uclibc: use builtin startfiles target

Now that we've dropped 0.9.31.x we can use the startfiles /
install_startfiles targets instead of building and installing them by
hand.
Fixes internal blackfin toolchain build errors regarding crtreloc.o

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

gcc/bfin: 4.{6,7,8}.x are broken

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

uclibc: make target utils optional

These are broken for blackfin unfortunately so they're disabled.

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

uclibc: reorder list of dependencies for NPTL option

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

uclibc/bfin: only supports linuxthreads.old

[Thomas: do not reorder options, as this is a separate change]
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

gcc/bfin: doesn't support --with-cpu defaults

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

tstools: new package

[Thomas: properly indent help text, add largefile dependency, remove
unneeded libpcap dependency, remove unneeded CLEAN_CMDS, reorder build
and install commands, remove uneeded <pkg>_INSTALL_STAGING,
<pkg>_INSTALL_TARGET and <pkg>_INSTALL_IMAGES variables, move from
package/multimedia/ to package/, add license details, rename patch to
remove version number in the patch filename.]

Signed-off-by: Tzu-Jung Lee <tjlee@ambarella.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

normalize separator size to 80

Signed-off-by: Jerzy Grzegorek <jerzy.grzegorek@trzebnica.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

fix header package name

Signed-off-by: Jerzy Grzegorek <jerzy.grzegorek@trzebnica.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

squid: bump to version 3.3.8

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

dhcp: install to staging

Install dhcp to staging so other applications can use its include files
and libraries.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

bind: install to staging

Install bind to staging so other applications can use its include files
and libraries.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain/buildroot: properly handle SSP

The current SSP handling is incomplete.

First we need to build uClibc with SSP support for a complete
"experience".

Second, it doesn't hurt to add -fstack-protector-all to the
CFLAGS/CXXFLAGS since most users would expect buildroot to do this
rather than adding the flags themselves.

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

directfb-examples: install rose.png for df_dok

The directfb example program df_dok now requires rose.png to be
installed, otherwise it gives the following error:

(#) DirectFBError [dfb->CreateImageProvider( dfb, DATADIR"/rose.png", &provider )]: A general or unknown error occured

Therefore add rose.png to the list of data files if df_dok is enabled.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

uclibc: arc: fix fstat for big-endian

Fixes "Value too large for defined data type" messages caused by fstat
syscall wrapper returning -EOVERFLOW.

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

u-boot: support custom u-boot.bin name

On some platforms, the standard U-Boot binary is not called u-boot.bin,
but u-boot<something>.bin. This is for example the case on the
Cavium-provided u-boot, where the output file is called
u-boot_<boardname>.bin.

This patch adds a 'custom format' choice in the existing u-boot binary
format selection, where this name can be put. In the Cavium example, this
option could be set to:
BR2_TARGET_UBOOT_FORMAT_CUSTOM_NAME="u-boot-$(BR2_TARGET_UBOOT_BOARDNAME).bin"

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/uclibc: fix big oops from fb354d5f

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

cpanminus: bump to version 1.6109

Signed-off-by: Francois Perrad <francois.perrad@gadz.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

perl-cross: bump to version 0.7.2

Signed-off-by: Francois Perrad <francois.perrad@gadz.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/python-pyro: adding Pyro 3.14

[Peter: fix LICENSE_FILES tag, adjust upstream URL]
Signed-off-by: Fabien Marteau <fabien.marteau@armadeus.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux: bump 3.10.x stable version

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

linux-headers: bump 3.{0, 4, 9, 10}.x stable versions

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

dropwatch: new package

[Peter: fix help text & license, needs host-pkgconf]
Signed-off-by: Tzu-Jung Lee <tjlee@ambarella.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

udisks: fix typos in license info

Signed-off-by: Jerzy Grzegorek <jerzy.grzegorek@trzebnica.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Revert "dropwatch: new package"

This reverts commit 8d228d30105079a91fed16097193601713c02198.

Was not supposed to have been pushed (yet).

dropwatch: new package

Signed-off-by: Tzu-Jung Lee <tjlee@ambarella.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

trivial: use geturischeme helper function where possible

pkg-download.mk contains some helper functions to obtain subparts of URLs,
like the URI scheme. In pkg-generic.mk, there is still one opportunity to use
that helper function, instead of hardcoding it.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ext-toolchain-wrapper: fix uboot/linux with hardfp

The linux kernel and uboot specify -msoft-float in order to prevent floating
point code from being generated.  This causes a conflict when -mfloat-abi=hard
or -mfloat-abi options are specified in the wrapper. This patch removes the
-mfloat-abi option from the options generated by the wrapper only when
-msoft-float, -mhard-float or -mfloat-abi are specified by the user.

[Peter: fix !BR_FLOAT_ABI case, simplify]
Signed-off-by: Spenser Gilliland <spenser@gillilanding.com>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ext-toolchain-wrapper: fix typo on s/BR2_FPU/BR_FPU/

Signed-off-by: Spenser Gilliland <spenser@gillilanding.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

fix white spaces

Signed-off-by: Jerzy Grzegorek <jerzy.grzegorek@trzebnica.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

zlib: bump to version 1.2.8

Signed-off-by: Jerzy Grzegorek <jerzy.grzegorek@trzebnica.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ebtables: fix download url

Signed-off-by: Jerzy Grzegorek <jerzy.grzegorek@trzebnica.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

uclibc/arm: doesn't build in thumb(1) mode with threads

It runs out of registers, it fails even in official form
(COMPILE_IN_THUMB_MODE=y) so just build it in ARM mode since EABI
mandates interworking.
Tested in an arm920t board.

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

eglibc/arm: doesn't build in thumb(1) mode

It runs out of registers, so build it in ARM mode, EABI mandates
interworking so this isn't an issue.
Tested in an arm920t board.

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

gcc/arm: disable mudflap for thumb(1)

libmudflap doesn't build for a thumb(1) toolchain so disable it.

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

trivial: add space in 'bool"package"' construct in Config.in

In early buildroot, it apparently was customary to have following style in
Config.in files:
    bool"expat"

Nowadays, only two packages remain with this style: diffutils and expat.
This trivial patch lines them up with the style:
    bool "expat"

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

dhcp: add config option for delayed-ack feature of dhcp server

This patch creates a config option for the ISC DHCP server to allow passing
--enable-delayed-ack to its configure script.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

uClibc: Add uClibc patch to fix MIPS64/n64 interpreter

uClibc picks the wrong interpreter for MIPS64/n64.
This patch fixes this problem by checking the selected
MIPS ABI instead of the MIPS variant.

The patch has been committed upstream:
http://git.uclibc.org/uClibc/commit/?id=603af30d6992e94ac30a66b953264076f4f2fd71

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

bind: security bump to version 9.6-ESV-R9-P1

Fixes CVE-2013-3919

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

gst1-plugins-good: unbreak (dyn|multi)udpsink for systems without IPv6 support

Patch from upstream git.

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

uclibc, eglibc: remove REALLY_NOSTDLIB references

These were required with ARC gcc 4.4.7; ARC gcc 4.8 has been changed
to behave the same as other architectures with -nostdlib, making this
option obsolete.

Signed-off-by: Mischa Jonker <mjonker@synopsys.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

arc: gcc: Go from 4.4.7 to 4.8 for ARC

This fixes multiple issues, including:
 - 'operand out of range' errors (due to defaulting to
   medium calls instead of short calls)
 - library dependencies that were not being found

Signed-off-by: Mischa Jonker <mjonker@synopsys.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

arc: binutils: Go from 2.19-arc to 2.23.2-arc

Signed-off-by: Mischa Jonker <mjonker@synopsys.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux-firmware: add support for b43xx firmwares

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: instrument external toolchain wrapper

If BR_DEBUG_WRAPPER is set in the envirnment, dump the actual command
being exec()uted, to ease debugging issues with the wrapper.

[Peter: adjust code style and output format to be cut'n'paste compatible]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package/nss-mdns: do not overwrite user's nsswitch.conf

Instead of over-writing the nsswitch.conf file (which may already contain
user's specific configuration if it comes from a custom skeleton), just
replace the 'hosts' line.

Remove now-useless custom nsswitch.conf file.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package/nss-mdns: fix avahi socket location

In buildroot, the avahi socket is located in /var/run/avahi-daemon,
not in /usr/var/run/avahi-daemon/socket.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

defconfig/rpi: bump kernel for misc fixes

Bump for a few fixes:
  - fix for slow SDcards
  - fix for USB interrupts
  - fix for tracing functions

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

xenomai: requires thread support

Fixes
http://autobuild.buildroot.org/results/c99/c9912b2c21594350945c8a64528bb6a9b34d6292/build-end.log.

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

ti-gfx: use BR2_ARM_EABIHF

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

uclibc: uclibc-menuconfig: don't depend on gcc/binutils

With the move of uclibc to package/uclibc, the uclibc-menuconfig target now
depends on uclibc-configure, which brings in the build of
binutils/gcc-initial/mpfr/gmp, causing it to take ages with a clean tree.

Fix it by moving the basic .config mangling to the patch step, and depend
on that for menuconfig instead.

Notice that this is broken for _OVERRIDE_SRCDIR users, just like for
busybox, but as that is presumably a small subset of the buildroot
user base (and these advanced users can just run menuconfig in their
local uClibc tree), so this is considered the lesser evil of the two.

The proposed out-of-tree build series will also cause problems with
this.

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

arch/arm: add support for thumb(1) mode

[Peter: also adjust BR2_GCC_TARGET_MODE]
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

arch/arm: update VFPv2 comment to mention ARMv5

Commit 6b3a0417c4 ('arch/arm: arm926 may have VFP') forgot to update
the help text of the VFPv2 option to mention ARMv5. This commit fixes
that.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain/toolchain-external: don't use x$(...) construct or ==

With modern shells, we can simply do test using the "$(...)" = "value"
form. This commit gets rid of the x$(...) = x"value" constructs and
replaces == by =, which is the correct operator to test the equality
of two strings with the test program.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: check ARM EABI vs. EABIhf for external toolchains

Following the introduction of the support of EABIhf as a second ARM
ABI, it is important to check whether the external toolchain provided
by the user actually uses the ABI that has been selected in the
Buildroot configuration. This commit introduces such a check by
looking at the 'Tag_ABI_VFP_args' tag of the architecture-specific
section of the ELF headers. This assumes that ELF is the binary format
used on ARM, which may not be the case on ARM noMMU systems (they use
the FLAT binary format), but Buildroot doesn't have support for such
systems at the moment.

Also ensure the correct CFLAGS are passed to the cross compiler for the
test, so the correct variant is used in case the toolchain is multilib.

[Peter: mention CFLAGS change]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain/helpers: don't use the x$(...) = x"value" syntax

As noted by Yann E. Morin, the x$(...) = x"value" syntax is old and
ugly, and the easier to read "$(...)" = "value" can now be used
without problems.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

pkg-autotools.mk: also use QUIET for host configure runs

Similar to how we do it for target configure runs. For make -s invocations.

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

gcc-final: ensure TARGET_DIR/usr/lib exists before copying to it

Got dropped in the toolchain/gcc -> package/gcc conversion.

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

arch/arm: arm926 may have VFP

The VFP9-S FPU (VFPv2) is optional for ARM926EJ-S, see:
http://www.arm.com/products/processors/classic/arm9/arm926.php?tab=Specifications+
Real silicon:
http://www.nxp.com/documents/data_sheet/LPC3180.pdf

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

configs: use new EABIhf option for beaglebone_defconfig

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

arch: use tabs instead of spaces in Config.in files

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

uclibc: add Thumb2 fixes

This commit adds three patches to uClibc that are needed to make
Thumb2 support work properly:

 uclibc-0006-arm-clone-restore-stack-pointer-just-after-return-fr.patch
 uclibc-0007-arm-clone.S-Add-missing-IT-instruction-for-Thumb2.patch
 uclibc-0008-arm-move-check-for-BX-to-its-own-header.patch

The first one is a necessary dependency of the second one. Both of
those patches have already been merged upstream, after 0.9.33.2. The
third one hasn't been merged upstream yet, but it has already been
submitted a while ago by Yann E. Morin, without receiving attention
from upstream.

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

uclibc: use numbered patches for 0.9.33.2

In preparation to the addition of more patches that require a correct
ordering, rename uClibc 0.9.33.2 patches.

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

arch/arm: add support for Thumb2

Until now, we were using the default ARM instruction set, as used by
the toolchain: the 32 bits ARM instruction set for the internal
backend, and for external toolchain, whatever default was chosen when
the toolchain was generated.

This commit adds support for the Thumb2 instruction set. To do so, it:

 * provides a menuconfig choice between ARM and Thumb2. The choice is
   only shown when Thumb2 is supported, i.e on ARMv7-A CPUs.

 * passes the --with-mode={arm,thumb} option when building gcc in the
   internal backend. This tells the compiler which type of
   instructions it should generate.

 * passes the m{arm,thumb} option in the external toolchain
   wrapper. ARM and Thumb2 code can freely be mixed together, so the
   fact that the C library has been built either ARM or Thumb2 and
   that the rest of the code is built Thumb2 or ARM is not a problem.

[Peter: fix empty BR2_GCC_TARGET_MODE check]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

gcc: create symbolic link to ld-linux in EABIhf/gcc 4.7/eglibc

When using eglibc, the ld-linux loader is installed as
ld-linux-armhf.so.3, but gcc 4.7.x was not yet updated, and generates
binaries that reference ld-linux.so.3. This was fixed starting with
gcc 4.8.x. In order to be able to use gcc 4.7, we create the
appropriate symbolic link.

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

toolchain-external: update config options after EABIhf introduction

The introduction of the EABIhf ABI requires a few updates to the
configuration options for external toolchains, in order to ensure that
the user doesn't do any invalid selection. In detail:

 * The Linaro ARM toolchains now depend on BR2_ARM_EABIHF, because
   that's the ABI they use, and it is incompatible with EABI. The
   comment about the availability of Linaro toolchains is updated to
   inform users selecting EABI that they should select EABIhf if they
   want to see Linaro toolchains.

 * The Sourcery CodeBench toolchains now depend on BR2_ARM_EABI,
   because that's the ABI they use. A comment is added to inform users
   that have selected EABIhf that Sourcery CodeBench are only
   available when EABI is used.

 * The Arago toolchains now depend on BR2_ARM_EABI, because that's the
   ABI they use. The description of the ARMv7 Arago toolchain is also
   slightly improved.

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

configs: update defconfigs after VFP option changes

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

toolchain: remove the old BR2_VFP_FLOAT option

Now that we have a much better way of selecting between the various
VFP versions and capabilities, the BR2_VFP_FLOAT version no longer
makes sense. This commit gets rid of it, and adds the appropriate
Config.in.legacy code.

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

gcc: take into account ARM floating point capabilities

The ARM EABIhf support was introduced in gcc 4.6.x, so versions
earlier than this one should not be used when EABIhf is selected.

The ARM VFPv4 support was introduced in gcc 4.5.x, so versions earlier
than this one should not be used when VFPv4 is used.

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

binutils: exclude binutils versions that don't support EABIhf

The ARM EABIhf support was introduced in Binutils 2.22, so earlier
versions should not be selected when EABIhf is used.

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

arch: improve ARM floating point support and add support for EABIhf

This commit introduces the support for the EABIhf ABI, next to the
existing support we have for EABI and OABI (even though OABI support
is deprecated). EABIhf allows to improve performance of floating point
workload by using floating point registers to transfer floating point
arguments when calling functions, instead of using integer registers
to do, as is done in the 'softfp' floating point model of EABI.

In addition to this, this commit introduces a list of options for the
floating point support:
 * Software floating point
 * VFP
 * VFPv3
 * VFPv3-D16
 * VFPv4
 * VFPv4-D16

and it introduces some logic to make sure the options are only visible
when it makes sense, depending on the ARM core being selected. This is
however made complicated by the fact that certain VFP capabilities are
mandatory on some cores, but optional on some other cores. The kconfig
logic tries to achieve the following goals:

 * Hide options that are definitely not possible.

 * Use safe default values (i.e for Cortex-A5 and A7, the presence of
   the VFPv4 unit is optional, so we default on software floating
   point on these cores)..

 * Show the available possibilities, even if some of them are not
   necessarily working on a particular core (again, for the Cortex-A5
   and A7 cores, there is no way of knowing whether the particular
   variant used by the user has VFPv4 or not, so we select software
   floating point by default, but still show VFP/VFPv3/VFPv4 options).

It is worth noting that this commit doesn't add support for all
possible -mfpu= values on ARM. We haven't added support for fpa, fpe2,
fpe3, maverick (those four are only used on very old ARM cores), for
vfpv3-fp16, vfpv3-d16-fp16, vfpv3xd, vfpv3xd-fp16, neon-fp16,
vfpv4-sp-d16. They can be added quite easily if needed thanks to the
new organization of the Config.in options.

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

arch: Refactor BR2_SOFT_FLOAT into per-architecture options

As we are going to introduced a more advanced support of floating
point options for the ARM architecture, we need to adjust how the
soft-float option is handled. We replace the current hidden option
BR2_PREFER_SOFT_FLOAT option and the visible BR2_SOFT_FLOAT option by:

 * A global hidden BR2_SOFT_FLOAT option, defined in arch/Config.in,
   that tells whether the architecture-specific code is using software
   emulated floating point. This hidden option can be used throughout
   Buildroot to determine whether soft float is used or not.

 * Per-architecture visible BR2_<arch>_SOFT_FLOAT options, for the
   architecture for which it makes sense, which allows users to select
   soft float emulation when needed.

This change will allow each architecture to have a different way of
presenting its floating point capabilities.

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

arch: introduce BR2_GCC_TARGET_{FPU, FLOAT_ABI}

Buildroot already has the BR2_GCC_TARGET_{TUNE,ARCH,ABI,CPU} hidden
kconfig strings that allow per-architecture Config.in files to feed
the appropriate values of --with-{tune,arch,abi-cpu} when building
gcc, or the appropriate flags for the external toolchain wrapper.

This commit has two additional options:
BR2_GCC_TARGET_{FPU,FLOAT_ABI}, that allows to define the
--with-{fpu,float} gcc configure options for the internal backend, or
the -m{fpu,float-abi} options for the flags of the external toolchain
wrapper.

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

qt: add powervr and opengl

Use the following to test:
/usr/share/qt/examples/opengl/hellogl_es2/hellogl_es2 -qws -display powervr

Signed-off-by: Spenser Gilliland <spenser@gillilanding.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Sundareson, Prabindh <prabu@ti.com>
Cc: Peter Korsgaard <jacmet@uclibc.org>
Cc: Sinan Akpolat <sinan@linkas.com.tr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ti-gfx: add new package

adds accelerated GPU support for the OMAP35xx, AM35xx, AM37xx, DM37xx,
AM387x, DM814x, AM389x, DM816x, and AM335x platforms.

[Peter: fix .pc files install]
Signed-off-by: Spenser Gilliland <spenser@gillilanding.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Sundareson, Prabindh <prabu@ti.com>
Cc: Peter Korsgaard <jacmet@uclibc.org>
Cc: Sinan Akpolat <sinan@linkas.com.tr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

powervr: add powervr virtual package

Signed-off-by: Spenser Gilliland <spenser@gillilanding.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Sundareson, Prabindh <prabu@ti.com>
Cc: Peter Korsgaard <jacmet@uclibc.org>
Cc: Sinan Akpolat <sinan@linkas.com.tr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package/libsvg-cairo: new package

Libsvg-cairo provides the ability to render SVG content from files or
buffers. All rendering is performed using the cairo rendering library.

[Peter: license is LGPLv2.1+]
Signed-off-by: Carsten Schoenert <c.schoenert@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ed: bump to version 1.9

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

gnutls: bump to version 3.2.2

Closes bug #6374

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

nettle: bump to version 2.7.1

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

collectd: bump to version 5.3.1

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

libidn: bump to version 1.28

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

gettext: bump to version 0.18.3

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

squid: bump to version 3.3.6

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

libosip2: arc: Undefine __arc__ when building for ARC processors

libosip2 appears to have support for the "ARC standard" (Advanced
RISC Computing), which is unrelated to Synopsys DesignWare ARC
processors. The check for this involves verifying the existence of
the "__arc__" macro. Unfortunately, this macro is also present
as a predefined built-in for the ARC processors, resulting into
build failures.

[Peter: add short comment in .mk files as well]
Signed-off-by: Mischa Jonker <mjonker@synopsys.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

uclibc: arc: add patch to fix strcmp/index susv3 legacy

This fixes build errors for 'keyutils'

Signed-off-by: Mischa Jonker <mjonker@synopsys.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package/uClibc: drop ARM EABI conditions

... since we now only support EABI for ARM, we only need to
force EABI unconditionally.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package/owl-linux: drop check for EABI

... since we only support EABI now.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain/helpers: only check for an EABI toolchain

... since we only support EABI now.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

pkg-infra: dump all ARM OABI references when setting ABI

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux: remove EABI conditional

We're only EABI now, so we want to always build an EABI kernel
when we're building for ARM.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain/helpers: ARM OABI is no longer supported

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package/uclibc: ARM OABI is no longer supported

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

arch/arm: remove setting gcc's apcs-gnu ABI (aka OABI)

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

arch/arm: remove OABI option

OABI is more than legacy, it's dead.

New developments should go with EABI, since it so much better.
>From the Debian EABI page [0] :
  - floating point performance, with or without an FPU is very much faster
  - mixing soft and hardfloat code is possible
  - structure packing is not as painful as it used to be
  - a more efficient syscall convention
  - more compatibility with various tools

[0] http://wiki.debian.org/ArmEabiPort

[Thomas: keep the ABI choice, as we are going to introduce EABIhf later].
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package/uclibc: fix indentation (use tabs, not spaces)

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

gcc: use BR2_EXTRA_GCC_CONFIG_OPTIONS in gcc-initial and gcc-intermediate

When refactoring the internal toolchain backend logic, the code was
changed to pass the custom configure options given through
BR2_EXTRA_GCC_CONFIG_OPTIONS only for the gcc final pass, with the
idea that we're only interested by user customization for the final
compiler.

However, the beaglebone_defconfig was passing --with-float=hard
--with-fpu=vfpv3-d16 as BR2_EXTRA_GCC_CONFIG_OPTIONS, and since the
refactoring, it was causing build failures of the beaglebone_defconfig
(with messages saying that Busybox is built to use VFP arguments, but
libc/libm are not). This is due to the fact that the gcc intermediate,
which is used to build the C library, wasn't built to generate hard
float, while the final compiler was generating hard float.

So, we get back to the original situation where the options in
BR2_EXTRA_GCC_CONFIG_OPTIONS are passed to all of the compiler
passes. Of course, the specific case of hard float will be fixed by
following patches in this area, but the idea still remains: the three
gcc should have the same options, if those options affected the ABI of
the generated code.

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

gstreamer1/gst1-*: bump version

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

linux: bump 3.10.x stable version

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

linux-headers: bump 3.{0, 2, 4, 9, 10}.x stable versions

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