Toolchain built for ARCv1 (read for ARC700 cores) by default has
disabled atomic ops (-mno-atomic). When we build Linux kernel for ARC770
which has LL/SC instructions and thus may handle normally atomic ops we
explicitly add "-matomic" in CFLAGS. But since user-space perf utility has
no way to extract CPU config options from Kconfig/defconfig it uses
compiler default settings.
In case of ARCv2 (read ARC HS38) atomics are enabled by default and so
perf builds perfectly fine thus reenabling perf for ARC HS38 (actually
for non-ARC700 targets).
Signed-off-by: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
arc-2016.09 is the most recent toolchain for ARC cores and
it is based on top of upstream Binutils 2.27 and GCC 6.2.1.
With updated major version of GCC we copied all GCC 6.x patches
for ARC as well as Bintils 2.27 patches.
Note that toochain sports ARCv4 ABI and so must be used
with 4.8+ Linux kernels. Even though it will build v4.4 kernel
perfectly fine on attempt to run user-space apps they won't
work with older kernel. That said previuosly sent RFC patches with
Linux kernel update are required:
[1] https://patchwork.ozlabs.org/patch/726686/
[2] https://patchwork.ozlabs.org/patch/726687/
Signed-off-by: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
binutils 2.27 checks if the target compiler supports -Wstack-
usage=262144, and also uses this setting for the host compiler. If the
host compiler is gcc < 4.9 binutils build will fail. This backports 2
commits which are fixing this problem for binutils 2.28.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Support for MIPS toolchains without FPU support was added upstream,
so remove our patch.
patches/310-mips-link-tool.patch was a backport form this version of valgrind
src/abort.c is not referenced anywhere
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Reboot of 519a199cbc
Which broke other builds.
This time, added compile flags to build only for POWERPC archs
Tested on mpc85xx, ar71xx and bcm2708.
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
Adds Google's mirror as primary source and kernel.org as fallback.
Same as commit 0d4f02dfd6
Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
This reverts commit 519a199cbc.
Multiple users have reported building failures on ARM architectures, e.g.:
In file included from defs.h:42:0,
from process.c:37:
process.c:51:22: error: 'struct user_regs' has no member named 'ARM_r0'
#define uoff(member) offsetof(struct user, member)
^
./linux/arm/userent.h:1:3: note: in expansion of macro 'uoff'
{ uoff(regs.ARM_r0), "r0" },
^
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
* Change git packages to xz
* Update mirror checksums in packages where they are used
* Change a few source tarballs to xz if available upstream
* Remove unused lines in packages we're touching, requested by jow- and blogic
* We're relying more on xz-utils so add official mirror as primary source, master site as secondary.
* Add SHA256 checksums to multiple git tarball packages
Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
We have to remove the FPU check, it will run in an endless loop on LEDE
when compile without FPU emulation support.
The second patch fixes this problem: valgrind: mmap(0x400000, 303104)
failed in UME with error 22 (Invalid argument).
valgrind still does not support mips16, build LEDE without mips16 support.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
In company networks everything except the http and https protocol is
often causes problems, because the network administrators try to block
everything else. To make it easier to use LEDE in company networks use
the https/http protocol for git access when possible.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
ARC port of GDB is not yet upstream so we need to use
sources from Synopsys GitHub repo.
Given Synopys' commitment to upstream ARC support in GDB
in the nearest future it might be simpler to add a separate
package for ARC GDB instead of patching generic GDB package.
This way once ARC GDB stuff gets uptreamed we'll only need
to remove that new "gdb-arc" package.
Note 1 very minor change in generic gdb package was done -
it now depends on !arc (while "gdb-arc" depends on "arc").
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Fixes build:
config/tc-i386.c: In function 'build_modrm_byte':
config/tc-i386.c:6143:31: error: logical not is only applied to the left hand side of comparison [-Werror=logical-not-parentheses]
&& !op.bitfield.reg64 != 1
Contains this fix:
https://sourceware.org/ml/binutils/2014-09/msg00095.html
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
SVN-Revision: 48003
This fix contains 2 parts:
- kernel 4.1: backport upstream patch "perf build: Do not fail on missing Build file"
- add NO_LZMA=1 to perf MAKE_FLAGS to disable LZMA support
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
SVN-Revision: 47338
The HOST_*FLAGS are for compiling programs which will run on the machine that is
running the build. Setting these flags is frequently required for unusual
cross-compiles.
Signed-off-by: Lawrence D'Anna <larry@elder-gods.org>
SVN-Revision: 46422
This updates valgrind to version 3.10.1 and make configure work with
kernel version > 4.0. This patch was taken from upstream valgrind.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 46302
- include upstream fixes for musl compatibility
Include <sys/stat.h> for S_I* macros
d34e00b293942b1012ddc49ed3ab379a32337611
Include <linux/ioctl.h> for _IOC_* macros
3460dc486d333231998de0f19918204aacee9ae3
strace 4.8 is broken with musl on some arch
(arm: omap,oxnas according to buildbot)
compile tested only
Signed-off-by: Dirk Neukirchen <dirkneukirchen@web.de>
SVN-Revision: 46124
fixes compile error:
(observed w. target arm64, all kernel build opts enabled)
error: call to '__open_missing_mode' declared with attribute
error: open with O_CREAT in second argument needs 3 arguments
Signed-off-by: Dirk Neukirchen <dirkneukirchen@web.de>
SVN-Revision: 45625
There are some places where there is a redundant declaration of
strlcpy() that prevents building perf otherwise.
Signed-off-by: John Szakmeister <john@szakmeister.net>
SVN-Revision: 44926
configure tries to guess pagesize by calling getconf, provided by eglibc and compiled for the target, not the host.
Signed-off-by: Nicolas Thill <nico@openwrt.org>
SVN-Revision: 44445
perf was in the oldpackages repository, but it makes more sense to have
it part of the default package set since we build it from the Linux
kernel sources.
Signed-off-by: Florian Fainelli <florian@openwrt.org>
SVN-Revision: 43418
Note, that licensing stuff is a nightmare: many packages does not clearly
state their licenses, and often multiple source files are simply copied
together - each with different licensing information in the file headers.
I tried hard to ensure, that the license information extracted into the OpenWRT's
makefiles fit the "spirit" of the packages, e.g. such small packages which
come without a dedicated source archive "inherites" the OpenWRT's own license
in my opinion.
However, I can not garantee that I always picked the correct information
and/or did not miss license information.
Signed-off-by: Michael Heimpold <mhei@heimpold.de>
SVN-Revision: 43155
Build libbfd and libopcodes as shared libraries, else they get linked
statically into every binary, increasing each binary size by ~650kB.
Reduces binutils package size to old expected values.
Fixes#17060.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 42337
