spsdk
spsdk copied to clipboard
NXPmicro
Make use of pypemicro optional or remove dependency on it altogether
Hi! As raised in https://github.com/NXPmicro/pypemicro/issues/10 I believe that the origins of the shared libraries provided by the pypemicro package are not trustworthy enough to be shipped to users.
Please consider removing this dependency altogether or at least making it an optional dependency for this project.
From a packaging perspective I do not feel comfortable packaging spsdk with it depending on pypemicro and would remove its use from the source code if necessary, until legal claims and claims of origin are resolved for the shared libraries in question.
@saper @mstarecek @mwsk sorry for pinging you here as well. The issue in https://github.com/NXPmicro/pypemicro/issues/10 is still unresolved, which means that packaging pypemicro somwhere downstream is still legally very questionable and from a security perspective also quite dubious.
It would be very good if NXP could address this issue.
That is very kind of you to tag me here but I don't have the powers to do anything about it. Still running a locally modified version of https://github.com/NXPmicro/spsdk/releases/tag/0.2.2, wonder how much it will take to update - so I can't even battle-test it.
That is very kind of you to tag me here but I don't have the powers to do anything about it.
My bad! I was under the impression that you were somehow affiliated, having contributed to the repository!
Still running a locally modified version of https://github.com/NXPmicro/spsdk/releases/tag/0.2.2, wonder how much it will take to update
ouf...
Don't we have the same problem with https://pypi.org/project/pylink-square/ ?
Maybe
those imports should be wrapped in try: ... except ImportError: and remove them from requirements.txt?
J-Link works with OpenOCD so one can mostly work it around, but I don't think OpenOCD can talk to pemicro stuff.
Don't we have the same problem with https://pypi.org/project/pylink-square/ ?
Why? Does it also bundle shared libraries with obfuscated code from unknown origin? ;-)
Maybe those imports should be wrapped in
try: ... except ImportError:and remove them from requirements.txt?J-Link works with OpenOCD so one can mostly work it around, but I don't think OpenOCD can talk to pemicro stuff.
That's largely what I'm doing for the downstream package: https://github.com/archlinux/svntogit-community/blob/19938f5cab9adf93da26c09ebeb8111ed1bdc59b/trunk/python-spsdk-1.6.0-remove_pypemicro.patch
But given that no pull requests in this repository ever get merged I don't think providing one would make much sense for me (also I don't know what the maintainers think about this change either).
Don't we have the same problem with https://pypi.org/project/pylink-square/ ?
Why? Does it also bundle shared libraries with obfuscated code from unknown origin? ;-)
https://pypi.org/project/pylink-square/
Python interface for the SEGGER J-Link.
Requirements
-
GNU ARM Embedded Toolchain (for functional tests)
maybe it works for you... *
(nxp) m.saper.info> /usr/local/bin/virtualenv-3.8 ~/nxp
Using base prefix '/usr/local'
New python executable in /home/saper/nxp/bin/python3.8
Also creating executable in /home/saper/nxp/bin/python
Installing setuptools, pip, wheel...
done.
(nxp) m.saper.info> pip install pylink-square
Collecting pylink-square
Downloading pylink_square-0.14.2-py2.py3-none-any.whl (82 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 82.3/82.3 kB 6.1 MB/s eta 0:00:00
Collecting psutil>=5.2.2
Downloading psutil-5.9.2.tar.gz (479 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 479.8/479.8 kB 6.5 MB/s eta 0:00:00
Preparing metadata (setup.py) ... done
Collecting six
Using cached six-1.16.0-py2.py3-none-any.whl (11 kB)
Collecting future
Using cached future-0.18.2.tar.gz (829 kB)
Preparing metadata (setup.py) ... done
Building wheels for collected packages: psutil, future
Building wheel for psutil (setup.py) ... done
Created wheel for psutil: filename=psutil-5.9.2-cp38-cp38-freebsd_13_0_release_p11_amd64.whl size=247198 sha256=7d2c122b11af7f441bd1f7265cea95c034cdcd1f893efa77fc1fa7b861cee0ea
Stored in directory: /home/saper/.cache/pip/wheels/c3/d9/d7/b55aee405934fcc854a0d75456d1116bd4a848bdda0a07df46
Building wheel for future (setup.py) ... done
Created wheel for future: filename=future-0.18.2-py3-none-any.whl size=491058 sha256=54a49a1f9fc3561275a181db69c960fa7b9979a946a619b965b360c9587549b5
Stored in directory: /home/saper/.cache/pip/wheels/8e/70/28/3d6ccd6e315f65f245da085482a2e1c7d14b90b30f239e2cf4
Successfully built psutil future
Installing collected packages: six, psutil, future, pylink-square
Successfully installed future-0.18.2 psutil-5.9.2 pylink-square-0.14.2 six-1.16.0
(nxp) m.saper.info> pip install pypemicro
Collecting pypemicro
Using cached pypemicro-0.1.9-py3-none-any.whl (3.7 MB)
Installing collected packages: pypemicro
Successfully installed pypemicro-0.1.9
(nxp) m.saper.info> python
Python 3.8.13 (default, May 10 2022, 01:15:55)
[Clang 11.0.1 ([email protected]:llvm/llvm-project.git llvmorg-11.0.1-0-g43ff75f2c on freebsd13
Type "help", "copyright", "credits" or "license" for more information.
>>> import pylink
>>>
(nxp) m.saper.info> python
Python 3.8.13 (default, May 10 2022, 01:15:55)
[Clang 11.0.1 ([email protected]:llvm/llvm-project.git llvmorg-11.0.1-0-g43ff75f2c on freebsd13
Type "help", "copyright", "credits" or "license" for more information.
>>> import pylink
>>> p = pylink.JLink()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/saper/nxp/lib/python3.8/site-packages/pylink/jlink.py", line 295, in __init__
raise TypeError('Expected to be given a valid DLL.')
TypeError: Expected to be given a valid DLL.
>>> import pypemicro
>>> dir(pypemicro)
['Any', 'CDLL', 'IntEnum', 'PEMicroArmRegisters', 'PEMicroException', 'PEMicroInterfaces', 'PEMicroMemoryAccessResults', 'PEMicroMemoryAccessSize', 'PEMicroPortType', 'PEMicroSpecialFeatures', 'PEMicroSpecialFeaturesSwdStatus', 'PEMicroTransferException', 'PyPemicro', '__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', 'byref', 'c_bool', 'c_byte', 'c_char_p', 'c_ulong', 'c_ushort', 'c_void_p', 'cdll', 'logger', 'logging', 'os', 'pemicro', 'pemicro_const', 'platform', 'sys']
>>> c = pypemicro.PyPemicro()
Traceback (most recent call last):
File "/home/saper/nxp/lib/python3.8/site-packages/pypemicro/pemicro.py", line 446, in get_pemicro_lib
raise PEMicroException("There is any suitable library for this OS.")
pypemicro.pemicro.PEMicroException: There is any suitable library for this OS.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/saper/nxp/lib/python3.8/site-packages/pypemicro/pemicro.py", line 537, in __init__
self.lib = PyPemicro.get_pemicro_lib(dll_path)
File "/home/saper/nxp/lib/python3.8/site-packages/pypemicro/pemicro.py", line 452, in get_pemicro_lib
raise PEMicroException(str(exc))
pypemicro.pemicro.PEMicroException: There is any suitable library for this OS.
>>>
But good news is, those packages do install, they just don't work - I don't see a problem here.
pylink-square installs just fine. If people then require some custom additional non-free blob, that is not my problem as a system's packager. However, if you think that this is problematic it would probably be good to bring this up with the upstream. E.g. the SEGGER downloads are only available behind a EULA.
With pypemicro it is more problematic, as the python package itself vendors a shared library of unknown origin and license (see https://github.com/NXPmicro/pypemicro/issues/10). This is nothing I can even package/redistribute legally as there is no proof whatsoever that it would be allowed.
I see what you mean:
> find nxp/lib/python3.8/site-packages/pypemicro -type f
nxp/lib/python3.8/site-packages/pypemicro/__init__.py
nxp/lib/python3.8/site-packages/pypemicro/pemicro_const.py
nxp/lib/python3.8/site-packages/pypemicro/libs/Linux/unitacmp-64.so
nxp/lib/python3.8/site-packages/pypemicro/libs/MacOS/unitacmp-64.dylib
nxp/lib/python3.8/site-packages/pypemicro/libs/MacOS/libusb.dylib
nxp/lib/python3.8/site-packages/pypemicro/libs/Windows/unitacmp-32.dll
nxp/lib/python3.8/site-packages/pypemicro/libs/Windows/unitacmp-64.dll
nxp/lib/python3.8/site-packages/pypemicro/__pycache__/pemicro_const.cpython-38.pyc
nxp/lib/python3.8/site-packages/pypemicro/__pycache__/pemicro.cpython-38.pyc
nxp/lib/python3.8/site-packages/pypemicro/__pycache__/__init__.cpython-38.pyc
nxp/lib/python3.8/site-packages/pypemicro/pemicro.py
I'd say this is a distribution policy problem. But if you want to disallow this, you have to patch. Maybe maintainers would be open to make both dependencies optional, i.e. removed from requirements.txt and wrapped with ImportError catch.
I'd say this is a distribution policy problem.
I'd go as far as saying that that is a supply chain nightmare. Spsdk is pulled in for projects such as pynitrokey which is used to interact with security tokens that store private key information. There should be no unknown shared lib involved, that can not be accounted for at all.
I was very happy to see that the maintainers changed their mind in regards to libusbsio (see https://github.com/NXPmicro/spsdk/issues/36).
When I was packaging it for myself for Gentoo as a dependency of pynitrokey I decided to delete the PEmicro code as I won't need it. There are two other choices of debug probes (any pyOCD probe, or JLink/pyLink) which are either fully open source or (pyLink) have the proprietary shared library as an optional part the user can install manually. https://github.com/jjakob/gentoo-overlay/blob/master/dev-python/spsdk/files/02-Delete-PEMicro-due-to-dependency-on-non-free-blobs.patch
Note that since 2.0.0 the patch has to be modified a bit to not modify spsdk/debuggers/__init__.py any more:
diff -ruN a/spsdk/debuggers/utils.py b/spsdk/debuggers/utils.py
--- a/spsdk/debuggers/utils.py 2022-02-04 14:27:29.000000000 +0100
+++ b/spsdk/debuggers/utils.py 2022-02-14 00:05:11.017196467 +0100
@@ -15,7 +15,6 @@
from spsdk import SPSDKError
from spsdk.debuggers.debug_probe import DebugProbe, SPSDKDebugProbeError, SPSDKProbeNotFoundError
from spsdk.debuggers.debug_probe_jlink import DebugProbePyLink
-from spsdk.debuggers.debug_probe_pemicro import DebugProbePemicro
# Import all supported debug probe classes
from spsdk.debuggers.debug_probe_pyocd import DebugProbePyOCD
@@ -23,7 +22,6 @@
PROBES = {
"pyocd": DebugProbePyOCD,
"jlink": DebugProbePyLink,
- "pemicro": DebugProbePemicro,
}
logger = logging.getLogger(__name__)
All tests pass successfully with this.
Note that pynitrokey isn't yet compatible with spsdk 2.0.0.
@PureTryOut thanks for the heads up! A new version of pynitrokey supporting spsdk > 2, <2.1 has just been released and adapting the patch comes in handy. :bow:
And with 2.1.0 the patch yet again would have to address spsdk/debuggers/__init__.py.
I'm less and less inclined wanting to package any of this stuff.. :roll_eyes:
This lib still has a dependency on pyocd-pemicro which depends on pypemicro
