gettext: make only available if BR2_NEEDS_GETTEXT is set

This allows to make sure that an user, or a randpackageconfig, will
not select the gettext or libintl package when using an external
toolchain based on glibc/eglibc.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

ltrace: fix build failure

ltrace failed to build because of missing arguments to gcc to find the
header files. This is due to the fact that the existing ltrace.mk was
setting CC and LD at build time to incorrect values. Keeping the
values set at configure time is just the right thing to do.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libcgicc: fix package after autotools infrastructure conversion

The patch had an incorrect name, and the libtool patch was applied
while it shouldn't, and the package wasn't autoreconfed while the
patch changes some Makefile.am files.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libcap: fix build failure

On a x86-64 host, libcap tried to install things in
$(STAGING_DIR)/lib64 and $(TARGET_DIR)/lib64. Therefore, pass lib= and
prefix=, as required by the strange build system used by libcap.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Make all package using gettext rely on BR2_NEEDS_GETTEXT

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Add a BR2_NEEDS_GETTEXT option

When using an external toolchain that uses the glibc or eglibc C
libraries, compiling a separate gettext and libintl is not needed and
is even a source of confusion, causing build failures. These build
failures are due to the fact that when libintl is compiled, it
replaces the C library libintl.h by its own, which does #define
gettext libintl_gettext. Then, when packages want to use gettext,
autoconf realize that gettext is available in the C library and
therefore do not add -lintl to the LDFLAGS, causing the build failure
because the program has been compiled to use libintl_gettext but this
function is not available.

Therefore, we should only use gettext if a uClibc internal toolchain
or a uClibc external toolchain. If an external glibc toolchain is
used, gettext shouldn't be used.

In order to implement that, we introduce the BR2_NEEDS_GETTEXT option,
which is hidden to the user, and whose value is computed automatically
from the rest of the configuration.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

mtd-utils: buildsystem is broken with parallel make (regarding libubi)

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

avahi: fix build on uclibc 0.9.31 without ipv6 support

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Config: terminate version number string

Reported-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

target/linux: bump 2.6.32 and 2.6.33 versions

New versions of the 2.6.32 and 2.6.33 kernel were released today
and it is suggested that all users should upgrade.

Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

busybox: 1.16.1 fixes for dhcpd and sed

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

kernel-headers: bump 2.6.32.x / 2.6.33.x stable versions

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Config.in: past 2010.05-rc2

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

update for 2010.05-rc2

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

CHANGES: add updated packages from Thomas' build-fixes branch

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Merge branch 'build-fixes' of git://git.busybox.net/~tpetazzoni/git/buildroot

libnl: include limits.h to fix build failure

Fixes the ULONG_MAX definition problem:

In file included from cache.c:43:
../include/netlink-local.h: In function '__str2type':../include/netlink-local.h:218: error: 'ULONG_MAX' undeclared (first use in this function)

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

netplug: disable -Werror to fix build failure

netplug uses nested functions, and gcc generates a warning that
is turned into an error:

cc1: warnings being treated as errors
if_info.c: In function 'ifsm_scriptdone':
if_info.c:289: error: generating trampoline in object (requires executable stack)

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

less: depends on curses

The less package already selects BR2_PACKAGE_CURSES, but it fails to
depend on it in the makefile. Therefore, there are cases where less
gets built before curses, and the build fails:

checking for working terminal libraries... Cannot find terminal libraries - configure failed

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

hostapd: use internal crypto library

When OpenSSL is not enabled, hostapd relies on an internal
library. However, this library must be enabled, otherwise the
compilation of hostapd fails with:

/home/test/ctng-arm-eglibc/bin/../lib/gcc/arm-unknown-linux-gnueabi/4.3.4/../../../../arm-unknown-linux-gnueabi/bin/ld: cannot find -ltommath

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

xvkbd: bump version

And fix dependencies while we're at it.

Signed-off-by: Paul Jones <paul@pauljones.id.au>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

shared-mime-info: missing dependency on host-intltool

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

ltp-testsuite: mark as broken

The install step, done with fakeroot, tries to install things in /opt:

make[2]: Entering directory `/home/test/buildroot/output.ctng-arm-eglibc-2010-05-07-19-39-56/target/root/ltp-full-20090630/lib'
install -D -m 644 libltp.a //opt/ltp/lib/libltp.a
install: cannot create directory //opt/ltp: Permission denied
make[2]: *** [install] Error 1

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

sed: bump to 4.2.1 to fix build failure

Version 4.1.5 fails to build due to the infamous getline() conflict
issue. 4.2.1 doesn't have this problem.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

vlc: mark as broken

Before 2010.02, VLC didn't build, because it didn't pass the
appropriate --disable- options according to the libraries available in
Buildroot. Now, 2010.05 is going to be released, and no one bothered
to fix VLC. Therefore, let's mark VLC as broken.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

linux-fusion: add dependency on !BR2_KERNEL_none

linux-fusion is composed of a kernel module, and linux-fusion.mk uses
several variable definitions that only exist when Buildroot compiles a
kernel.

We also add the same dependency to BR2_PACKAGE_DIRECTFB_MULTI, which
relies on linux-fusion.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libxcb: relax host-python regexp

Some python versions print x.y rather than x.y.z as version number, so
make regexp more robust.

Reported-by: Paulius Zaleckas <paulius.zaleckas@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package/games/vice: remove

Has been marked broken since Nov 2009, so removing as announced.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

libgcrypt: specify where gpg-error-config is

libgcrypt depends on libgpg-error, but it needs to know where the
gpg-error-config utility is.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

hal: add dependency on host-libxml-parser-perl

hal integrates its own version of intltool tools, but they require
libxml-parser-perl to be installed on the host.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

gawk: remove CC override

For some unknown reason, gawk.mk was overriding CC to $(TARGET_CC) at
build time, while it had already been set at configure time to
"$(TARGET_CC) $(TARGET_CFLAGS)" which is already the good value.

Setting it to just $(TARGET_CC) breaks the compilation with external
toolchains because we must pass the --sysroot option. But anyway, this
was incorrect as we were loosing some CFLAGS set by Buildroot.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

gnuchess: add a patch to fix getline() conflict

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

php: make sure either CLI or CGI is selected

When neither CLI nor CGI is selected, PHP's configure script fails
with:

checking whether to build CGI binary... configure: error: No SAPIs selected.

Of couse, the help text of the options says that at least one of them
should be selected, but when doing testing with randpackageconfig,
noone is reading these help texts.

Therefore, based on the suggestion of Yann E. Morin, modify the
Config.in organization so that at least one of the option is selected.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Merge branch 'squashfs-fixes' of git://git.busybox.net/~tpetazzoni/git/buildroot

CHANGES: #1531 and #1669 are now resolved

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

CHANGES: update with changes from misc-fixes2

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Merge branch 'misc-fixes2' of git://git.busybox.net/~tpetazzoni/git/buildroot

CHANGES: update with changes from misc-fixes

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Merge branch 'misc-fixes' of git://git.busybox.net/~tpetazzoni/git/buildroot

hostapd: Fix build with external toolchain

Closes #1717

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

squashfs3: fix build when zlib isn't available on the host

We need to pass CFLAGS and LDFLAGS, and to rename the patch so that it
actually gets applied.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

squashfs: fix build when zlib isn't available on the host

Pass CFLAGS and LDFLAGS so that when building squashfs, it finds the
zlib installed in $(HOST_DIR).

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

xlib_libX11: re-add a patch to fix the keysymdef issue

When xlib_libX11 was bumped to 1.3.2 in commit
1d956c9190dafbe425e9e02255b540641e27f690, the keysymdef patch was
dropped. However, this patch is still needed in order to be able to
tell ./configure where the keysymdef header file is installed.

The patch has been updated for 1.3.2.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libxcb: fix path to Python modules

When building libxcb, the variable XCBPROTO_XCBPYTHONDIR must point to
the location where the Python modules needed to run the c_client.py
program are installed. The path
$(STAGING_DIR)/usr/lib/python2.6/site-packages was hardcoded. However,
it doesn't work when the version of Python installed on the host is
Python 2.5.

Therefore, add a little bit of magic to compute the host Python
version.

We also verify that Python is available on the host, as we don't build
it in Buildroot.

Fixes bug #1531.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

busybox: disable MTD utils in default configuration

The MTD utils require MTD headers in the toolchain, and in several
external toolchains, they are not present. In order to make the build
work by default, let's disable the MTD Busybox applets in our default
configuration.

Fixes bug #1669

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

customize: fix copy paths

Closes #849, #1135

customize.mk used to copy files with a wrong command, resulting in one
directory level in excess. For example, source/etc/myfile would have
produced /etc/etc/myfile in the target filesystem instead of /etc/myfile.

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

wpa_supplicant: Fix build with external toolchain

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Fix the computation of REAL_GNU_TARGET_NAME

When the selected C library is glibc, the C library shouldn't be
mentionned in REAL_GNU_TARGET_NAME. In other words:

 arm-unknown-linux-uclibcgnueabi must be used for uClibc
 arm-unknown-linux-gnueabi must be used for glibc

This fixes the build of GDB on the target, as reported by Quotient
Remainder <quotientvremainder@gmail.com>.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

pciutils: Fix cross compiling

Closes #1705

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

usbutils: fix recursive LIBUSB Kconfig issue and bump to 0.87

Closes #1699

[Peter: Use BR2_KERNEL_MIRROR]
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package/sed: make HOST_SED_BINARY a simply expanded variable (:=)

HOST_SED_BINARY is used 2 times during startup, and is relatively
expensive to run, so make it a simple expanded variable (:=) rather
than a recursively expanded one (=).

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

boa: use BOA_VERSION instead of spawning a shell/basename for build dir

And as it is a recursive variable, this is done twice during startup.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

check-host-sed.sh: add missing shebang line

Bash seems to be smart enough to source the file when execve returns
ENOEXEC, but other shells might not be.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Makefile: make GNU_HOST_NAME a simply expanded variable (:=)

GNU_HOST_NAME is referred more than 100 times in the tree, used 5
times during startup, and is relatively expensive to run - so make
it a simple expanded variable (:=) rather than recursively expanded
one (=).

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Makefile.package.in: optimize UPPERCASE macro

As noticed by Thomas, we call the UPPERCASE macro a lot, and it slows down
startup quite a bit.

Optimize it by implementing it in make, rather than forking a shell + tr.
The implementation is heavily based on the 'up' macro from gmsl
(http://gmsl.sf.net)

With this in place, startup time is ~5 times lower.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

xfonts: fix dependencies and build process

Instead of assuming that mkfontdir and mkfontscale are available on
the development system, use the one installed in $(HOST_DIR).

The FCCACHE variable is removed, because, for some reason, the
installation process of xfonts disables the call to fc-cache when a
DESTDIR is passed. See the definition of XORG_FONT_FCCACHE in
aclocal.m4 in font packages:

AC_DEFUN([XORG_FONT_FCCACHE],[
        AC_PATH_PROG(FCCACHE, fc-cache)
        FCCACHE_WARN='echo "** Warning: fonts.cache not built" ; echo "** Generate this file manually on host system using fc-cache"'
        if test x"$FCCACHE" = x ; then
                RUN_FCCACHE="${FCCACHE_WARN}"
        else
                RUN_FCCACHE='@(if test -z "$(DESTDIR)"; then echo $(FCCACHE) $(fontdir); $(FCCACHE) $(fontdir); else'
                RUN_FCCACHE="${RUN_FCCACHE} ${FCCACHE_WARN} ; fi)"
        fi
        AC_SUBST([RUN_FCCACHE])
])

In addition to this, we update the font dependencies to include
font-utils for the host, mkfontdir for the host, mkfontscale for the
host and bdftopcf for the host.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

xapp_mkfontdir: enable for the host

When compiling xfonts, mkfontdir is needed for the host, therefore,
enable it.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Enable xapp_bdftopcf dependencies for the host

Now that xapp_bdftopcf is enabled on the host, we also need its
dependencies to be available for the host. The dependency of
host-xapp_bdftopcf is host-xlib_libXfont, which itself depends on a
bunch of other packages. Some of them were already available for the
host, some not (xfont_encodings, xlib_xtrans, xproto_fontcacheproto,
xproto_fontsproto) and are therefore added by this patch.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

xapp_bdftopcf: enable for the host

Compiling xfonts require having bdftopcf installed on the
host. Therefore, enable host support for this package.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

xfont_font-util: enable host package

When compiling xfonts, host utilities like bdftopcf or ucs2any, part
of xfont_font-util are needed. Until now, xfont_font-util was
installing ugly symbolic links, assuming that these tools were
available on the development machine.

This patch enables the host package for xfont_font-util, removes the
post-install.sh script, and keeps only the useful part of this script
in a post install hook.

host-xfont_font-util will be used later by xfont packages.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Add staging libraries directories to -L in external case

In order to solve issues of libtool trying to link target components
against host libraries, it seems that specifying -L$(STAGING_DIR)/lib
and -L$(STAGING_DIR)/usr/lib works.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Add -rpath option for host package compilation

In c1b6242fdcf2cff7ebf09fec4cc1be58963e8427, we added
$(HOST_DIR)/usr/lib to LD_LIBRARY_PATH when building target packages,
because the build of target packages sometimes require host tools
installed in $(HOST_DIR)/usr/bin which themselves require host
libaries installed in $(HOST_DIR)/usr/lib.

Unfortunately, this solution didn't work, as libtool then tried to
link target binaries against host libraries. So $(HOST_DIR)/usr/lib
got removed from LD_LIBRARY_PATH in
0d1830b07db4ebfd14e77a258de6fb391e57e960.

However, this meant that we went back to the previous situation, in
which host tools used during compilation of target components might
require host libraries. An example :

make[2]: Entering directory `/home/test/buildroot/output.ctng-arm-eglibc-2010-05-04-12-41-00/build/xfont_font-adobe-100dpi-1.0.1'
/home/test/buildroot/output.ctng-arm-eglibc-2010-05-04-12-41-00/host/usr/bin/mkfontdir /home/test/buildroot/output.ctng-arm-eglibc-2010-05-04-12-41-00/target/usr/share/fonts/X11/100dpi
/home/test/buildroot/output.ctng-arm-eglibc-2010-05-04-12-41-00/host/usr/bin/mkfontscale: error while loading shared libraries: libfontenc.so.1: cannot open shared object file: No such file or directory

Therefore, we try another solution: make sure that host binaries are
linked with an -rpath option, so that $(HOST_DIR)/usr/lib doesn't need
to be in LD_LIBRARY_PATH for them to find their libraries.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Remove gtk-doc documentations from target

On my target rootfs it is installed documentation for:
atk, cairo, gdk, gdk-pixbuf, gio, glib, gobject, gtk, pango

It ocupies 34M of space. Configure option --disable-gtk-doc doesn't
prevent installation of it. So we should remove them manually if
documentation is disabled.

[Peter: also remove manual rm's from hal/libxml2/libglade]
Signed-off-by: Paulius Zaleckas <paulius.zaleckas@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

hal: cleanup package

Closes #1525

Kconfig selects are not recursive, so ensure we select the dependencies
of our dependencies as well.

Cleanup target installation:
- Remove tasks handled by the generic post-build cleanup
- Hal doesn't install etc/rc.d, so don't try to delete it

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

mplayer: disable parallel builds (make -jX)

The build system races with itself when building the internal vidix stuff
(x86 only), so use MAKE1.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

usb_modeswitch: Bump to version 1.1.2

And correct kconfig dependency.

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

mplayer: internal copies of libdvdread/libdvdcss require largefile support

So disable when we don't have largefile support in toolchain.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

X11 OpenChrome driver depends on libxcomposite

Signed-off-by: Paulius Zaleckas <paulius.zaleckas@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

update for 2010.05-rc1

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

CHANGES: update list of new/updated packages

Mainly package changes done by Thomas' pull requests.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ccache: fix ccache-target dependencies

Ccache for the target doesn't make sense without a toolchain, so make
it depend on it.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

bzip2: fixup build after ea448fe (convert to gentargets, add host package, bump version)

Fix a few typos/mistakes introduced in commit ea448fe:

- bzip2 soname is not the same as package version
- It is PKG_INSTALL_STAGING, not PKG_STAGING_INSTALL
- Typo: BZIP2_INSTALL_TARGET_CMDS, not B2IP_INSTALL_TARGET_CMDS

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package: Add $(HOST_DIR)/usr/lib to LD_LIBRARY_PATH for host but not target

Some packages like icu requires to be compiled against the host system
first to be able to compile against the target. This is due to the
usage of self generated binaries by the package to build itself. When
the generated tools also depends on generated libraries it is required
to add the path to these libraries in the library path
(LD_LIBRARY_PATH) especially for the configure step.

Adding $(HOST_DIR)/usr/lib to LD_LIBRARY_PATH for target compilation
might break the link step by mixing host libraries and target
binaries.

Signed-off-by: Lionel Landwerlin <llandwerlin@gmail.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

target/generic: inittab: fix devpts mounts when devtmpfs is used

If devtmpfs (the kernel-maintained /dev filesystem) is used, no
/dev/pts directory is created, causing the devpts mount to fail, which
in term causes stuff like dropbear to fail.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ncurses: fix path to patch ncurses5-config

Signed-off-by: Lionel Landwerlin <llandwerlin@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux-fusion: bump to 8.1.1

Also convert to generic package infrastructure

Signed-off-by: Lionel Landwerlin <llandwerlin@gmail.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

directfb: added directfb tests installation option

Signed-off-by: Lionel Landwerlin <llandwerlin@gmail.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

directfb: bump to 1.4.3

Signed-off-by: Lionel Landwerlin <llandwerlin@gmail.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

docs/news: slides from FOSDEM

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ncurses: convert to autotools infrastructure

Signed-off-by: Lionel Landwerlin <llandwerlin@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain/gdb: get rid of 6.4 / 6.5

These are ancient (2006) and upstream strongly discourage using them:
ftp://sourceware.org/pub/gdb/old-releases/README

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

autoconf: Unverbosify autoreconf output

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Fix autotools configuration cache handling

It got broken by the introduction of the new autotools infrastructure
based on the generic infrastructure.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Update defconfigs after BR2_GNU_TARGET_SUFFIX removal

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove BR2_GNU_TARGET_SUFFIX and compute GNU_TARGET_NAME

Instead of asking the user about the GNU target suffix, just compute
it automatically from the other configuration options.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Update defconfigs after BR2_ROOTFS_SUFFIX removal

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Get rid of the code computing ROOTFS_SUFFIX

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove last users of the ROOTFS_SUFFIX thing

The ROOTFS_SUFFIX thing has been removed in
325bfd1cbae03fe286796d3fa7de7c0a3227f7a5, so get rid of the last users.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Get rid of KERNEL_CROSS

In both internal and external toolchain cases, KERNEL_CROSS was
defined to *exactly* the same value as TARGET_CROSS. It isn't modified
anywhere, and is just used by kernel compilation and pcmcia
compilation.

Therefore, get rid of KERNEL_CROSS and use TARGET_CROSS instead.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Commonalize the definition of TOOLCHAIN_DIR

The definition of TOOLCHAIN_DIR is the same regardless of whether
external or internal toolchains are used. Moreover, move its
definition together with all the other *_DIR definitions.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

uclibc: make sure the default configuration doesn't ask questions

Since new configuration options have been added in 0.9.31, the value
of these configuration options should be determined, either by the
default configuration file we provide, or by uclibc.mk process.

The locale generation process should probably be improved in order to
allow building other locales than just en_US.

[Peter: fixup locale handling, add PROGRAM__NAME to defconfig]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

uClibc: add UCLIBC_HAS_NFTW to default configuration

UCLIBC_HAS_NFTW is a new knob in 0.9.31, which allows the obsolete and
deprecated ftw() to be compiled-out separatly from nftw(), which is
part of POSIX. nftw() should probably be enabled by default in uClibc,
and a bug has been opened about this on uClibc bug tracker
(https://bugs.busybox.net/show_bug.cgi?id=1597).

nftw() is, for example, used in Gtk+.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Paulius Zaleckas <paulius.zaleckas@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

autotools: fix referencing to LIBTOOL_PATCH variable

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

packages: remove useless HOST_*_LIBTOOL_PATCH

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

autotools: inherit LIBTOOL_PATCH for host packages

When FOO_LIBTOOL_PATCH is defined, assume that the same value should
be used for HOST_FOO_LIBTOOL_PATCH.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

openssl: convert to the generic infrastructure

OpenSSL is not using the autotools as its build system. Therefore, we
must use the generic infrastructure instead of the autotools one.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

libxslt: fix host compilation

Do not build against libgcrypt on the host, since we don't build
libgcrypt ourself, and it might fail even when libgcrypt is installed
on the host:

...configure...
checking for libgcrypt-config... /usr/bin/libgcrypt-config
Crypto extensions will be available.

...build...
gcc: /usr/lib/libgcrypt.so: No such file or directory

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

autotools: fix autoreconf check

The autoreconf check was incorrect, missing a $ sign to properly
reference a package-specific variable. There was no visible effect
until now since :

 * The existing syntax allowed to access the value defined in the
   package specific .mk file, so when AUTORECONF was set to YES by a
   package, it was working.

 * The default value in Makefile.autotools.in was NO. In fact, when a
   package .mkf file wasn't defining the AUTORECONF variable, the
   Makefile.autotools.in test was testing the empty string against
   'YES', which was false, leading to the AUTORECONF not being done,
   which was the desired effect.

However, in a later patch, we intend to change this default value.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

lua: fix pkgconfig installation

install -D needs the destination filename, not just it's directory,
otherwise we end up with a pkgconfig file instead.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain/gcc: bump 4.4.x version to 4.4.4

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>