Hi Hongxu,

On Wed, Dec 16, 2020 at 2:09 PM Hongxu Jia <hongxu.jia@...> wrote:

Signed-off-by: Hongxu Jia <hongxu.jia@...>
---
.../0001-customize-for-yocto.patch | 28 +++++++++++++++++++
.../tensorflow/tensorflow-estimator_1.13.bb | 12 ++++++--
2 files changed, 37 insertions(+), 3 deletions(-)
create mode 100644 recipes-framework/tensorflow/tensorflow-estimator/0001-customize-for-yocto.patch

diff --git a/recipes-framework/tensorflow/tensorflow-estimator/0001-customize-for-yocto.patch b/recipes-framework/tensorflow/tensorflow-estimator/0001-customize-for-yocto.patch
new file mode 100644
index 0000000..e9b66d5
--- /dev/null
+++ b/recipes-framework/tensorflow/tensorflow-estimator/0001-customize-for-yocto.patch
@@ -0,0 +1,28 @@
+From a1bcf09a43fc44ad5e04c441ee45cc23d16cf7d2 Mon Sep 17 00:00:00 2001
+From: Hongxu Jia <hongxu.jia@...>
+Date: Wed, 9 Dec 2020 17:59:01 +0800
+Subject: [PATCH] customize for yocto
+
+Upstream-Status: Inappropriate [oe specific]
+
+Signed-off-by: Hongxu Jia <hongxu.jia@...>
+---
+ tensorflow_estimator/tools/pip_package/build_pip_package.sh | 2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+diff --git a/tensorflow_estimator/tools/pip_package/build_pip_package.sh b/tensorflow_estimator/tools/pip_package/build_pip_package.sh
+index d4953a6..e08cd8a 100755
+--- a/tensorflow_estimator/tools/pip_package/build_pip_package.sh
++++ b/tensorflow_estimator/tools/pip_package/build_pip_package.sh
+@@ -38,7 +38,7 @@ function prepare_src() {
+
+ # Verifies all expected files are in pip.
+ # Creates init files in all directory in pip.
+- python tensorflow_estimator/tools/pip_package/create_pip_helper.py --pip-root "${TMPDIR}/tensorflow_estimator/" --bazel-root "./tensorflow_estimator"
++ nativepython3 tensorflow_estimator/tools/pip_package/create_pip_helper.py --pip-root "${TMPDIR}/tensorflow_estimator/" --bazel-root "./tensorflow_estimator"
+ }
+
+ function build_wheel() {
+--
+2.18.2
+
diff --git a/recipes-framework/tensorflow/tensorflow-estimator_1.13.bb b/recipes-framework/tensorflow/tensorflow-estimator_1.13.bb
index f3a5098..5b2fe5d 100644
--- a/recipes-framework/tensorflow/tensorflow-estimator_1.13.bb
+++ b/recipes-framework/tensorflow/tensorflow-estimator_1.13.bb
@@ -3,9 +3,10 @@ learning programming."
LICENSE = "Apache-2.0"
LIC_FILES_CHKSUM = "file://LICENSE;md5=01e86893010a1b87e69a213faa753ebd"

-SRC_URI = "git://github.com/tensorflow/estimator.git;branch=r1.13 \
+SRC_URI = "git://github.com/tensorflow/estimator.git;branch=r2.4 \
+ file://0001-customize-for-yocto.patch \
"
-SRCREV = "340703eed78ba4854862f749ed94f91598826e79"
+SRCREV = "c3e7f2b5bbcc35185ef71797955a28cadce28f60"
S = "${WORKDIR}/git"

inherit python3native bazel
@@ -19,12 +20,18 @@ DEPENDS += " \
python3-astor-native \
python3-gast-native \
python3-termcolor-native \
+ python3-wrapt-native \
+ python3-opt-einsum-native \
+ python3-astunparse-native \
+ flatbuffers-native \

I'm getting issue when building tensorflow-estimator (I add small
changes to be able to build this patches on top of dunfell) and I'm
getting:
| /bin/bash -c 'source
external/bazel_tools/tools/genrule/genrule-setup.sh;
bazel-out/k8-opt-exec-2B5CBBC6/bin/tensorflow_estimator/python/estimator/api/create_tensorflow_estimator.python.estimator_api_1_estimator_python_api_gen_compat_v1
--apidir=bazel-out/k8-fastbuild/bin/tensorflow_estimator/python/estimator/api_v1/
--apiname=estimator --apiversion=1
--package=tensorflow_estimator.python.estimator
--output_package=tensorflow_estimator.python.estimator.api._v1
bazel-out/k8-fastbuild/bin/tensorflow_estimator/python/estimator/api/_v1/v1.py
bazel-out/k8-fastbuild/bin/tensorflow_estimator/python/estimator/api/_v1/estimator/__init__.py
bazel-out/k8-fastbuild/bin/tensorflow_estimator/python/estimator/api/_v1/estimator/experimental/__init__.py
bazel-out/k8-fastbuild/bin/tensorflow_estimator/python/estimator/api/_v1/estimator/export/__init__.py
bazel-out/k8-fastbuild/bin/tensorflow_estimator/python/estimator/api/_v1/estimator/inputs/__init__.py
bazel-out/k8-fastbuild/bin/tensorflow_estimator/python/estimator/api/_v1/estimator/tpu/__init__.py
bazel-out/k8-fastbuild/bin/tensorflow_estimator/python/estimator/api/_v1/estimator/tpu/experimental/__init__.py')
| Execution platform: @local_config_platform//:host
| /home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/h5py/__init__.py:72:
UserWarning: h5py is running against HDF5 1.8.21 when it was built
against 1.8.19, this may cause problems
| _warn(("h5py is running against HDF5 {0} when it was built against {1}, "
| Traceback (most recent call last):
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/bazel/output_base/execroot/org_tensorflow_estimator/bazel-out/k8-opt-exec-2B5CBBC6/bin/tensorflow_estimator/python/estimator/api/create_tensorflow_estimator.python.estimator_api_1_estimator_python_api_gen_compat_v1.runfiles/org_tensorflow_estimator/tensorflow_estimator/python/estimator/api/create_python_api_wrapper.py",
line 26, in <module>
| from tensorflow_estimator.python.estimator import estimator_lib
# pylint: disable=unused-import
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/bazel/output_base/execroot/org_tensorflow_estimator/bazel-out/k8-opt-exec-2B5CBBC6/bin/tensorflow_estimator/python/estimator/api/create_tensorflow_estimator.python.estimator_api_1_estimator_python_api_gen_compat_v1.runfiles/org_tensorflow_estimator/tensorflow_estimator/python/estimator/estimator_lib.py",
line 22, in <module>
| from tensorflow_estimator.python.estimator.canned.baseline
import BaselineClassifier
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/bazel/output_base/execroot/org_tensorflow_estimator/bazel-out/k8-opt-exec-2B5CBBC6/bin/tensorflow_estimator/python/estimator/api/create_tensorflow_estimator.python.estimator_api_1_estimator_python_api_gen_compat_v1.runfiles/org_tensorflow_estimator/tensorflow_estimator/python/estimator/canned/baseline.py",
line 53, in <module>
| import tensorflow as tf
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/__init__.py",
line 55, in <module>
| from ._api.v2 import compat
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/_api/v2/compat/__init__.py",
line 39, in <module>
| from . import v1
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/_api/v2/compat/v1/__init__.py",
line 34, in <module>
| from . import compat
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/_api/v2/compat/v1/compat/__init__.py",
line 39, in <module>
| from . import v1
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/_api/v2/compat/v1/compat/v1/__init__.py",
line 51, in <module>
| from tensorflow._api.v2.compat.v1 import lite
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/_api/v2/compat/v1/lite/__init__.py",
line 11, in <module>
| from . import experimental
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/_api/v2/compat/v1/lite/experimental/__init__.py",
line 10, in <module>
| from . import nn
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/_api/v2/compat/v1/lite/experimental/nn/__init__.py",
line 10, in <module>
| from tensorflow.lite.python.lite import TFLiteLSTMCell
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/lite/python/lite.py",
line 40, in <module>
| from tensorflow.lite.python.convert import
build_toco_convert_protos # pylint: disable=unused-import
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/lite/python/convert.py",
line 33, in <module>
| from tensorflow.lite.python import util
| File "/home/ubuntu/projects/test/build/tmp/work/aarch64-poky-linux/tensorflow-estimator/2.4-r0/recipe-sysroot-native/usr/lib/python3.8/site-packages/tensorflow/lite/python/util.py",
line 30, in <module>
| import flatbuffers
| ModuleNotFoundError: No module named 'flatbuffers'

FLatbuffers from openembedded are available version 1.12. Do you have
an idea what is can cause?

tensorflow-native \
"

do_compile () {
unset CC
export TMPDIR="${WORKDIR}"
+ export PYTHON_BIN_PATH="${PYTHON}"
+
${BAZEL} build \
--subcommands --explain=${T}/explain.log \
--verbose_explanations --verbose_failures \
@@ -32,7 +39,6 @@ do_compile () {
--python_path="${PYTHON}" \
//tensorflow_estimator/tools/pip_package:build_pip_package

- PYTHON_BIN_PATH="${PYTHON}" \
${S}/bazel-bin/tensorflow_estimator/tools/pip_package/build_pip_package \
${WORKDIR}/estimator_pip
}
--
2.21.0

Thanks and BR,

marek
--
as simple and primitive as possible
-------------------------------------------------
Marek Belisko - OPEN-NANDRA
Freelance Developer

Ruska Nova Ves 219 | Presov, 08005 Slovak Republic
Tel: +421 915 052 184
skype: marekwhite
twitter: #opennandra
web: http://open-nandra.com

From: Paul D. DeRocco

Using Yocto 3.2.1 on an Intel target, trying to use the x32 model. I'm
getting this compile error when I use the SDK to separately compile my
application. I get no such errors when I build Linux, or the SDK.

In file included from

/opt/poky/3.2.1/i64/sysroots/core2-64-poky-linux/usr/include/sys/cdefs.h:453

,
from

/opt/poky/3.2.1/i64/sysroots/core2-64-poky-linux/usr/include/features.h:465,

from
/opt/poky/3.2.1/i64/sysroots/core2-64-poky-linux/usr/include/dirent.h:25,
from ../my_header_file.h:7,
from ../my_source_file.cpp:3:

/opt/poky/3.2.1/i64/sysroots/core2-64-poky-linux/usr/include/bits/long-doubl

e.h:23:10: fatal error: bits/long-double-32.h: No such file
or directory

etc., etc.

I solved this myself by building an SDK specific to the x32 configuration,
rather than trying to use the plain 64-bit SDK. What led me to believe that
the 64-bit SDK should work is that the host sysroot includes
/usr/bin/x86_64-poky-linux and /usr/bin/x86_64-poky-linux-gnux32
subdirectories. The latter contains symlinks into the former, which is no
surprise, since one toolchain should be able to do 32, 64, and 64x32, but
the target sysroot include files don't quite work. The only reason I care is
that each SDK is about 2.4GB. So I'm not sure if this is a bug, or whose bug
it is, but 2.4GB seems awfully large for an SDK, especially since this is
the standard SDK, not the ESDK.

--

Ciao, Paul D. DeRocco
Paul mailto:pderocco@...

I've been resurrecting an old Pyro project under Gatesgarth. It's an Intel
32-bit system that needs maximum speed, so I decided to try to build the
system and application as 64-bit, for more registers. It pretty much worked
on the first try, and is about 8% faster. Now I'm trying to do it as x32, to
see if that speeds it up even more.

Unable to find any specific instructions, I set the DEFAULTTUNE to
"core2-64-x32" in my BSP conf file. Also, in the old project I had tinkered
around with this a bit, and had found some kernel config settings somewhere,
so I tried using them again:

CONFIG_X86_X32=y
CONFIG_COMPAT=y
CONFIG_COMPAT_FOR_U64_ALIGNMENT=y
CONFIG_SYSVIPC_COMPAT=y
CONFIG_KEYS_COMPAT=y
CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS=y
CONFIG_ARCH_MMAP_RND_COMPAT_BITS=8
CONFIG_BLOCK_COMPAT=y

Other than the first one, I don't know whether these are correct, or where I
originally got them, or if there's a stock .cfg file that does all this.

The system built fine, the SDK built fine, and my application built fine,
but when I boot, I get a kernel panic, preceded by some message about "kmod
busy with 50 threads for more than 5 seconds now".

Is there some instruction on how to do a proper x32 build? I couldn't find
one Googling. Barring that, do any of those kernel configs look bogus?

--

Ciao, Paul D. DeRocco
Paul mailto:pderocco@...

Using Yocto 3.2.1 on an Intel target, trying to use the x32 model. I'm
getting this compile error when I use the SDK to separately compile my
application. I get no such errors when I build Linux, or the SDK.

In file included from
/opt/poky/3.2.1/i64/sysroots/core2-64-poky-linux/usr/include/sys/cdefs.h:453
,
from
/opt/poky/3.2.1/i64/sysroots/core2-64-poky-linux/usr/include/features.h:465,
from
/opt/poky/3.2.1/i64/sysroots/core2-64-poky-linux/usr/include/dirent.h:25,
from ../my_header_file.h:7,
from ../my_source_file.cpp:3:
/opt/poky/3.2.1/i64/sysroots/core2-64-poky-linux/usr/include/bits/long-doubl
e.h:23:10: fatal error: bits/long-double-32.h: No such file or directory

long-double.h includes either long-double-32.h or long-double-64.h based on
the __WORDSIZE macro, which is 32. This works fine when compiling straight
32-bit or 64-bit code, but fails in x32 code because only long-double-64.h
exists. I don't see why the characteristics of a long double should have
anything to do with the "word size" of a pointer or long int. And indeed,
the two headers it chooses from are basically empty, except for defining
__LDOUBLE_REDIRECTS_TO_FLOAT128_ABI. I'm also not sure why it would be
harmful for both of these files to exist always.

So this is an obvious bug. But my question is: what's the cleanest way to
get over this hump? Patch cdefs.h when building the SDK? Use a bbappend on
some SDK recipe to copy long-double-64.h to long-double-32.h? I have no idea
where this recipe would be.

--

Ciao, Paul D. DeRocco
Paul mailto:pderocco@...

Hi Nikhil:

I searched for a solution to this for a very long time.  I never did find a solution.  However I also learned I really did not need this for my .hddimg.  

For early development booting from a USB stick, I simply used a WIC image partitioned in the way that I wanted my target's flash to be partitioned.   I referenced this thread for the clues I needed to create the partitions I needed on the wic image.

https://stackoverflow.com/questions/56187209/yocto-create-and-populate-a-separate-home-partition

Of course on the WIC boot menu, there is no install option, so that doesn't quite get you to a standalone target running from internal storage.

Then I figured out that if I only use the .hddimage for installation to internal storage rather than for operation, you can modify (by using a bbappend) the installation script to create the partitions you want during installation.  

For us that script was located at meta/recipes-core/initrdscripts/files/init-install-efi.sh .  Its actually unusually well commented so you can probably follow along after a few re-reads and understand what it is doing.  

Getting a software update scheme working (that's a whole other subject) while you are still working on the wic image will allow you to basically install from the .hddimg once (or whenever you trash your target drive :) ) and then use your update scheme to install a new image, and skip the USB stick step altogether.

Sorry to be so wordy, but hope this helps.
Howard

Hi,

To remove completely ipv6 from distro disable all the kernel configurations related to ipv6, add the the below line in the local.conf file
DISTRO_FEATURES_remove = "ipv6".

If above doesnot help create your own distro file owndistro.conf and add below lines to remove unwanted features as follows

DISTRO_FEATURES_remove = " ipv6 x11 wayland"

and change local.conf as follows

DISTRO ?= "owndistro"



Regards,
Sadiq

Hi Guys,

        I am trying to build core-image-sato using dunfell sources. i am facing bug quilt-native configure issue. its present in meta/recipes-devtools.

Below i am attaching  my issue odt file. can please check it . help me to solve this issue.

I'm upgrading a project from Pyro to Gatesgarth. Building the toolchain, it
barfs on a patch which isn't needed. Here's one of the error messages:

ERROR: binutils-native-2.35-r0 do_patch: Command Error: 'quilt --quiltrc
/home/pauld/yocto-gatesgarth/build-intel32/tmp/work/x86_64-linux/binutils-na
tive/2.35-r0/recipe-sysroot-native/etc/quiltrc push' exited with 0 Output:
Applying patch gas_as.h.patch
patching file gas/as.h
Hunk #1 FAILED at 486.
1 out of 1 hunk FAILED -- rejects in file gas/as.h
Patch gas_as.h.patch can be reverse-applied

There are two more just like it, for the cross and target versions. All it
is is a patch to add one external function definition, but it's already
there.

Is there a preferred way to fix this with a bbappend? Perhaps replacing the
patch file with an empty one?

--

Ciao, Paul D. DeRocco
Paul mailto:pderocco@...

On Tue, 29 Dec 2020, Howard wrote:

Yup, thank you Bruce:

Using
IMAGE_FSTYPES_remove += "wic"

Didn't do it for me, but

IMAGE_TYPES_remove += "wic"

did the trick. 

pedantry, but i'm fairly sure "=" would have sufficed; there was no
need to use "+=" here.

rday

Yup, thank you Bruce:

Using
IMAGE_FSTYPES_remove += "wic"

Didn't do it for me, but

IMAGE_TYPES_remove += "wic"

did the trick. 

I would recommend that you go for a systemd drop-in file [1].
That would be something like a small snippet placed to /etc/systemd/system/dbus-org.bluez.service/override.conf

In that file place something like
[Service]
ExecStart=
ExecStart=<new command line with your custom options>

this way you don't need to alter the upstream recipe with weird sed hacks or similar.

This drop-in file could be packaged by any recipe or bbappend.

[1] https://coreos.com/os/docs/latest/using-systemd-drop-in-units.html

Are you able to resolve this issue ? I am seeing the similar issue
p: cannot stat ‘Module.symvers’: No such file or directory
WARNING: /home/jojan.vazhaeparampil/workspace/distro/build/tmp/work/cyclone5-poky-linux-gnueabi/linux-altera/5.8+gitAUTOINC+eb446787ab-r0/temp/run.do_shared_workdir.21085:1 exit 1 from 'cp Module.symvers $kerneldir/'