by request of wbx@:

make it possible for users to do
| /etc/init.d/<foo> restart
instead of having to do
| /etc/init.d/S*<foo> restart
by moving the order into a magickal™ comment.

Untested.

Also remove the svn:executable properties, because rcS and rcK will
always call them with /bin/sh, and bump dashvers accordingly.

• const char[] is .rodata (.text on a.out) not .data
• missing dependency

heck, make them their own files

move some of the common stuff to the sqlite dir
not strictly sqlite related, but hey

apps using it must define
• -DNFOTISER to get the extra externs (strlcpy, etc.)
• -DNFOTISER_foo to get your own externs (e.g. NFOTISER_CREATE
  for these from categories.c and dbscheme.S)
  XXX this might better be in a directory-local one

improve debugging

unbreak -ddd mode

put the “all” target at the top here, too

no we do NOT want this threadsafe, damnit!

crazy warning flags indeed help to spot errors ☺

coerce this into compiling into a demo app

oops

steal a sweet thing from mksh

• add the database scheme as a string to the source
• data.h -> extern.h, it'll contain prototypes of functions etc. too
• add protos for fgetln(3), strlcat(3), strlcpy(3) which are missing
  on glibc and thusly added to the build (from ../paxmirabilis/)

categories for the menu – sorted

first draft of the actual database scheme

runs through without syntax errors; feel free to point out
any semantical errors or other improvements to me, as I know
barely enough SQL for a select statement and can't even do
insert without having the docs in front of me…

reading .nfo files and writing out the generated stuff can
be done by different programmes, ok wbx@

• wbx@ wants an URL(s) field
• we need a category name->description relationship
• some thoughts about additional input

we need a field for additional “dependencies on a CSYM value”,
like these generated for info.mk in trunk

sort subdir list

build an SQLite3 static library for use by the nfotiser and other programmes

• don't rm info.mk on clean
• don't split the config at the moment (devel only)
  later we'll split again, but not generate info.mk from that, only .cfg/*

remove the svn:ignore stuff from the packages, the info.mk files
generated by nfotiser will be kept in the repository

add a dummy, so that make prereq can succeed

WE WILL *NOT*, I REPEAT, N̲O̲T̲, USE PARALLEL MAKES BY DEFAULT!

Import the current SQLite Version 3 amalgamation
from http://www.sqlite.org/sqlite-source-3_3_17.zip

Licence: Public Domain

the order preserving stuff

• CATEGORY moves from SOURCES to CSYMS; DIRECTORY is copied there
  this is necessary so that the non-package (pure conf) symbols
  show up in the right place (category) and output file (directory)

• some of the relationships drawn are pure text, no RDBMS,
  e.g. the dependencies are just written out to the target
  files ⇒ we don't need the intermediate tables for m:n
• the primary key for SOURCES is of course the name of the
  directory where the info.nfo file lives in, not the name
  of the primary package

about reading: I first thought of using fts(3), but then,
that doesn't exist in glibc (fwcf has similar problems and
just provides an alternative implementation which can be
used on glibc based OSes); I'll probably either popen a
“find package -name info.nfo | sort” command or require
the list of info.nfo files on standard input

first draft of a possible database design

now if anyone has an idea how to preserve the order of con-
figuration menu entries WITHIN THE FILE, please tell me so…

of course, only flavours can have build deps

add some thoughts

-> SOURCES, RELEASE, SUBPACKAGES renamed
-> clarify which items MULTIPKGS can have
-> clarify which items FLAVOURS can have
-> new item CONFSYMS
-> clarify which items CONFSYMS can have

more example info.nfo conversion

add a working copy of the draft
• new: PERMIT_BIN_<x> – required for multipkgs,
  optional (defaults to ${PERMIT_BIN}) for flavours (in case we
  have flavours like idea/no_idea)
• more thoughts on the database issues

this is where we want to end (comments stripped, beware!)

• add CSYMDEPS for explicit dependency on config symbols
• rename MENUDEP to MENUDEPS for clarity

oops @base-files-arch forgot to svn add

• base-files: oops, wrong whitespace
• base-files-arch: sample .nfo conversion

this is how a conversion to info.nfo could look like
(except the /bin/sh selection stuff)

changes to the proposal:
• PERMIT_SRC      <error>         could be n, but we want to catch it
• PERMIT_BIN      <error>
-> while we're converting, we can check the licences
-> allowed value: 'y' or a reason why not
• CONFFILES       <optional>
-> just to get rid of more files
• CATEGORY        misc
-> to autogenerate package/Config.in, as discussed with wbx@

why is that here?

merge trunk (rev 2330) into adk branch, requested by wbx@

revert, thx dirk@

merging from trunk 1493:1683

remove package, peerguardian seems not maintained upstream, it uses libstdc++ gnu extensions, pg2 is windows only, enough reasons to drop, okay @no-1

remove package, peerguardian seems not maintained upstream, it uses libstdc++ gnu extensions, pg2 is windows only, enough reasons to drop, okay @no-1

remove superflous touch

convert to broken, to be fixed by wbx@

fix build

convert to new style

fix some symbols

convert to broken, checkin for wbx@ to check

second part of conversion, patch is not needed

convert to new style

fix build

madwifi tools are multi-binary, use 

select madwifi

convert to new style

convert to broken - checkin for wbx@ to check

convert to new style

convert to new style

convert to new style

convert to new style

convert to new style

corrected my conversion

convert to new style

convert to new style

convert to new style

convert to new style

convert to new style

convert to new style

convert to new style

older gnu make do not like else ifeq in one line. add missing endif

convert to new style, add new link to tar archive

convert to new style

convert to new style

convert to new style, fix BSD compile

* '%' as line addressing is a BSD extension to the text editor,
  which GNU ed 0.3+ doesn't support, but isn't listed in the
  ChangeLog relative to GNU ed 0.2 - use ',' instead
  bug cought by n0-1@
* correct line addressing of "TOPDIR:=..." line in prereq.mk

convert to new style

convert to new style

try to convert to new style, need to check on debian/stable

fix my last commit

convert to new style

convert to new style

simplify minimal

convert to new style

convert to new style

convert to new style, remove mini version, we need a better solution, i don't like the -mini packages

fix build under BSD

introduce TCXXFLAGS

convert to new style

upgrade; another case of upstream simplification

convert to new style

convert to new style

corrected my own workaround

convert to new style

fixing glib

Copying files directly into ${STAGIN_DIR} is dangerous,
because the content can vary a lot. This case here worked
when running the second time, but for a clean install it
doesn't. It simply creates a file called 'glib-1.2' instead
of copying glibconfig-sysdefs.h into it. Installing into
${WRKINST} instead should be a sane solution, as it's content
should be very predictable.

fix openobex and convert obexftp

convert to new style

because we patch with -p1 not -p0, if upstream versions
(reads: ${WRKDIST} with ${WRKDIR}/ stripped up front)
changed, the first component of the pathname changed,
leading to a totally new diff being created, not keeping
comments from the old diff. this sucks.

because it would be impractical to change _all_ diffs
to -p0 since some of them are maintained outside, even
if that would be the TECHNICALLY CORRECT solution, and
to not offend a certain co-developer, kludge the upda-
te-patches script to accept _any_ WRKSRC leading com-
ponent instead of only packagename-version and hope it
doesn't go wrong.

tested on mksh, results pleasing

upgrade to new upstream snapshot; saves 80 bytes relative to
the uncommitted 20070611 snapshot (more relative to the last
committed version) and simplifies Makefile a lot

convert to new style

convert to new style

convert to new style, dependency fixup

convert to new style

convert to new style

convert to new style

convert to new style

convert to new style, id3lib have some compile problems at least on bsd

convert to new style

also look at the massive ppp changes, I think this
is quite a good solution for creating the sub-packages

convert to new style

* add -fwrapv to the default target cflags, we don't use gcc <3.3 which
  is the first one to support it (atm we use gcc 3.4), this is required,
  see http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30475#c2

Note: this still does not eliminate all problems in gcc3, see
      http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30477 for details

* also add -fwrapv to the default *build* cflags, people whose build
  OSes have gcc <3.3 or non-gcc are SOL unless they predefine CFLAGS.
  http://gcc.gnu.org/ml/gcc/2006-12/msg00749.html implies that gcc it-
  self with -On (n>0) or -fno-wrapv miscompiles itself(!), and in com-
  bination with gcc PR#30477 this means that any gcc2 or gcc3 on the
  *build* os potentially miscompiles gcc for the target too.

complete removal of unmaintained packages in r1600

fix compile

remove unmaintained packages

convert to new style

convert to new style

* make cgi.h available for the scripts, so it is installed to ${STAGIN_DIR}
  (first tried this as bugfix, but could be useful if code relies on it)
* rrdtool converted to new style, but still doesn't compile. There are
  problems with cgilib, which I couldn't fix.

convert to new style

add back required CPPFLAGS which wbx@ took away in r1572

possibly fix the bug described by n0-1@ in
Message-ID: <20070113230904.GB24931@nuty.aurisp.office>
by changing -I to -isystem (which is only searched if
there is no matching file in -I arguments, kind of like
the default /usr/include which comes last)

note this could probably be fixed by using --with-sysroot
on gcc compiles as well

this idea from MirPorts Framework, as usual. BUT there is
one missing thing here: -L${STAGING_DIR}/usr/lib is also
added to ldflags, we need something like
-Wl,--library-after,${STAGING_DIR}/usr/lib instead (the
-Lfoo form is expanded to -Wl,--library,foo by gcc front
end anyway), which has been done in MirPorts with a wrapper
around ld… I could commit that as well if that's desired

big bang whitespace cleanup かめはめ-は:
* convert stuff like
  FOO := bar
  to
  FOO:=<tab(s)>bar
* no whitespace at end of line!
* eight subsequent spaces are a tab
* space then tab doesn't make sense

I suspect there are a few vi(m) users here. Please change
your editor if you can't get it under control; vim users
are most likely to introduce this kind of nonsense.

Consider using jupp (package/joe, for the target), especially
since version joe-3.1jupp7 and up have a mode where whitespace
is made visible (as utf-8 characters · (space) and → (tab), in
an unicode terminal).

* revert unwanted change that crept in during last commit
  (local modification - you see how new upstream features
  are making fwrt makefiles even smaller ;) especially if
  upstream starts supporting cross compiles proper…)
* fix style: $(command $(var)) -> $(command ${var})

fix the libtool problem, tested with mksh (no .la file),
libgpg-error (creates .la file), libgcrypt (uses libgpg-error)

introduce ${WRKDIST} from MirPorts, relieve ${WRKSRC} from its jobs
agreed wbx@ "if it solves our problem" n0-1@

info dir triggers a problem in postgresql, we do not use info pages anyway, so remove the default for now

convert to new style

convert to new style

convert to new style

warn if files are installed into /lib
-> this is obsolete, every package should use
   /usr/lib for it's installation

convert to new style

convert to new style

convert to new style

convert to new style, added a new feature

if the tarbal-internal structure doesn't fit, one can set ${WRKSRC}
to something different than ${WRKDIR}/${PKG_NAME}-${PKG_VERSION}.
By contrast, one can override ${WRKBUILD} to specify the directory
inside which building is done (configure, make, fake).
My new feature let's one override ${CONFIGURE_BIN} to specify the
absolute path to the configure script.
(This is/was needed/tested by/with the samba package.)

convert to new style

convert to new style, fix mysql package

remove strange include path, seems not generic

convert to new style, remove some whitespace

convert to new style

convert to new style, finetune bluez-libs

convert to new style

convert to new style, bitlbee does not use autotools

convert to new style

convert to new style

update to new style

convert to new style

convert to new style

convert to newstyle

convert to new style

convert to new style

add libiconv dependency

convert to new style

convert to new style

convert to new style

still problems with the following packages:
mysql: some strange error, seems to have sth. to do
       with CPPFLAGS
tntnet: doesn't build in trunk, still doesn't build :)
cxxtools: did build _right after_ building uclibc++
          I tried before, after building some other stuff
	  and it complained about sys/dl.h would be present
	  but not compileable.
	  -> I think there are still problems with headers/
	     libs in $(STAGING_DIR), this could get a major
	     problem soon if there are many conflicts like
	     this one

convert to new style

convert to new style, upgrade to latest upstream

convert to new style

convert to new style

convert to new style

convert to new style

convert to new style

add workaround from tg for the libtool problems, works fine so far

fixing this due to tg's hint

convert to new style

convert to new style

also did some tuning to the makefile patch of vncrepeater,
so all necessary variables can be passed via environment

converted to new style 

sync and sort list of developers

convert some more, update to latest upstream. some libtool problems unsolved, sameas in avahi package

convert to new style

convert to new style

convert to new style, still not perfect, can we abstract this tg?

update to new style, update to latest upstream

convert libaudiofile to new style

convert libart to new style

new version of wrt-radauth, updated only for our
patch, so no need for it anymore

convert to new style

explicit dependencies on directories are for the weak,
and be consistent with $ vs $$ in a gmake template

move package: target to pkg-bottom.mk,
because it depends on ${ALL_IPKGS}, and
dependencies are - in GNU make - immediately
expanded, and the package template is not yet
called here. Also move other stuff which re-
fers to ${ALL_IPKGS} etc. to pkg-bottom.mk,
and remove dependency of package on the
fake cookie because the IPKGs already depend
on it.

add PKG_CONFIG PATH to configure envs

convert libao to new style

convert to new style, update to latest upstream

convert to new style

fix the post-package md5 scan round

typo

add HOST_CFLAGS global for later use by packages that need it

*experimental* stuff to check
- all IPKGs for a specific package
- all files in the target root filesystem
for md5 collisions and report them, so they can be fixed.

enhancements:
* check the entire contents of bin/packages/ too
  (cross-package checks) -> needs persistent cache file
* ignore 0-byte-file MD5

move most magic for the "install-prepare" target into a
common makefile include; this also fixes permissions for
/tmp on brcm boards (were 0777, need to be 1777) as nice
side effect, probably has been a typo

converting to new style

the user may override CFLAGS and TARGET_CFLAGS (e.g. in prefs.mk),
but add -fhonour-copts to it (new gcc option) and provide defaults

fix typo

* add licence to fgetln.c from hector Garcia <hector@debian.org>
* sync strlfun.c and update-patches with latest miros licence template
* replace older miros licence template on the other files with
  easier freewrt © notice because they're not that "deep"

converting to new style

rename .defconfig into Config.default, so that .config* stuff is deletable

relatively automatic (some conflicts) merge of trunk 1308:1492 into branch

correct spelling

add base-files-arch to the list of actually configured packages

convert to new style

convert to new style

correct spelling (some weird dialect from overseas -> english)

add CC/CXX to configure/make env
use '...' not "..." if we don't use shell vars

convert to new style

convert

convert to new style

convert to new style, still not ready, libtools freaks me out

fix perms on initial random seed delivered with images

unbreak full build

convert to package template; work around changing package name

by request of wbx@ collect some common kernel flags
XXX to be amended later, even variable names are
XXX subject to changes, etc.

ignore info.mk properly

convert to new style

convert to new style

convert to new style

not needed either

not needed

convert to new style

convert to new style

converted to new style

upgrade to devel version, which has utf-8 support,
bug fixes, and is >2K smaller

converted wx200d and wrt-radauth

wrt-radauth needs patching, should be sent upstream

unbreak

convert autossh

prevent installing multiple times

unbreak 'make v'

convert atftp

update to new upstream, convert to new style

unbreak

unbreak update-patches

openssl converted

convert arpd

start with arptables

convert libpcap

CFLAGS et al. shall not be set as make arguments,
but as environment variables - if they're not taken,
fix the upstream makefiles

convert and update

-L${STAGING_DIR}/{include,lib} are of importance as well

fix nvram, requested by wbx@

handle cppflags and ldflags the same as cflags

agreed n0-1@

fixed wrong stampfile creation

convert libeevent, simplify libdnet after n0-1 commit

added extensible CPP- and LDFLAGS

update and convert libncurses: rewritten from scratch ;)

more conversions

move cookies to the dirs where they belong

corrected cp syntax, the .so file is not needed

fix properties (ignore info.mk) here

move more patch stuff into patch-kernel.sh to simplify and unbreak

better way of not matching *.orig patches,
also unbreaks packages with no patch files :$

if a ${TOPDIR}/prefs.mk exists, use it

converted zlib makefile to new style

if WRKDIR_BSD=1 in environment, make it bsd-style
and do the WRKDIR in the current (package) directory,
instead in TOPDIR/build_mipsel

:D

* unbreak patching: typo
* prevent warning

allow to say MD5SUM=NO to *temporarily* add new packages

convert libdnet

prevent patches/*.orig to be picked up (I hope)

_also_ copy over stuff from ${WRKINST} to ${STAGING_DIR},
because stuff like <ncurses.h> won't be found on the target...

(this is a little new for me, not having includes together
with libs... that's why... sorry)

don't copy usr/{share,man,info} because it's unneeded

simplify

because stuff like ncurses needs 'install-dev' and 'uninstall-dev',
files within ${IDIR_foo}/{lib,usr} will be copied to ${STAGING_DIR}
on package creation now, and removed on package cleaning.

This means: it's done _automatically_ and for _every_ package,
so there's no need to explicitly add these now any more.

move files from stagingdir/usr/bin to stagingdir/bin
to avoid conflicts with packages

agreed wbx@

to add cflags, use
| TCFLAGS+=		-foo
in the package Makefile

requested by n0-1@

convert aircrack and aiccu

* mention this contains (an ever increasing fraction of) material
  derived from the MirPorts Framework by The MirOS Project
* mention years of copyright

agreed wbx/n0-1

* little changes to openntpd/Makefile 
  (we use curly brackets for variables
   and round brackets for make-internal functions 
   (e.g. $(eval ...)))
* first own Makefile conversion (radvd)

add XAKE_FLAGS, document the new vars

wbx@ needs MAKE_ENV, MAKE_FLAGS, FAKE_FLAGS

use new NO_DISTFILE

convert 6tunnel

document the ${WRKfoo} variables, requested by n0-1@

make ${WRKBUILD} when it's needed (i.e. between
the patch and the configure stage) - didn't break
until now because it defaults to ${WRKSRC}, but
what the hey? we'll need it soon enough

add a few comments, derived from explaining the new
structure to wbx@ and n0-1@

implement a means for packages with no distfiles to work
instructions: use a do-extract: target (default: empty),
and set NO_DISTFILES to 1 before including package.mk

convert some basic packages

fixed little typo

add missing file

* split libgcc and uclibc from base-files
* try to use the new layout for base-files and dropbear

@tg: How to use the new stuff for packages without DISTFILE?
DISTFILE:= seems to be not enough...

clean, cleandir, and distclean should remove .cfg/ and the
info.mk (and other auto-generated files) in their entirety

* split-cfg.sh:
  - add explanatory comment about slowness
  - add progress meter
  - enable the dependency code
  - add missing -l option to fgrep(1)
* property changes to accomodate new info.mk generated file;
  nuke superfluous antique .cvsignore too
* package.mk:
  - switch 'all' from 'build' to 'build-all-ipkgs', ok n0-1@
  - move includes futher below, so that all variables that
    could be of use are defined before... stupid GNU make!
  - include the new "info.mk" from ${.CURDIR} if exists
  - add some kind of ipkgs cookie and fix the rebuild-if-needed
    logic wbx@, n0-1@ and I talked about on saturday

impact: if a .ipk exists, it's cleaned if the .config options
that are mentioned in the package's Makefile and (new, relative
to what we agreed on on saturday) Config.in are changed; if you
interrupt a package build (^C) you can continue; if you're de-
veloping on a package and break it, you must make package=foo clean
it manually (the new stuff only affects if the .ipk already
exists but its "dependencies" change); this is consistent with
preexisting behaviour

fix both an ipkg control file corruption and a dependency bug

add (commented out) code for dependency-on-.config information
generation - too slow at the moment though

absolutise path

oops, need target stubs

we're testing:
* first: ifeq (${FOO_STYLE},+foo+) and
  ifneq ($(filter foo bar baz,${FOO_STYLE}),)
* then: ifeq ($(filter-out manual,${FOO_STYLE}),)
* then: else error

to be consistent, configure build and install style default to manual,
and build and install have to be set to 'auto' for the automatisms

* fix CONFIGURE_STYLE
* make BUILD_STYLE=auto equivalent to till-now empty
* introduce concept of INSTALL_STYLE

convert openvpn; introduce pre/post-install
make ipkg control files available to *-install targets (make fake)

remove *.ipk generation instructions from the Makefiles for
the packages, move them into the template

use do/post-install stuff if you need that

introduce CONFIGURE_STYLE
* empty or manual: use do-configure target (which defaults to no action)
* gnu: call ${WRKSRC}/configure with some default opts
  (this might need tuning)
* future extension: simple, autoconf, automake, autogen (like MirPorts)

also introduces
* pre-configure, post-configure targets
* CONFIGURE_ENV, CONFIGURE_ARGS variables

add new variable BUILD_STYLE, for now possible values:
* empty (default): build by cd'ing to ${WRKBUILD}
  and issuing gmake -f ${MAKE_FILE} ${ALL_TARGET},
  defaulting to Makefile and all, respectively
* "manual": consider do-build target

also introduce {pre,post}-build hooks

unify naming of WRKINST and IDIR_*

(machinally) flesh out some stuff into pkg-bottom.mk

* convert openntpd
* get rid of WRKINST (it'll probably now only be
  the IDIR_* root) and simplify the .ipk target,
  so that it can be moved into the template

cool side effect: make package=openntpd V=99 package
works now even though I didn't select it in menuconfig

convert the first package :) to the new style, and fix it

restore default target behaviour of 'all'

warn if we encounter (possible) breakage

being proactive rules :)

tg@flyingfish:~/adk $ make oldconfig
make: *** No rule to make target dconfig'.  Stop.
2|tg@flyingfish:~/adk $ make config
package/ltp/Config.in:1:warning: multi-line strings not supported
#
# using defaults found in .config
[...]

first draft of restructuring build process to use new targets,
as discussed on saturday and requested by wbx@

stupid: gmake doesn't allow us to test if a target (rule) is
already defined... so I'll use a different approach... but not now

simplify 'clean' target

sync BSDmakefile w/ GNUmakefile changes
add some requirement comments

add DEVELOPER=1 to the command line for recursive makes
for 'make package=foo', 'make subdir=bar/foo' -> allows
us to create .ipk files from here

a variable ALL_IPKGS containing full dependable pathnames
for all .ipk files created by the port

while here, we can _easily_ check if someone uses the same
Config.in option twice (coding error - consistency check)
doesn't hurt IMHO

the .cfg/* contents depend on the split engine

handle the case of "old .config contains FOO, new doesn't"

implement splitting of .config into .cfg/* line by line
the files are only updated if their content changes, to
preserve mtime, except FWRT_HAVE_DOT_CONFIG which is magic

this will be used to forcefully rebuild packages during
an automatic build if their configuration options change

original idea by n0-1@, discussed on saturday

remove validation of the rule that this file must be
usable by non-GNU makes

this commit is more likely to break things:
* remove superfluous whitespace at EOL and EOF,
  possibly brought in due to use of vi/vim/nvi ;)
* remove all uses of other stamp files too, in
  this case: tools and toolchain

possible quirk: tools and toolchain are built every
time you start a compile now, and not only if needed;
unlike packages, they don't keep inventory of their
own status; we'll have to take them into consideration
once packages stuff is done

for now, I'd suggest a ${TOPDIR}/.tools_done and a
${TOPDIR}/.toolchain_done single stamp to be added.

remove all uses of STAMP_DIR as well as its definition
n0-1@ and I want to get rid of the old stamping stuff first

this commit, as well as almost all of the following, is
not tested - the common-adk branch is *not* supposed to
be compileable right now - but should work

Create a branch (owner set to "common" because at least
n0-1, wbx and myself will be working on it) for the ADK
development, from trunk (revision 1308)

requested by wbx@

fix descriptions

rb2.6 kernel patches cleanup

* removed all references to <linux/config.h>
* integrated my getLatchU5State ksym into 
  100-rb5xx_support.patch
* removed pcikorina.h as it is referenced nowhere
* used '-p' diff flag (makes diff's a little more
  readable)

add binutils package, we want to use it for native compiles on Routerboard, it is huge (3MB), tested on Asus Premium

remove unused code

Update iptables ipopt documentation

Add ipt_CLASSIFY target, closes #157

fix package build, recognized by wib autogenerated mails

add bash, side effect of a small workshop to show how simple integration of new packages is.

remove unneeded variable

missing restore by target snapshotconfig

first part of patches preparation

As we are currently working together with the
mtdlinux staff to get the rb500 support upstream,
this is the first part of preparing the patches
to do so. What I did was applying 'indent' on them,
using `indent -kr -i8 <filename>' (kernighan & 
ritchie, indentation level is 8 spaces).
The second part consists of some minor code fixes,
therefore it's good having these changes here committed
already.

Integrated stress into buildroot. This is another
package to generate load to the system, like cpuburn
but not written in asm, and also having support to
stress the io and vm subsystem.

rb-2.4 kernel upgrade to 2.4.34-rc1

finished the upgrade process started by Dirk with
porting of the patches and testing the result, everything
should be fine.

Arrgh, typo...

Revert 2.4.34-rc1 change for rb-2.4. I cannot regression testing all
combinations by my own and this platform is time critical. It's on Phil
to update to the new version.

Migrate to 2.4.34-rc1
* update include.patch
* drop gcc4 patch (it's now integrated)
* drop pppoe-mtu patch (integrated too)

Tested with WRT54G3G.

Finally integrated YAFFS2 support for rb532.

Downgrading to the older version taken from
OpenWrt's patch was a big mistake. No comment
about the nfs problems, they just can't have
something to do with it. The other problem I
mentioned (having 2MB reserved by the fs on each
partition) is not a bug, but a feature. It's
called "checkpointing" and helps speedup mount
time using some sort of index over the contained 
files.
As this is working flawlessly now (even flash_erase
works fine), support for NAND and YAFFS2 was added
to the kernel configs.
To get around the reserved blocks problem, I did
some rework of the YAFFS2 support to export the
default count of reserved blocks into the kernel
config (setting it to zero will use checkpointing,
but use only as much blocks as there are free in the
fs).
I also added ISDN support to the nfsroot kernel
config.

Initial commit of web image builder code.

fix annoying crash when setting interface up and down twice

My wgt survived 5 minutes doing:
while true; do ifup -v ath0 ; ifdown -v ath0;done

Patch from here http://madwifi.org/attachment/ticket/914/phydisable.2.diff

closes #133 in FreeWRT.

fix wan port initialization bug on Asus WL500g deluxe. Seems to be the only supported model, which needs this.

Major rebuild of YAFFS2 support for linux-2.6.18

Initially I used the current YAFFS2 sources from CVS,
but they included some problems:
* the filesystem used 2MB for itself on each partition
  (after formating NAND via RouterBOOT and initially mounting)
* while using a nfsroot setup for testing, some nfs timeouts
  occured right after bootup (totally weird)
this patch is based on the one by OpenWrt, and was ported
to linux-2.6.18 (more a hack, but seems to work).

Imported final NAND support for rb500

As this patch depends on the one of OpenWrt for linux-2.6.17
and the MTD interface has changed for linux-2.6.18, some rework
had to be done. Especially rbmips_hwcontrol500() was refactored
to support the new design.

first attempt to fix support - non-functional yet, but at least chip is successfully detected

Sync with changeset:1279

netio needs libpthread, included dependency handling

- added ip-up scipt to restart openntpd after a connection has been established
  otherwise the clock stays wrong if we had no WAN connection so far.

synced with changeset:1272 to run in deamon mode and script cleanup

squeezed the ipkg, resulting in 23MB smaller size

- synced with changeset:1269 

imported netio package

Imported package ltp (big, fat testsuite). As the
installed version consumes about 150MB, better not
try installing it onto your embedded device (unless
you have a CompactFlash or NFSRoot setup.

drop --passive-ftp from ipkg patch

damnit, I forgot the renameing :(

synced with changeset:1263 which upgrades to new upstream release

ipkg depends on passive-ftp, fixes #141

wget long options for ipkg

- synced with changeset:1258 from branch to vendor and product ID for Novatel XU870

added package sysstat

made installation of iptables initscript conditional,
thx to austriancoder

Remove CONFIG_DEVEL from various Config.in, whitespace cleanup

Update ppp to 2.4.4, removing patches which are very special (precompiled
filter, special default route parameter, pppoatm tuning, ppp reconnect
reimplementation). patches are now usable again.

There are some confusion about default services enabled. To calm down
discussion, reverting DHCP-Services on by default is a good idea. Revert
r1240.

synced changeset:1243 from branch. Sorry for the syncing trunkt to branch,
but it does not matter in this case

Try to revert r976

Update wpa_supplicant/hostapd package

Start dnsmasq by default on LAN when unconfigured

* fixed modules.mk for isdn on 2.6
* added isdn support to rb-2.4 config

some dot-correction

* initially included isdn4k-utils (propably needs some more fine tuning)
* some parts of capi4k-utils integration also included here (rest will follow soon)

new patch to automate logfile creation if non-existent, thx to austriancoder

fixed these wrong conversions, thx to Ulrich

Introduced a new make target `snapshotconfig'.

This target is quite equal to the well-known `allmodconfig', 
with the extension that all possible target devices are selected
at the same time. This leads to also enabling packages which depend
on certain target devices, what is useful for snapshot builds.

fix buffer problem, closes bug #142

enable tcp_wrapper

Shorewall package from Ticket #134, big thanks to Christian Gmeiner

some strange sort of typo?

DO NOT REMOVE TOTALLY UNRELATED LINES IN YOUR COMMITS
WITHOUT DOCUMENTING THAT IN THE COMMIT MESSAGE, AND
MAYBE ASK THE SUBSYSTEM MAINTAINER FIRST WHETHER THE
LINE DOES SOLVE A PURPOSE, WHICH THIS ONE INDEED DOES.

remove python dependency, ipkg-make-index.sh from OpenWrt, get rid of All package collection, use allmodconfig instead

add basic isdn hisax support

sync with stable branch

fix mtd update for netgear images

fix for ccache

sync with 1.0

owfs is broken

sync with 1.0

fix a small 2.4/2.6 issue

fix problems on OpenBSD host, reported by Ulrich Kahl

disable tntnet and cxxtools, not ready, waiting for patches from upstream

ussp-push need openobex

use /tmp as www-root, closes #146

Finished the "allmodconfig" target. As `conf' is quite stupid, with flag
"-m" it selects just everything, also every Busybox feature. This leads
to at least one broken package, so I wrote a little script replacing
config symbols by pattern, taking the default from .defconfig (which I
updated using `rm .config; make menuconfig (just save); mv .config
.defconfig`).
Also corrected a mistake regarding Fuse (sneaked back in with the last
update).

* included yaffs2 support, sources taken from
  the cvs repository at aleph1.co.uk
* ported openwrt's nand patch from 2.6.17 to 
  2.6.18.2 and also included it

thx to Aleph One and OpenWRT

no need for this dependency

fixed calculation, thx @tg :)

fixed for different content of /proc/meminfo with kernel-2.6

fixed a bug with fuse module for 2.6

Enabled the well-known "allmodconfig" target.
The idea is merely to build something that fits everywhere,
but rather for testing builds including everything. (This
way one can deselect senseless packages, like e.g. busybox
which is totally useless with everything enabled.)

Also synced Config.in.collections with the one from 1.0.

found some bugs when testing, here the fixes

* no need for preinit.arch, as it was loading modules now builtin
* took mounting of system-critical partitions into mount_root, so
  fstab can be fully user-managed
* merged mount_root in rb-2.4 - it's not target fs dependant

fixed escaping

added a little documentation to my 'tuned' KMOD_template

fixed nls dependencies - I see no need for a package depending on itself, correct me if I am wrong

fixed nls stuff for 2.6

Optimized configs a bit. Access to the shared memory
from userspace should not be necessary, so I removed it.
Also rb-reset and rb-uled were merged to a single module
which is built into the kernel, so no need for config
options for them.

19:49⎜<tntnet:#freewrt> beim bauen von cxxtools kommt: cannot find -luClibc++
19:49⎜<tntnet:#freewrt> vor cxxtools muß uClibc++ gebaut werden

merged rb-reset and rb-uled into a single module,
and integrated it into the kernel (as it's needed
always on that board)

Overriding base-files doen't work like that. Besides,
if it's universal, it should stay that. Fixed my prob
using a single line in mdev.conf which likely will grow
in future anyway.

sync with stable

fixed inittab syntax for 2.6 naming scheme

when changing something, don't forget to remove the old
code when introducing new code ;)

cought by wbx@

add ussp-push, cxxtools and tntnet (cxxtools and tntnet nonfunctional

some additions and corrections

rb-2.6 now uses mdev

unbreak tree, afdisk needs to be tested first

update uclibc++

merged files as they are the same

added ntfs support

experimental support for having CC="ccache gcc", etc.
reported by tntnet in IRC, not tested

fix url to faq

Update openswan to 2.4.7

next one, forgotten again :-/

improved configs

forgot to add the new kernel module control files

Merged modules-2.4.mk and modules-2.6.mk into modules.mk

* different module paths and names are solved with variables
 -> no if-statements around eval's allowed
 -> could help getting configuration more flexible without having
    to split every functionality into 1000 packages

* changing package content leads to new KMOD_template syntax
 -> the parameter referenced $(7) inside is now gone
    (specification of module names to echo into /etc/modules.d/$(6)-$(2)
 -> module names are now fetched from parameter $(3) 
    (the actual modules to integrate into the package)

* changing paths to iptables modules are solved inside mk/netfilter.mk
  (affects only xt_modules)
 -> the variables IPT_CONNTRACK-m, IPT_EXTRA-m, etc. now hold module 
    names including subpaths, so they have to be converted inside 
    package/iptables/Makefile

* still things to fix/reorganize, but should work for now

fixed syntax for busybox ip applet (does not understand abbreviated COMMAND params)

sync with stable branch

cleanup, use gcc builtin defines to recognize host os

fix typo for kmod-nls-utf8 package

unify Makefile header, remove obsolete afdisk headers

new target: allnoconfig - creates a config with everything disabled

Fixed for builds without busybox.
This is seldom, but not for me as i like to test 
stuff with minimal setup to reduce build time.

some 2.6 infrastructure (builds now)

jffs2 code fixes (conflicted with some earlier patch)

remade "oldconfig" for mini_fo config symbol

and here follow the patches themselfs including a (rather important)
addition to the README

added the patches READMEs (very minimalistic for now)

Initially imported afdisk (neither the Linux, nor the FreeBSD one).
Designed for automated hard disk partitioning it should fit perfect
for FreeWRT to provide a portable and flexible tool for paritioning
CompactFlash disks as target devices inside a script. 
Still needs some rework, anyway thanks alot to David Roetzel (the
author of afdisk).

Added FUSE support for kernel-2.6. This is a little 
tricky, as FUSE support is shipped with the 2.6 sources.

added mtd support to configs

* reorganized patches again (hope now finally)
* added yaffs2 support

fixed for nfsroot

corrected accidentally overwriting flash.sh at last commit

Ok, moved the handbook to the right place now.

Remove link to obsolete sip-module, common whitespace fixes again

added temporary place for the freewrt handbook (including contents)

sync with stable branch

sync with stable branch

removed old (OpenWRT?) patches, FreeWRT will start from scratch

raised version of kernel-2.6 to 2.6.18.2

improved patch layout a little

sync with 1.0

sync net-snmp with 1.0

enabled threads for iperf, needed for bidirectional speed tests
enabled multicast for iperf
added patch for a know bug when building iperf with threads
added libpthread dependency

sync with 1.0

sync with 1.0 branch, fix mini-fo bugs

update banner, development version, fwcf note

update from upstream - fixes bugs in strlcat() implementation in rare case

Revert quiet patch of 1094, choosing shell quiet

make openvpn also usable with standalone ip-package

moved dropbear startup config to /etc/rc.conf, synced double quotes from trunk

yet another dot too much

cleanup, removed dot

added dependancy for hostapd on WGT634U devices

silence busybox, added NETMAP iptables target

fix removal of *.old files. This reduces the package size by 50%.

sync with 1.0 branch (revisions 999-1085), still not complete

add patches from #135, thx christian gmeiner

merge nls stuff to trunk

Applied corrections suggested by wbx and tg. The module
stuff is just search-and-replace, there may still be
syntactical errors (e.g. different module names) which
I want to correct as kernel integration advances.
Thx for cross-reading!

initially integrated kernel-2.6 support for RouterBoard 500

keeping compat to non-broadcom systems (as discussed with wbx)

mount_root should be run before using /dev/urandom,
at least my installations don't provide a static
one. As mount_root mounts devfs it should be there 
afterwards.

some cosmetic changes

sync with 1.0 branch

sync with 1.0 branch

sync with 1.0 branch

fix owfs build, sync alsa fix in trunk

* some rather cosmetic changes to rb-reset.c
* integrated user led support

New banner font for FreeWRT

enabled tmpfs support

integrated support for reset button of RouterBoard

patch to fix libtool call, needed for ccache

ccache integration, makefile fixes

update gmp to 4.2.1, needed for ccache

update nmap to version 4.11, fix for ccache

remove debug version, fix dependency

remove rrdtool1 package

small error in help text, thanks atari

drop .orig file from patch

updated patches to fit 1.2.2.1 and upstream Config.in changes

Update all hostap/supplicant packages for madwifi

fix busybox build

bumb to new version (untested)

Sync with current version

sync with mainstream

change * to IPv4 any address

* ntpd-listener activated by default
* synchronizing is way too slow when difference <180 sec. Forcing
  immediate time update on startup

Markus don't like my style, so I give up ;-)

small fixes to openvpn.init

keep in sync with branch

drop updetach completely

commented out 'updetach' as it blocks startup on failure

add prerouting_rule for NAT

we're almost at revision 1000, adapt the banner width for that

some tuning to default busybox parameters

merged changes from target/linux/rb-2.4/nfs as they also apply here

integrated preinit into boot process

busybox upgrade to 1.2.1

added assertion on request from dirk

improved nfsroot integration

preinit now supports target filesystems without fwcf

added dependency for Broadcom-specific packages

fixing ticket #126

In accordance with wbx make squashfs-overlay default choice

corrected the variable name

delete special failsafe, default script is fine

disable ppp interface renaming: breaks defaulroute (now fixed) and ppp accounting. More testing needed.

Fixed behaviour of Package Collections to create 
soft dependencies rather than hard dependencies, 
as there still should be a way to change earlier
taken preselections.

drop --enable-largefile from makefiles. 2GB should be enough in most cases

fix build

Fix OpenVPN build

* Implementation of wifi interface without using nvram
* dropping wificonf, replacement is broadcom-wl-util
* sample config interface for 1.0

now hopefully merged with branch without breaking history :-)

I need to learn merging :( reverted to 841

I need to learn merging :( reverting to 841

missed stuff with last merge

merged with 1.0 branch

for now, do *NOT* run 'fwcf commit' automatically, EVER.
there are concurrency issues when run twice at the same time,
e.g. if the second ssh-keygen is faster than the first fwcf commit,
and then the device becomes unusable

error report in IRC by "frop"

as of now, 'fwcf commit' is only to be run manually by the admin.

merged with branch

dnsmasq update, disable SRV filter (breaks SIP), enable explicit interfaces, added ntp server

few bugfixes, should be operational now

sync with stable branch

fixed file properties

merged ppp cleanups from 1.0 branch

- target/linux/Config.in ; package/pcmcia-cs/Config.in
 -> fixed some dependency mistakes
 -> pcmcia-cs now is no longer dependant to the used hardware, 
    but to the use of the pcmcia kernel module

- target/image/rb/Makefile ; scripts/* ; docs/* 
 -> finished the nfsroot-target
 
- target/linux/rb-2.4/{ext2,ext2-cf} ; mk/targets.mk
 -> fixed this naming error

mv target/linux/board-kernel/Makefile to mk/, add nfrastructure for brcm2.4s

mtd is a hard dependency here

fix most parts of my braindead misuse of "select".

There are two methods to depend on different packages/options.

Soft dependency, user can deselect:
add to the package "foo" dependencies to device or filesystem "bar":
config FWRT_PACKAGE_FOO
 default y if FWRT_TARGET_BAR
 default n
 depends on FWRT_TARGET_BAR

If you want a hard dependency, user can't deselect, then
add a select to FWRT_TARGET_BAR
config FWRT_TARGET_BAR
 select FWRT_PACKAGE_FOO

For menus wich are only relevant for specific devices, use a
depends on line (see USB kernel modules for example).

After discussion with Phil, who said it really sucks to remove a .config before 
switching between devices or filesystems. He was right.

fix dependencies, thx martin ley

fix spelling of my name

add new developers

startup script cleanup

sync with FSF version (address change, etc.)

update to fwcf-1.01, requested some time ago by wbx@ for both branches
tested only a rebuild of the package in 1.0 branch, but the changes in
the code are minimal...

Merged back support for Mikrotik's RouterBoard532 from branches/n0-1_rb500 r849:933

remove debug version from snapshot build

remove mtd static from snapshot build, the static version is only needed if you want to update from a different firmware to FreeWRT and can be found in http://www.freewrt.org/downloads/tools

fix PR#121 and a plethora more problems with this package

make ntp_server an rc.conf variable, as asked by wbx@ in
Message-ID: <20061010233127.GB30104@freewrt.org>

XXX the default (in ntpclient.postinst) should be changed to our
XXX vendor zone, which we haven't received yet

sync with stable branch

sync with stable branch

add broadcom-wl-util as package

remove duplicate patch

cleanup watchdog init script

add to startup, bump release number

fix avahi startup script, if rc.conf variable is empty, do not start

remove duplicate binaries

add else

fix copy and paste mistake, thx cjay

openswan build fix

merge config files: request from wbx

Fix wifi scripts: now we are using /etc/network/interfaces

there is no 'pre-down' hook in ifupdown

fix kernel meta package dependency, close #110

cleaned up makefile and added hotplug script

fixed path to hotplug.d directory.

added nozomi driver for UMTS support

fixed typo, removed output policy

For 1.0: migrate openswan to ip, remove unneeded patches

remove libamsel and amwall, no upstream. cvs version. no kernel 2.6 support. okay @dirk

for 1.0: sync tcpdump/libpcap with upstream

for 1.0: cleaning up firewall script, separate config

* OpenVPN: wrong parameter order
* PPPoE: MTU propagation

apparently dead code - .config.cmd does nothing
might fix PR #118 and open different can of worms

changed ifconfig to ip

add mmc driver for linksys wrt54g mmc hack, contibuted by ulmen, thx

upstream is no longer available

add a default timezone, patch from dirk

* all: fix whitespace
* Config.in.runtime: fix english spelling, etc.
* Config.in.runtime: move the configure option
  FWRT_RUNTIME_SSH_DISABLE_PASSWORD_FOR_ROOT
  to the top so it is selected by default, instead
  of an unsafe default

- added a selection to edit ssh login configuration
- added evaluation of the selection to the makefile

remove part of the linux-mips patch, needed to get ext3 loaded, fixes ticket #109

patch from ticket #114, fixes some gcc 4.1 issues, thx

add openvpn patch from dirk, we are using ip for network configuration, add an example configuration

rename MODEL to DEVICE to keep in sync with database naming and docs

fix bug from ticket #112, thx for the report

ipsec-tools is 2.6 only, deactivate for snapshot buld for now

sync with stable

fix startup script, report from dirk

add PPPoE support to new network scheme

fix, hopefully last time

fix hostname, does not work with ash, thx AndyCap for reporting

sync with 1.0 branch

fix wlan example, it is eth2 on premium, the strange eth0/eth1 problem for b44 driver still exist, no need for up wifi

remove autobuild and snapshot targets. snapshots and releases will be done with ib.sh

add missing ipkg control file

fix initscript for openssh

sync with freewrt_1_0 branch. It was a little bit to early to branch,
but now everybody can work again. New smaller stuff can be committed 
directly in trunk.
Bigger stuff, like a new architecture or a new board should be
developed via a separate branch. Kernel and compiler updates, too.

Any bugfixes or security updates should go into branch and trunk. 

Still send me a patch first and do not commit directly into the branch.

eJunky requests feodora core 4 to be allowed

try: disable check for pkg-config

dnehring@ please test if this fixes your problem

default to FWRT_NO_PACKAGE_SELECTION according to wbx' policy
to keep the default to as few as possible

(rationale: the others select packages that can't be deselected)

more cleanups

revert r794, cf. Message-ID: <Pine.BSM.4.64L.0609302348160.11130@odem.66h.42h.de>

fixes lzo2 compile errors

fix squid ssl build

reflect renaming in file

more renameing

first cleanups

experimental: clean up stamp files
requested by eJunky, NOT to end up in 1.0 branch

fixed ARCH to KARCH for package creation

added alias ll="ls -l"

fix for broken rintf() function

- added x-wrt stuff to a seperate package
- merged some files with webif package files

TODO:
- cleanup the webif2 pages
- remove every use of nvram from this damn package...

sync with latest ipp2p patches

sync iptables with upstream version

switch package selections to openntpd as suggested by tg

VPN package selection added

update openswan to 2.4.6 (2.4.5 is unstable with many connections)
sync lzo to upstream version 
sync openvpn to upstream version, needs testing

yet another bug *sigh*

another little bug in the debian script

package selection from 1.0 branch
sync iproute2 with upstream
disable IPV6 in busybox by default

added ppp example

fix typo, name of the file is rc.conf

added example interfaces for asus without wlan bridging

two little matching bugs in vlan hook, and bad grammer in bridge hook

added loopback interface

added busybox feature to recognize 'manual' in interfaces config

added missing rc.d evaluation

sync gcc4 patches with kernel

even better: put fwcf's output to logger
so it can be seen with logread

like I announced in an eMail, re-run the init script even
on 'fwcf commit' failure, so that admin can still login

cf. Message-ID: <Pine.BSM.4.64L.0609252212330.520@herc.66h.42h.de>

strongly suggested for 1.0 after testing!

- added two hook scripts for ifupdown to configure vlans and bridges
- added hook scripts to Makefile
- added rc.conf query to broadcom network.init script

example for interface with bridging and vlan configuration.

sync kernel with linux-mips.org

some cleanup in fuse package, add wpa_supplicant dependency to madwifi, remove util-linux menu

just use one package for shfs

please test before you commit a tree break, thanks

fwcf 1.00 with LZO1X-1 compression, without dependency
on the ~150 kB zlib

ok to switch wbx@

PS: pkg-depend.mk belongs into package/ not into mk/

strip the .comment section too, shaves some bytes off the executables
also add licence boilerplate

add back "specific packages" suffix

added basic umts startup scripts and separate package available for g3g routers

updated static hierachy numbers and added firewall page

yet another renaming

added small documentation of the menu layout

renamed webif pages to reflect their menu structure

added some webif-pages from x-wrt, updated some old ones. This ist totaly not tested yet and needs some review/rewrite

it's really difficult to quote correctly, isn't it?

clean up .config for new kernel

mv kernel version to 2.4.33.3

add fwcf commit after ssh key generation, automatically save ssh hostkeys on fwcf partition

add some missing stuff from last commits, add asus wl500g premium files for non-nvram based configuration

add function delete fwcf on update

better remove some more packages from list, so that no essential packages are removed accidentally via web interface

make broadcom-scripts default for all models, but not netgear wgt634u, add some depends lines for broadcom specific packages

* add PKG_VERSION to every package in default selection
* add broadcom-scripts package as default for 1.0, it adds
  network and switch configuration scripts to base
  suggested by markus to separate these from webif package
* move ed to application and disable by default, sorry tg, but
  not everybody needs it
* be sure that if broadcom-scripts are installed, new ifup/ifdown
  network configuration is not used

add /etc/network/interfaces for wrt54gs 1.0

remove some fprintf

removed mk/targets.mk include... I copied a wrong template

added empty hotplug directory for button event

remove verbose output

remove boot_wait activation, asus wl500g has recovery mode and boot_wait always on

remove accidently commited stuff

add a sensible text editor, because busybox' vi is horribly
broken, even according to wbx@ (vi fanboy)...

add -fno-ident, on the host too (I wonder what host-built
files get into the target... libgcc probably)

add asus trx for webupgrade of asus wl500g/wl500gx/wl500gp, make cleandir is a good idea before compiling firmware for one of these models

my bad, I still had underscores in mind
re-order so we don't need to hardcode brcm et al.

optimise for size by default

even so, everything is much much larger than on i386-bsd,
is that the libraries, the compiler or the platform's fault?

fix the script for asus-wl500g-deluxe

add small example for addon configuration

fix wireless network configuration on netgear

* remove unneeded login script, just call telnetd in failsafe mode
* remove non-portable nvram check for failsafe variable

PR #86

diff -I gets an extended regular expression, but grep uses basic regexen
cf. http://www.mirbsd.org/man7/re_format.htm

* configure scripts are not always executable, prefix them with ${BASH}
* WRKBUILD is not always WRKSRC (e.g. for Berkeley DB, gcc)

do not build mypackage as default, save some space

* INSTALL_DATA is not the right thing if you have symlinks 

very stupid idea from me.

use up commands to configure wireless, pre-up scriüpt does not work correctly

switch to FWCF

ok to enable now wbx@
bugreports to tg@

allow -F to force to not check, for non-TRX partitions

ssh is for root only

fix squashfs-symlink, add network configuration for asus wl500g deluxe, add pre-up wireless script

we need to flash to different addresses too

fix ifup dirnames

switch to /etc/network/interfaces, if webif is not installed, add default network config for wl500g, more to add soon

fix the ipkg file, oops, forgot that one

we do determine the hostname slightly differently from the rest

determine type of board from the file name

"fix it if you can" wbx@

add a few dependencies the shell script has

add initial fwcf package, from pkg example derived

DO NOT SELECT THIS YET!

But positive critiques are welcome.

add the missing v3.1/v4 from linksys wrt54g, which are very equal to v3 and should be supported