We Keep Coding

How do I avoid an ELFCLASS64 error when building Android 8 from source?

I have an old tablet I'm toying around with for the fun of it. I'm trying to build Android 8 from source just to tinker with it and see if I can install it on the tablet. I'm using Lubuntu 20.04 LTS.
So far, all I've done is set up my environment, sync to the android-8.1.0_r81 repo, and run m, all per the instructions here: https://source.android.com/setup
My tablet uses an ARMv7 processor, so when I ran lunch I chose aosp_arm-eng, which corresponds to a generic 32-bit ARMv7 processor.
However, when I run m, it eventually fails with the following error:
[ 0% 1/9730] build out/target/product/generic/obj/ETC/sepolicy_tests_intermediates/sepolicy_tests
FAILED: out/target/product/generic/obj/ETC/sepolicy_tests_intermediates/sepolicy_tests
/bin/bash -c "(python out/host/linux-x86/bin/sepolicy_tests.py -l out/host/linux-x86/lib64 -f out/target/product/generic/obj/ETC/plat_file_contexts_intermediates/plat_file_contexts -f out/target/product/generic/obj/ETC/nonplat_file_contexts_intermediates/nonplat_file_contexts -p out/target/product/generic/obj/ETC/sepolicy_intermediates/sepolicy ) && (touch out/target/product/generic/obj/ETC/sepolicy_tests_intermediates/sepolicy_tests )"
Traceback (most recent call last):
File "out/host/linux-x86/bin/sepolicy_tests.py", line 73, in <module>
pol = policy.Policy(options.policy, options.file_contexts, options.libpath)
File "/home/username/Android/Development/Oreo/out/host/linux-x86/bin/policy.py", line 255, in __init__
self.__InitLibsepolwrap(LibPath)
File "/home/username/Android/Development/Oreo/out/host/linux-x86/bin/policy.py", line 190, in __InitLibsepolwrap
lib = CDLL(LibPath + "/libsepolwrap.so")
File "/home/username/Android/Development/Oreo/prebuilts/python/linux-x86/2.7.5/lib/python2.7/ctypes/__init__.py", line 365, in __init__
self._handle = _dlopen(self._name, mode)
OSError: out/host/linux-x86/lib64/libsepolwrap.so: wrong ELF class: ELFCLASS64
...
From what I understand, the out directory is the output of the make command--out isn't in the original repo. I also understand (I think) that the wrong ELF class: ELFCLASS64 error means that the libsepolwrap.so library is expecting a 32-bit OS and getting a 64-bit one. What I don't understand is why libsepolwrap.so is expecting a 32-bit OS.
According to the docs,
A 64-bit environment is required for Android 2.3.x (Gingerbread) and higher versions, including the master branch. You can compile older versions on 32-bit systems.
Given that I haven't actually modified anything yet and I'm building a pure vanilla Android 8 ROM, I'm stumped as to how to even start trying to fix this.
Side Note: I am already aware that I can find a vanilla Android 8 ROM somewhere online; I'm more interested in building it myself so I can learn the process and have something to tinker with.

Buildroot Python Dependencies (_sqlite3)

Full Disclaimer: I have been using buildroot for the last 6 weeks. This is my first introduction to embedded Linux, thus I am still very green but have been able to solve 99% of my problems myself. for the most part the process has been straightforward.
So to occupy myself while I am stuck and home and cant work I have been working on an embedded hardware project. I've selected my hardware, built a prototype, learned buildroot basics, brought up the basic system, optimized the kernel config, built a custom device tree for my hardware and I am happy up until this point.
In parallel I have been programming the target application in python on my desktop, its dependent on a few libraries (hardware abstraction, communication, display etc) but is relatively straight forward.
I've got to the point where I have a list of requisite packages to build into my buildroot system. The buildroot tools are great here using the scanpypi tool:
~/buildroot$ utils/scanpypi diskcache -o package
Simply adding all the dependencies into /package/config.in has allowed them to be selected in menuconfig and added to the recipe.
The problem comes at build time where the building of the python module fails for the module above python-diskcache.
It has dependencies on a few things but one is slqite3, this has been added as:
the core python module "sqlite module"
external package "python-pysqlite3"
libraries > database > sqlite
However, it fails at build:
>>> python-diskcache 4.1.0 Building
Traceback (most recent call last):
File "setup.py", line 5, in <module>
import diskcache
File "/home/buildroot/output/build/python-diskcache-4.1.0/diskcache/__init__.py", line 9, in <module>
from .core import Cache, Disk, EmptyDirWarning, JSONDisk, UnknownFileWarning, Timeout
File "/home/buildroot/output/build/python-diskcache-4.1.0/diskcache/core.py", line 14, in <module>
import sqlite3
File "/home/buildroot/output/host/arm-buildroot-linux-uclibcgnueabihf/sysroot/usr/lib/python3.8/sqlite3/__init__.py", line 23, in <module>
from sqlite3.dbapi2 import *
File "/home/buildroot/output/host/arm-buildroot-linux-uclibcgnueabihf/sysroot/usr/lib/python3.8/sqlite3/dbapi2.py", line 27, in <module>
from _sqlite3 import *
ModuleNotFoundError: No module named '_sqlite3'
make[1]: *** [package/pkg-generic.mk:269: /home/buildroot/output/build/python-diskcache-4.1.0/.stamp_built] Error 1
make: *** [Makefile:84: _all] Error 2
In looking for a solution it seems like _sqlite3 is the C module external to python for communicating with an sqlite database. It should be installed with python (using 3.8) and should be enabled in buildroot with the enabling of the core sqlite module.
discussion 1
discussion 2
There are several fixes for dealing with this issue on a host (e.g. apt get install libsqlite3-dev and reinstall/reconfigure python). Obviously this is not possible in the image, and both the sqlite and python3 installs are latest builds and installed to the image at build time.
I'm really struggling to understand the problem or how I might patch it. I have a few theories based on the discussion, but I am unsure.
1) python is getting installed to the image before sqlite, so the appropriate module is not getting cp or symlink'ed to the python install.
2) there is some other unknown dependency not being met at build time and its failing silently
any ideas or assistance would be hugely appreciated.
Thanks

The problem is that Python on your build machine finds the cross-compiled _sqlite3 module, which it can't load because it's for the wrong architecture.
This usually doesn't occur, because the setup script normally doesn't go and load the package it is trying to build/install.
One workaround could be to install the host version of all the dependencies of diskcache, and set PYTHONPATH=$(HOST_DIR)/lib/python$(PYTHON3_VERSION_MAJOR)/:$(PYTHON3_PATH) in DISKCACHE_ENV. However, this is liable to lead to all kinds of other breakage.
A better solution, therefore, is to patch the setup.py script of diskcache so it doesn't try to import diskcache itself. It probably only does that to get a version number or something similar; that can be solved by moving the version number into a separate file and load that one instead.

As suggested by Arnout, I got my diskcache building by creating a following patch
diff --git a/setup.py b/setup.py
index b49d9c8..22cd3c8 100644
--- a/setup.py
+++ b/setup.py
## -3,8 +3,6 ## from io import open
from setuptools import setup
from setuptools.command.test import test as TestCommand
-import diskcache
-
class Tox(TestCommand):
def finalize_options(self):
## -23,8 +21,8 ## with open('README.rst', encoding='utf-8') as reader:
readme = reader.read()
setup(
- name=diskcache.__title__,
- version=diskcache.__version__,
+ name='diskcache',
+ version='5.4.0',
description='Disk Cache -- Disk and file backed persistent cache.',
long_description=readme,
author='Grant Jenks',
--
2.25.1

How can I source two paths for the ROS environmental variable at the same time?

I have a problem with using the rqt_image_view package in ROS. Each time when I type rqt_image_view or rosrun rqt_image_view rqt_image_view in terminal, it will return:
Traceback (most recent call last):
File "/opt/ros/kinetic/bin/rqt_image_view", line 16, in
plugin_argument_provider=add_arguments))
File "/opt/ros/kinetic/lib/python2.7/dist-packages/rqt_gui/main.py", line 59, in main
return super(Main, self).main(argv, standalone=standalone, plugin_argument_provider=plugin_argument_provider, plugin_manager_settings_prefix=str(hash(os.environ['ROS_PACKAGE_PATH'])))
File "/opt/ros/kinetic/lib/python2.7/dist-packages/qt_gui/main.py", line 338, in main
from python_qt_binding import QT_BINDING
ImportError: cannot import name QT_BINDING
In the /.bashrc file, I have source :
source /opt/ros/kinetic/setup.bash
source /home/kelu/Dropbox/GET_Lab/leap_ws/devel/setup.bash --extend
source /eda/gazebo/setup.bash --extend
They are the default path of ROS, my own working space, the robot simulator of our university. I must use all of them. I have already finished many projects with this environmental variable setting. However, when I want to use the package rqt_image_view today, it returns the above error info.
When I run echo $ROS_PACKAGE_PATH, I get the return:
/eda/gazebo/ros/kinetic/share:/home/kelu/Dropbox/GET_Lab/leap_ws/src:/opt/ros/kinetic/share
And echo $PATH
/usr/local/cuda/bin:/opt/ros/kinetic/bin:/usr/local/cuda/bin:/usr/local/cuda/bin:/home/kelu/bin:/home/kelu/.local/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin
Then I only source the /opt/ros/kinetic/setup.bash ,the rqt_image_view package runs!!
It seems that, if I want to use rqt_image_view, then I can not source both /opt/ros/kinetic/setup.bash and /home/kelu/Dropbox/GET_Lab/leap_ws/devel/setup.bash at the same time.
Could someone tell me how to fix this problem? I have already search 5 hours in google and haven't find a solution.

Different solutions to try:
It sounds like the first path /eda/gazebo/ros/kinetic/share or /home/kelu/Dropbox/GET_Lab/leap_ws/src has an rqt_image_view package that is being used. Try to remove that dependency.
Have you tried switching the source files being sourced? This depends on how the rqt_image_view package was built, such as by source or through a package manager.
Initially, it sounds like there is a problem with the paths being searched or wrong package being run since the package works with the default ROS environment setup.

Python Evdev binding for OpenWrt

Good day,
I'm a student and I would just like to ask for a minute of your time.
I'm working on a barcode reader connected via USB port to a board name Arduino Yun. This board runs a version of embedded linux derived from OpenWrt using a microprocessor named Atheros AR9331
I would like to ask you, what's necessary to make the Python Evdev binding (python-evdev.readthedocs.org/en/latest/), to be able to run in this type of MIPS microarchitecture? At the momento, it's only for Ubuntu and ArchLinux.
I'm kind of guessing that cross compilation would be needed, or the indication of the usage of a specific C compiler inside this linux.
The current python version supported for OpenWrt is 2.7.3
I already know , if you compile C code in your PC, the resulting executable will only run in this type of architecture. If you use that compiled program inside the microprocessor, it wont work.
I've used this binding without trouble within ubuntu in my PC. I followed the instructions, python setup.py install, with a previous installation of setuptools, and it worked just fine.
But regarding OpenWrt, this was not the case.
The python script I'm using requires this library within the first line of code in order to reach the data from the device (it works like a keyboard /dev/input/event0):
#!/usr/bin/env python
from evdev import InputDevice, ecodes, list_devices
from select import select
I've seen suggestions of copying the entire library inside the arduino, and run the script inside the same folder. But it doesn't work, since the evdev module has files created with the architecture of the PC and not the MIPS.
So, what are the messages displayed for the error?
If you run python setup.py install in Openwrt to try to install the evdev binding, this appears on screen:
File "setup.py", line 10, in <module>
from setuptools.command.develop import develop
ImportError: No module named setuptools.command.develop
It's obvious from here that you need the module aforementioned. So, I tried to install it with this script (pypi.python.org/pypi/setuptools):
python ez_setup.py
And the output shows this:
Downloading https://pypi.python.org/packages/source/s/setuptools/setuptools-11.3.1.zip
Traceback (most recent call last):
File "ez_setup.py", line 332, in <module>
sys.exit(main())
File "ez_setup.py", line 327, in main
downloader_factory=options.downloader_factory,
File "ez_setup.py", line 287, in download_setuptools
downloader(url, saveto)
File "ez_setup.py", line 209, in download_file_curl
_clean_check(cmd, target)
File "ez_setup.py", line 169, in _clean_check
subprocess.check_call(cmd)
File "/usr/lib/python2.7/subprocess.py", line 511, in check_call
raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['curl','https://pypi.python.org/packages/source/s/setuptools/setuptools-11.3.1.zip', '--silent', '--output', '/mnt/sda1/evdev-0.4.6/setuptools-11.3.1.zip']' returned non-zero exit status 60
I pressume this output is due to the fact that pypi doesn't exist for the python 2.7.3 in OpenWrt , only for newer versions and other architectures. Evedv binding is requiring the setuptools module in order to make things easier and standard, but if the binding is not supported for the target architecture, what's needed to be able to use it anyways?
Thanks for your time,

Good day everyone,
The solution was provided by Georgi Valkov. He is the creator of the python-evdev binding. I contacted him directly, and he was so kind that he cross compiled a version for the OpenWrt / Yun .
You can install the package using the openwrt package manager - opkg. The installation process is along the lines of:
$ opkg update
$ opkg install /path/to/python-evdev_0.4.7-1_ar71xx.ipk
To verify that the install was successful:
$ opkg files python-evdev
/usr/lib/python2.7/site-packages/evdev-0.4.7-py2.7.egg-info
/usr/lib/python2.7/site-packages/evdev/genecodes.py
/usr/lib/python2.7/site-packages/evdev/ff.py
/usr/lib/python2.7/site-packages/evdev/_input.so
/usr/lib/python2.7/site-packages/evdev/device.py
/usr/lib/python2.7/site-packages/evdev/events.py
/usr/lib/python2.7/site-packages/evdev/__init__.py
/usr/lib/python2.7/site-packages/evdev/ecodes.py
/usr/lib/python2.7/site-packages/evdev/_ecodes.so
/usr/lib/python2.7/site-packages/evdev/util.py
/usr/lib/python2.7/site-packages/evdev/uinput.py
/usr/lib/python2.7/site-packages/evdev/_uinput.so
This works just fine. Thanks.
PS. If someone needs the file, please contact me. Georgi sent me this address, but I didn't download the file from there because he sent it to me over email.
https://github.com/gvalkov/openwrt-packages-yun/blob/master/lang/python-evdev/Makefile

In the output, you can see that curl returned the status code 60. According to man curl
60 Peer certificate cannot be authenticated with known CA certifi‐
cates.
According to the setuptools page, you can instead use python ez_setup.py --insecure but obviously do that at your own risk. Alternatively you could do what the advanced instructions say and manually download the setuptools tarball, verify its md5 hash yourself, and install it using its setup.py .

Analytics API + Python Server, NotImplementedError Hello Analytics

A little background: I've been trying to make a restful server that can query and insert via the management API. After banging my head against the wall using node.js and javascript I switched over to python knowing it has more support.
Currently I am trying to follow the GA Tutorial: Hello Analytics API with a slight twist. I trigger the what was the main method in hello_analytics_api_v3.py when trying to access a particular end point on the server. This is only for testing. The method is called insertExperiment and attempts to insert an experiment into a private Google Analytics account.
However I'm always receiving the exception:
NotImplementedError('The gflags library must be installed to use tools.run(). Please install gflags or preferrably switch to using tools.run_flow().',)
Here is the full trace stack:
File "/0zzz/bottle.py", line 764, in _handle
return route.call(**args)
File "/0zzz/bottle.py", line 1575, in wrapper
rv = callback(*a, **ka)
File "server.py", line 39, in server_static
hello_analytics_api_v3.insertExperiment("xxxxxxx", "xxxxxxxxx", "xxxxxx", experiment_body)
File "/0zzz/hello_analytics_api_v3.py", line 11, in insertExperiment
service = hello_analytics_api_v3_auth.initialize_service()
File "/0zzz/hello_analytics_api_v3_auth.py", line 32, in
initialize_service
credentials = prepare_credentials()
File "/0zzz/hello_analytics_api_v3_auth.py", line 25, in
prepare_credentials
credentials = run(FLOW, storage)
File "/0zzz/oauth2client/tools.py", line 241, in run
'The gflags library must be installed to use tools.run(). '
NotImplementedError: The gflags library must be installed to use
tools.run(). Please install gflags or preferrably switch to using
tools.run_flow().
A few extra notes:
I have tried this exact same thing (invoking main on pinging the correct address) when copying the exact code from the Tutorial.
I have write access with the management API -- so that's not a problem here at all.
I thought there might be a namespace issue with one of my libraries (which also imports a run) but resolved this by renaming the other library.
GA Client Library is installed! as per the instructions on Hello Analytics API.
Thank you for any help you can provide!

I received the same error while running through a tutorial for google APIs called 'Google APIs Console Help'. The fix was simple in my case, just update the gflags library:
easy_install --upgrade python-gflags

You need to install gflags library.
$ wget https://python-gflags.googlecode.com/files/python-gflags-2.0.tar.gz
$ tar xfvz python-gflags-2.0.tar.gz ; cd python-gflags-2.0
$ python setup.py install