexec: if started from ssh, the connection will be closed during
	the normal Alt-F shutdown (dropbear terminated); use telnet instead.
	make the 'debian' script detach. 
	now create sh script and execute with nohup in background

FIXED debian: the install/run is susceptible to device enumeration ordering,
	and if it is mounted by label the install/run will fail.

FIXED /etc/init.d/*cleanup/*backup:
	crontab: chdir(/var/spool/cron/crontabs): No such file or directory
	root: Stopping cleanup: OK.
	crontab: chdir(/var/spool/cron/crontabs): No such file or directory
	root: Stopping backup: OK.

UNCOVERED ntfs-3g_ntfsprogs still need libfuse?
	NO, is configured to use internal (light) libfuse (which is not created as a library)
	The library still needs the kernel module

Debian fsck failed at boot, out of swap/tmp memory? 

DONE rcE: umount tmpfs before turning swap off? to avoid running out of memory if the tmpfs
	filesystems uses a lot of space?
	if /tmp is full or with many files, it will be using swap; when using swapoff,
	/tmp might not have enough space on RAM, or will use all or most RAM.
	So, unmount /tmp before swapping off.

FIXED ntfs-3f-ntfsprogs: fix configfs BR2_PACKAGE_NTFS_3G_NTFSPROGS
	base fw has usr/sbin/fsck.ntfs usr/sbin/fsck.ntfs-3g usr/sbin/mkfs.ntfs and shouldnt

MB/MiB, GB/BiB: uniformize disk units, use MB/GB

FIXED sysctrl: spindown timeout of zero constantly puts the disk to sleep!

FIXED browse-dir: when nfs is mounted, 'find' gives a lot of 'not permited' errors
	no permittion read users directories...

DONE backup: do the first backup using a plain 'cp -a', faster than rsync
	use instead 'find | cpio -p' instead of 'cp -a', as 'cp -a' uses too much memory,
	using swap too much which slows down the copying
	apply 'nice' to cpio and rsync

DONE filesystems: aplly 'nice' when creating/cleaning/... filesystems

DONE directory setup: use 'find|cpio -p' instead of 'cp -a', as 'cp -a' uses too much memory,
	using swap too much which slows down the copying. Also apply nice to cpio.

FIXED busy_start/stop: error on "rm" tmp file used to generate a random obj id 

FIXED status page: remove bind mounts from filesystem section (or at least keep only one
	of the device mounts.
	Also, the RW status appears duplicated.

FIXED rcbackup: 
	more fields in /etc/backup.conf, crontab generation is wrong!
	May 29 03:17:01 nas cron.err crond[1212]: user root: parse error at /Backup
	May 29 03:17:01 nas cron.err crond[1212]: user root: parse error at /home/jcard

backup: watchout 4GB problem in cpio

ipkg install fails when ipkg.conf is inherited from a previous installation and does
	not has a valid feed

DONE /tmp size can grow... use other method to increase/decrease when adding/removing swap
	/tmp size is now 10% os swap space

icu: libicudata.so.48.0 is not stripped

/etc/inittab: remove "logging junk" entry

web smb: share permissions not right

reloaded: fix proc-arm926.S. asm/asm-offsets.h should be used for 2.6.15
	#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,15)

FIXED diskpart: js updatesect() display wrong current partition maximum size

DONE diskwizard: raid0 message: "with two (or three, one an external USB) disks."
	raid1: "duplicate everything on two disks (and use if an external USB disk if available as a spare)

raid: when adding a component that has/was failed, mdadm gave error, but no error
	appears in the web page, and it is keep blank.
	mdadm report:
	mdadm: /dev/sdb2 reports being an active member for /dev/md0, but a --re-add fails.
	mdadm: not performing --add as that would convert /dev/sdb2 in to a spare.
	mdadm: To make this a spare, use "mdadm --zero-superblock /dev/sdb2" first.

ipkg packages: when installing from /tmp, the web interface should show as
	installed "None", but it shows the header and even the remove all button 

NONEED, current gptdisk doesn't use ICU
	diskpart, gpt: the current method of copying icu libraries (*) needed by sgdisk fills
	almost all /rootmnt fs space, leaving only 2.5MB available (icudata is 18MB!), 
	One could increase the tmpfs capacity, as is done with /tmp, whenever
	swap is added. The problem is that when partitioning disks swap can be
	non-existent.
	The real problem is the 18MB icudata library!
	(*) when gpt partitioning the disk where Alt-F packages are, a copy
	of sgdisk and needed libraries is made from /Alt-F to /rootmnt.
	Currently this copy is not removed after the operation ends, as the user
	might want to repartition again and the gptdisk/icu file packages have gone.

DONE diskpart, mbrtogpt: run sgdisk -zap before, to clean traces of previous GPT data,
	or an "invalid partition" error might happens

DONE atexit: make it visible in uclibc (for 0.9.28 compatibility)
	twonkymedia at least needs it

status: when FS shrinking, in FS Maintenance does not shows all progress,
	only the shrinking one

status: when checking(?), in FS Maintenance a awk division by zero occurs

samba: the patch applied to the Makefile might fail because of parallel makes?

sysctrl: -if a disk is active, ignore the fan_off temperature.
	-add temp hysteresis as a setting in the control file,
	 but not to the web page 

FIXED samba35-doc package: there are missing dirs compared with samba-doc

DONT gptfdisk: after releasing RC1, change package to not include sgdisk, and 
	buildroot package configuration 

DONE minidlna: use kill -2 to terminate.
	Not a nice thing to do, but otherwise it takes 2 minutes to stop

ipkg: returns 0 when there is a package checksum error

FIXED fstab: hot*.sh changes it, "save settings" message appears in status page 

sysctrl: if temp is lower than fan_off_temp, the fan if not updated. If it is at a high speed,
	it will remain at that speed. Not always.

DONE minidlna: add a button to the webgui to force a rescan. Predate minidlna.conf to signal this,
	using a commented line to not fool minidlna: #rescan=yes

ipkg: the web page has ipkg_0.99.163_arm.ipk harcoded, if the package change, an old version
	will be retrieved. This is problematic, if uclibc has changed, as the old ipkg will
	not run.
	The easy solution is to not version ipkg? Better, version it, but make the base version
	a copy of the last version! This solve all problems:
	Users with a old version will see the version number increase and will update,
	new users will have the recent version and will made a (dummy) update afterwards.

packages: allow specifying ipkg feed or ffp site when first installing