Ed Bartosh [Thu, 20 Aug 2015 11:10:29 +0000 (14:10 +0300)]
wic: deferred call of hlp.get_wic_plugins_help()
If get_wic_plugins_help is called from wic main module
it calls git_bitbake_var at some point. This fails when
wic is called from bitbake as 'bitbake -e' can't be
run.
Moved call of this method to help.py in order to call it
later, when BitbakeVariables singleton is properly initialized
to get variables from .env files.
Signed-off-by: Ed Bartosh <ed.bartosh@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Ed Bartosh [Tue, 18 Aug 2015 15:41:33 +0000 (18:41 +0300)]
wic: remove undescore from function name
Renamed __exec_cmd -> _exec_cmd as double underscores cause
strange behaviour when function is called in class method.
Python complains that __exec_cmd method(!!!) of the same class
doesn't exist.
Signed-off-by: Ed Bartosh <ed.bartosh@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Mariano Lopez [Tue, 18 Aug 2015 14:52:38 +0000 (14:52 +0000)]
testimage: Run commands in target and host when test fails
This patch modify three files altought two of them
are minimal modifications. This version includes
the changes proposed by Paul.
testimage.bbclass:
Create new vars for easy modification of the dump
directory and commands to be run on host and target
when a test fails
TESTIMAGE_DUMP_DIR: Directory to save the dumps
testimage_dump_target: Commands to run on target
testimage_dump_host: Commands to run on host
oetest.py:
- Allow to use the vars defined in testimage class
- Now able to run commands in the host and dump the
results
- Fix an issue with the condition where to run the
dump commands (Before it run the commands every
test after a failure, now it runs the commands only
in tests that failed)
- Fix the output to stdout
Ross Burton [Wed, 26 Aug 2015 20:32:45 +0000 (21:32 +0100)]
byacc: fix compilation on Ubuntu 14.04
Ubuntu defaults to passing _FORTIFY_SOURCE=2 which breaks byacc as it doesn't
pass enough arguments to open():
inlined from 'open_tmpfile' at byacc-20150711/main.c:588:5:
/usr/include/x86_64-linux-gnu/bits/fcntl2.h:50:24: error: call to '__open_missing_mode' declared with attribute error:
open with O_CREAT in second argument needs 3 arguments
Add a mode of 0666 to fix this.
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Bruce Ashfield [Thu, 20 Aug 2015 20:24:56 +0000 (16:24 -0400)]
linux-yocto/3.14: cleanups and gcc5 ARM build fixes
Updating the 3.14 SRCREVs to match the latest kernel meta data updates
and also to merge four patches Richard Purdie located that fix the
gcc 5.x ARM build (we still have boot issues, but building is the
first step).
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
833214b33303 intel-quark-tiny: Add tiny KTYPE .scc file 45ecab18d462 amd.cfg: Add X86_AMD_PLAFORM_DEVICE to default 9a31d2cbc251 intel-quark-standard: add mass storage and boot live support 4ea43922558a bsp: add coretemp to intel-common-drivers
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Paul Eggleton [Fri, 28 Aug 2015 09:13:20 +0000 (10:13 +0100)]
lib/oe/patch: handle patches that delete files with older git
When running "git add" on a directory with older git versions, deleted
files aren't added to what is to be committed unless you explicitly
specify the -A option. The result of this is that when applying a patch
from a recipe which doesn't apply with "git am" (and we fall back to
applying the patch through other means then "git add" following by
a "git commit") these deletes weren't committed with the patch, leaving
them sitting deleted but not committed at the end.
This should fix test_devtool_modify_native (which unwittingly exercises
this scenario by attempting to run "devtool modify" on apt-native) on
machines with older git versions.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Richard Purdie [Tue, 25 Aug 2015 16:58:55 +0000 (17:58 +0100)]
os-release: Exclude DATETIME from sstate signature
I'm torn over this, I can see someone wanting this to vary in each
build. What pushed me to believe its incorrect in its current form
is that it changes for every MACHINE build, making a complete mess
if you try and generate package feeds using it.
The alternative would be to give up on allarch in this case and make
it MACHINE specific which at least would then be more consistent when
the package makes it to a package feed. If someone wants to do that,
the can propose another patch but this change at least make package
feed usage 'sane' again and avoids the perpetual rebuilds.
Incidentally, its worth noting that changes in source metadata revision
used for the build and included in this recipe would still trigger
rebuilds which is likely the common use case people actually care about.
Changes to OVERRIDES (e.g. from changing MACHINE) should not change
the sstate signatures of do_package_write_*. Exclude the variable
explicitly in the package classes to avoid this.
Richard Purdie [Tue, 25 Aug 2015 16:56:12 +0000 (17:56 +0100)]
oeqa: Optimise sstate tests
We can use "none" here instead of printdiff for some small performance
gains. This also means we can remove the ignore exit code hack which
leads to safer code.
Signed-off-by: Alexander Kanavin <alexander.kanavin@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
libatomic-ops: move to recipes-support, fix recipe name
- This recipe is useful for more than just pulseaudio, so move it to
recipes-support.
- Rename to the correct upstream name, which corresponds to the library name.
Keep a PROVIDES of libatomics-ops for compatibility.
- Depend on libatomics-ops unconditionally for all archs, as it's used by more
than just arm, and currently the configure script auto-detects and enables
its use based on the existence of its header, resulting in non-deterministic
builds.
- Drop the ldflags bits which link against libatomic_ops, because libunwind
doesn't actually link against it, nor does it need to, as it currently uses
all inline functions from the header file.
The sbindir vs bindir separation really doesn't mean much in this context, and
toolchain-scripts.bbclass doesn't add sbindir to the PATH, so set the former
to the latter.
update-rc.d.bbclass: explicitly dep on initscripts
update-rc.d doesn't just want an initscripts package, it wants
initscripts-functions. In the case where VIRTUAL-RUNTIME_initscripts is set to
nothing, update-rc.d still needs initscripts-functions to satisfy its runtime
dependencies. Further, we shouldn't be using a *runtime* virtual in a *build
time* variable. This should resolve image creation failures with
configurations without sysv initscripts when a recipe inherits update-rc.d and
doesn't inherit systemd.
image.bbclass: re-run do_rootfs when feature packages change
Currently, FEATURE_PACKAGE_<feature> isn't in any vardeps, so changing the
packages for a feature won't change the checksum for do_rootfs. Rather than
explicitly adding those to vardeps, just use the expanded form of
FEATURE_INSTALL and FEATURE_INSTALL_OPTIONAL, so the actual list of packages
from the features goes into the checksum.
Automatically spawning a devshell, rather than doing so on demand, is
questionable default behavior, and is potentially problematic in headless
builds. Further, there are problems with the patch resolver today. Default to
noop, and the user can always opt-in to use of the patch resolver by setting
PATCHRESOLVE to user.
The user's shell shouldn't be allowed to affect the build (and it can break
the flock command and possibly more, if the user's shell isn't POSIX
compliant).
where it now pokes at the system to find
out the path of mount/umount tools, this caused the builds to fail
because it ended up with identifying these programs from native sysroot
but they were actually meant for target, this lead to boot failures due
to none of mount worked because the paths are encoded into systemd
binaries during build time.
Correct few others while here, these are not yet detected wrongly in my
build, because those binaries are not found in my native sysroot but if
some one staged the native providers of these packages they will fail too.
Khem Raj [Sat, 15 Aug 2015 22:02:16 +0000 (15:02 -0700)]
core-image-minimal: Add 4M extra space when using systemd
qemu images fail to rebuild the hwdb due to no space letf on device
systemd-hwdb-update.service - Rebuild Hardware Database
Loaded: loaded (/lib/systemd/system/systemd-hwdb-update.service;
static; vendor preset: enabled)
Active: failed (Result: exit-code) since Sat 2015-08-15 21:39:28
UTC; 28s ago
Docs: man:hwdb(7)
man:systemd-hwdb(8)
Process: 92 ExecStart=/bin/systemd-hwdb update
(code=exited, status=1/FAILURE)
Main PID: 92 (code=exited, status=1/FAILURE)
Aug 15 21:39:28 qemux86 systemd-hwdb[92]:
Failure writing database //etc/udev/hwdb.bin:
No space left on device
Aug 15 21:39:28 qemux86 systemd[1]:
systemd-hwdb-update.service: Main process
exited, code=exited, status=1/FAILURE
Aug 15 21:39:28 qemux86 systemd[1]: Failed to
start Rebuild Hardware Database.
Khem Raj [Mon, 24 Aug 2015 01:31:20 +0000 (18:31 -0700)]
diffstat: Add TOOLCHAIN_OPTIONS to Linker calls
configure in this case does a novel thing where it drops all additional
compiler options from CC variable into CFLAGS which in OE's case include
the options added by TOOLCHAIN_OPTIONS and then it does exactly same for
LD as well, the problem starts when we miss the --sysroot options and
ABI options which are part of CC variables usually in OE, in the end it
assigns LD = CC but doesnt add newly separated CFLAGS to LDFLAGS hence
the compile still works since all those needed options are getting into
compiler cmdline but link step fails especially when using gold, since
it does not find --sysroot option on cmdline it starts to complain about
missing libraries and paths
This shows up with clang more so because clang does not have implicit
sysroot that it silently passes down to linker like gcc does when
confgured with sysroot
We see errors like
arm-angstrom-linux-gnueabi-clang -c -I.
-I/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/work/armv7at2hf-vfp-neon-angstrom-linux-gnueabi/diffstat/1.60-r0/diffstat-1.60
-DHAVE_CONFIG_H -D__extern_always_inline=inline -D_GNU_SOURCE -O2
-pipe -g -feliminate-unused-debug-types -march=armv7-a -mthumb
-mfloat-abi=hard -mfpu=neon -no-integrated-as
--sysroot=/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/raspberrypi2
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/work/armv7at2hf-vfp-neon-angstrom-linux-gnueabi/diffstat/1.60-r0/diffstat-1.60/diffstat.c
NOTE: make -j 16
arm-angstrom-linux-gnueabi-clang -Wl,-O1 -Wl,--hash-style=gnu
-Wl,--as-needed -o diffstat diffstat.o
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/x86_64-linux/usr/bin/arm-angstrom-linux-gnueabi/arm-angstrom-linux-gnueabi-ld:
error: cannot open crt1.o: No such file or directory
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/x86_64-linux/usr/bin/arm-angstrom-linux-gnueabi/arm-angstrom-linux-gnueabi-ld:
error: cannot open crti.o: No such file or directory
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/x86_64-linux/usr/bin/arm-angstrom-linux-gnueabi/arm-angstrom-linux-gnueabi-ld:
error: cannot open crtbegin.o: No such file or directory
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/x86_64-linux/usr/bin/arm-angstrom-linux-gnueabi/arm-angstrom-linux-gnueabi-ld:
error: cannot open crtend.o: No such file or directory
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/x86_64-linux/usr/bin/arm-angstrom-linux-gnueabi/arm-angstrom-linux-gnueabi-ld:
error: cannot open crtn.o: No such file or directory
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/x86_64-linux/usr/bin/arm-angstrom-linux-gnueabi/arm-angstrom-linux-gnueabi-ld:
error: cannot find -lgcc
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/x86_64-linux/usr/bin/arm-angstrom-linux-gnueabi/arm-angstrom-linux-gnueabi-ld:
error: cannot find -lgcc
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/x86_64-linux/usr/bin/arm-angstrom-linux-gnueabi/arm-angstrom-linux-gnueabi-ld:
error: cannot find -lgcc
/mnt/home/kraj/work/angstrom/build/tmp-angstrom-glibc/sysroots/x86_64-linux/usr/bin/arm-angstrom-linux-gnueabi/arm-angstrom-linux-gnueabi-ld:
error: cannot find -lgcc
Khem Raj [Mon, 24 Aug 2015 01:40:43 +0000 (18:40 -0700)]
liburcu: Fix atomic function parameter types in calls
__sync_val_compare_and_swap family specifically asks for proper types of
pointer type parameters, gcc does not notice it but clang does and sends
out errors calling it out.
Khem Raj [Sun, 23 Aug 2015 02:46:23 +0000 (19:46 -0700)]
systemd: Remove exporting special CPP
This is no more needed.
it was done long ago while systemd lived in meta-openembedded
http://lists.openembedded.org/pipermail/openembedded-commits/2012-August/141061.html
The accompanying patch has been applied to systemd already so we were
not needing to set CPP for sometime now.
as a nice side effect it helps compiling systemd with clang
Khem Raj [Sat, 22 Aug 2015 14:16:20 +0000 (07:16 -0700)]
psplash: Turn psplash_fb_plot_pixel() into static inline
This function is not used anywhere except psplash-fb.c so make it static
inline function which is portable across compilers
Fixes issues like
psplash-fb.o: In function `psplash_fb_draw_rect':
|
/mnt/home/kraj/work/angstrom/sources/openembedded-core/build/tmp-glibc/work/i586-oe-linux/psplash/0.1+gitAUTOINC+14c8f7b705-r15/git/psplash-fb.c:363:
undefined reference to `psplash_fb_plot_pixel'
Khem Raj [Sat, 22 Aug 2015 15:58:11 +0000 (08:58 -0700)]
alsa-lib: Lift the limitation to use gcc only cross compiler
alsa-lib configure tries to see that if CC variabled doesnt contain
'gcc' then it tries to guess and redefine CC to be something gcc, which
doesnt work with OE since we define CC to have sysroot and options
attached to it. So all tests fail when configure constructs the CC for
cross compiler like this. This portion is not needed at all
Ed Bartosh [Fri, 21 Aug 2015 13:00:57 +0000 (16:00 +0300)]
devtool: implement build-image plugin
Implemented new plugin to build image from workspace packages.
Plugin creates <image>.bbappend file, adds
all workspace packages to the image using IMAGE_INSTALL_append
variable in bbappend file. After that it runs 'bitbake <image>'.
Signed-off-by: Ed Bartosh <ed.bartosh@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Octeon II/III binaries can contain instructions that are not compatible
with MIPS64 processors. Thus Octeon II/III packages should go to
separate directories. Set MIPSPKGSFX_VARIANT_tune-* to Octeon-specific
values and update PACKAGE_EXTRA_ARCHS_tune-* accordingly.
Robert Yang [Thu, 20 Aug 2015 06:48:10 +0000 (23:48 -0700)]
oprofile: 1.0.0 -> 1.1.0
* Remove backport patch filemode-fix.patch.
* Update --with-kernel=${STAGING_DIR_HOST}/${prefix} to find kernel
headers (linux/*.h) to fix the error:
| checking kernel supports perf_events... unknown -- perf_event.h not found
| ERROR: You requested to build oprofile with '--with-kernel=/buildarea/lyang1/test_f2/tmp/work-shared/qemux86/kernel-source',
| but headers were not accessible at the given location.
| Be sure you have run the following command from within your kernel source tree:
| make headers_install INSTALL_HDR_PATH=<kernel-hdrs-install-dir>
| Then pass <kernel-hdrs-install-dir> to oprofile's '--with-kernel' configure option.
| configure: error: Unable to build oprofile. Exiting.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Aníbal Limón [Fri, 28 Aug 2015 16:15:01 +0000 (11:15 -0500)]
systemd: Increase devices timeout in QEMU machines to avoid failures on serial-getty
Systemd serial-getty is failing because dev-tty ends with timeout, systemd
uses a default timeout of 90 secs that it's reached on AB's production environment
due to high I/O loads (nfs).
When use QEMU is used without KVM support, the machine code is executed by TCG
(software code execution) that is dependent on devices layer and causes locks
between TCG/Devices increasing the machine code execution time [1].
QEMU don't support configuration of device timeout always uses a default one that is
90 secs, so the next patch increases the device timeout to 240 secs [2] in order give
enough time to get devices ready. It ONLY applies on QEMU machines.
Randy MacLeod [Fri, 28 Aug 2015 03:09:47 +0000 (23:09 -0400)]
gcc-source: exlude from world
Exclude all versions of gcc-source from world builds so that:
bitbake -c <stage> world
will work. gcc-source deletes most bitbake build stages
since it is a source-only package.
Signed-off-by: Randy MacLeod <Randy.MacLeod@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Randy Witt [Fri, 28 Aug 2015 02:42:43 +0000 (19:42 -0700)]
qemuarm64.conf: Make the second serial console /dev/hvc0
Since the qemu for aarch64 must use a virtual console for the second
serial port rather than emulating actual hardware, make sure the correct
device is specified so that a tty is actually started.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Randy Witt [Fri, 28 Aug 2015 02:40:18 +0000 (19:40 -0700)]
sysvinit-inittab: Run ttys on ttys that don't have tty in the name
On qemuarm64 the second serial port in SERIAL_CONSOLES will be hvc0.
Since that doesn't have tty in the name, a correct label didn't get
added to inittab.
This change makes both names with tty and without work.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Thu, 27 Aug 2015 15:58:52 +0000 (16:58 +0100)]
bind: remove dependency on bind in bind-dev
bind doesn't ship shared libraries only static libraries, so the default
dependency on PN from PN-dev is pointless and means that an image with bind-dev
installed (via dhcp-dev's automatic dependency) ends up with named installed and
started on boot which is rarely intended.
If and when we ship bind's shared libraries we should ensure that the libraries
go into a separate package.
Also remove an old comment about --enable-exportlib which isn't supported by
configure anymore.
[ YOCTO #8216 ]
Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Randy Witt [Thu, 27 Aug 2015 09:04:10 +0000 (02:04 -0700)]
qemurunner: Don't loop on EWOULDBLOCK in logging thread.
EAGAIN/EWOULDBLOCK can be followed by no data. So don't tight loop
waiting for data.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Randy Witt [Wed, 26 Aug 2015 22:15:34 +0000 (15:15 -0700)]
runqemu-internal: Make sure tcpserial is always last
If this is not the case, sometimes the additional tcpserial will be
enumerated as ttyS0, which is not what we want. Because then it would be
the console, and qemurunner would not log things properly.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This patch fixes the ability to correctly identify syslog's package
name for the built image. It is derived from modifying oeqa/oetest.py
for [YOCTO #8170]
Signed-off-by: Costin Constantin <costin.c.constantin@intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
oeqa/oetest.py: add better package search for hasPackage()
Modified hasPackage() to split the content of pacakage manifest file
in containing lines and search at the begining of each line the
existance of the needed pkg.
[YOCTO #8170]
Signed-off-by: Costin Constantin <costin.c.constantin@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Randy Witt [Wed, 26 Aug 2015 07:15:15 +0000 (00:15 -0700)]
qemurunner: In the logging thread retry on EAGAIN, EWOULDBLOCK
On a nonblocking socket an exception can be generated for the EAGAIN
and EWOULDBLOCK errnos. Since these aren't actually errors make sure to
retry rather than bailing out.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Randy Witt [Tue, 25 Aug 2015 20:58:47 +0000 (13:58 -0700)]
qemurunner: Shut down logging thread successfully when test fails
Before this change on a test failure an exception would be generated due
to runqemu being killed before the logging thread which was on the other
end of the socket.
The exception was actually correct saying there was no data on a socket
marked readable, but this was because the qemu process was killed before
the listener thread.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Randy Witt [Mon, 24 Aug 2015 07:25:03 +0000 (00:25 -0700)]
runqemu: Add a tcpserial option
The option was added so that the qemurunner could start a second tcp
serial port without adding machine conditional logic to qemurunner.
The issue that made this necessary was that when "virt" is passed to
qemu-system-aarch64, the normal mechanism for specifying a tcp serial
port does not work. This is because the hardware for the "virt" machine
is hardcoded in the device tree blob and the addition devices must be
virtio devices.
So runqemu can specify virtio for qemuarm64 whereas it seems all other
qemu machines work with the "-serial tcp*" option.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Randy Witt [Thu, 20 Aug 2015 20:01:25 +0000 (13:01 -0700)]
qemurunner: Use two serial ports and log console with a thread
qemu can freeze and stop responding if the socket buffer connected to a tcp
serial connection fills up. This happens of course when the reader of
the serial data doesn't actually read it.
This happened in the qemurunner code, because after checking for the
"login:" sentinel, data was never again read from the serial connection.
This patch solves the potential freeze by adding a thread to continuously
read the data from the console and log it. So it also will give a full log
of the console, rather than just up to the login prompt.
To simplify this patch, another serial port was also added to use for the
sole purpose of watching for the sentinel as well as being the interactive
serial port. This will also prevent the possibility of lots of debug
data on the console preventing the sentinel value from being seen due to
interleaved text.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Randy Witt [Thu, 20 Aug 2015 17:55:17 +0000 (10:55 -0700)]
qemurunner: Make create_socket() return data and use exceptions
So that create_socket() can be called more than once to create sockets,
it now returns the socket and port rather than setting class variables
directly.
create_socket() now only uses exceptions for errors, not the return
value from the function.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Randy Witt [Tue, 18 Aug 2015 23:44:25 +0000 (16:44 -0700)]
qemurunner.py: Move some class variables that should only be local
The bootlog and qemusock variables were set in the class as part of the
create_socket() routine. However those variables are never used outside
of the same function and thus serve no purpose as class variables.
This initializes those variables near where they are used.
Signed-off-by: Randy Witt <randy.e.witt@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
The libcap-ng library is intended to make programming with posix
capabilities much easier than the traditional libcap library.
It's not a replacement to libcap, it provides different library
(libcap-ng.so) while packages explicitly look for libcap.so. It
could be used by qemu, util-linux, libvirt, audit ...
With adding it to oe-core, the copies from following layers could
be removed:
Reinette Chatre [Thu, 20 Aug 2015 21:34:33 +0000 (14:34 -0700)]
xf86-video-intel: enable xvmc feature by default and fix dependency
XvMC extends the X Video extension (Xv) and enables hardware rendered
motion compensation support. In a test build enabling this feature
increased the size of the xf86-video-intel package from 1386841 to 1847154 bytes.
When we enable the xvmc feature in xf86-video-intel we see the
following QA issue reported:
QA Issue: xf86-video-intel rdepends on xcb-util, but it isn't
a build dependency?
We fix this by ensuring the build dependencies are set correctly when
the xvmc feature is enabled.
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
package_rpm.bbclass : escape "%" in files and directories name
The rpm process replace all the "%name" in the spec file by the name of
the package. So, if the package is composed of some files or directories
named "%name...", the rpm package process failed.
Replace all "%" present in files or directories names by "%%%%%%%%" to
correctly escape "%" due to the number of times that % is treated as an
escape character. Jeff Johnson says this is the Right Thing To Do.
Paul Eggleton [Wed, 5 Aug 2015 08:41:08 +0000 (09:41 +0100)]
hello-mod: drop PV and PR
* PR = "r0" is the default, no need to set it (and most people won't
need to set it anyway with the PR service)
* PV = "0.1" is already set by virtue of 0.1 being in the filename
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Paul Eggleton [Wed, 5 Aug 2015 13:37:13 +0000 (14:37 +0100)]
oe-selftest: replace assertTrue(False, ...) with fail(...)
I'd somehow missed the existence of fail() when I wrote these. It's
preferable here so you don't get the somewhat useless "false is not
true" message in the case of failure.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Paul Eggleton [Thu, 23 Jul 2015 13:56:02 +0000 (14:56 +0100)]
scripts/create-recipe: drop
We have "recipetool create" (or "devtool add" which wraps it) instead
which works pretty well now, is written in Python and thus can call
into BitBake/OE code easily, and can do things that create-recipe can't;
any future improvements should go into recipetool and therefore I think
we'd rather people used it instead.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Otherwise QA check will fail.
Some schemas in gsettings-desktop-schemas (such as proxy and locale)
are still using deprecated paths, as of 3.16.1. This causes warning
messages, and meta/lib/oe/rootfs.py complaints about them.
Signed-off-by: Alexander Kanavin <alexander.kanavin@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Khem Raj [Wed, 19 Aug 2015 17:44:56 +0000 (10:44 -0700)]
kernel-arch: Force BFD kernel when using gcc for linking
We redefine LD to point to ld.bfd when building kernel, which works in
most cases since kbuild system calls out for bare LD most of the time,
however some of newer kernels e.g. 4.1+ have some code added which can
call gcc directly to do the linking job e.g. arm vdso code
This causes build failures when we have configured the default cross
toolchain to use gold linker as default. Errors like
BFD: arch/arm/vdso/vdso.so: Not enough room for program headers, try
linking with -N
| arm-angstrom-linux-gnueabi-objcopy:arch/arm/vdso/vdso.so[.hash]: Bad
value
start happening.
With this patch we force gcc to choose bfd linker as well
Signed-off-by: Khem Raj <raj.khem@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
