Cannot connect to jupyter/ipython instance

[jzuhone]

When attempting to run a jupyter notebook or qtconsole, I get error messages like this:

Invalid argument (bundled/zeromq/src/tcp_address.cpp:171)

The normal ipython prompt works fine.

[sunilmut]

Thanks for trying this out. Could you collect a strace for the failing command to what might be failing?

[chrisgorgo]

I can confirm this bug. You can find the strace here: https://gist.github.com/chrisfilo/2dd2c47592423291bf16db6348d26c6a

[chrisgorgo]

It seems that this bug is caused by jupyter/notebook#1331

[arthurhjorth]

Trying to get the same thing to run too, and with the same result. I'm on a Surface 3 4GB/128GB and happy to aid with testing/pinning things down. Jupyter is such an excellent tool and being able to run this on Bash on Windows would be amazing.

[guydepauw]

Same setup as @arthurhjorth here and also happy to help out. Loving bash on windows so far. Getting jupyter up and running would be the icing on the cake.

[aseering]

I came up with a workaround based on @MadcowD 's investigation on jupyter/notebook#1331 . I submitted it upstream as zeromq/libzmq#2013 , and posted an updated Ubuntu 14.04 package to a PPA in case anyone would like to try to use it.

Standard disclaimers -- I make no guarantees about either security or correctness; if it eats your data or steps on your cat I take no responsibility. But it seems to work for me.

Of course, getting a fix into WSL would be ideal. My cursory understanding of this code is that WSL would need to implement getifaddrs(), and would I think need to at minimum report on the existence of a loopback interface.

[sgraphics]

@aseering can you please explain how to use this fix? I can 'make install' for libzmq from repo source (including this fix) but how do I setup Jupyter to use the custom libzmq build? - it is still giving the same error that is quoted in the original post. Thank you!

[aseering]

@sgraphics are you familiar with Ubuntu PPAs? I would suggest using the PPA that I created that contains the fix. The simplest way to do so would be, I believe:

sudo add-apt-repository ppa:aseering/wsl
sudo apt-get update
sudo apt-get install libzmq3

That last command would normally install Ubuntu's libzmq; but because you've added my PPA containing the updated package, it will get the updated package instead.

If you want to inspect the source code, you're welcome to download it from the PPA's website (see the link above). You can also follow the first two steps above, then run apt-get source libzmq3, which will download the source code for the package to the current directory. Note that PPAs can contain updated versions of any package; you should only add PPAs that you trust.

Unfortunately, make install will not work in this case: The upstream version that contains the fix is also too new to work on Ubuntu 14.04. The PPA contains the same fix but applied to the source code used by the official Ubuntu 14.04 package. (I downloaded the Ubuntu source using apt-get source, then used quilt to apply the diff containing my fix, and uploaded the updated package.)

[guydepauw]

@aseering this works perfectly. Thanks a million!

[pulinagrawal]

@aseering
I am getting an error 'Invalid argument' in line 190 of tcp_address.cpp even after updating the libzmq3 as you described above.
I couldn't build from source because it was asking for source URI when I was trying to download source.

[jzuhone]

I had the same issue as @pulinagrawal.

On Jul 13, 2016, at 11:22 AM, Pulin Agrawal notifications@github.com wrote:

@aseering

I am getting an error 'Invalid argument' in line 190 of tcp_address.cpp even after updating the libzmq3 as you described above.
I couldn't build from source because it was asking for source URI when I was trying to download source.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or mute the thread.

[aseering]

@pulinagrawal @jzuhone -- hm... Could you post the output of dpkg -l libzmq3? It should indicate that you have version 4.0.4+dfsg-2wsl1 installed. (The wsl1 at the end is the key.)

[aseering]

I've posted a bug report about this issue to the Ubuntu libzmq3 bugtracker:

https://bugs.launchpad.net/ubuntu/+source/zeromq3/+bug/1602900

The fix is known; hopefully they will agree to backport it so that WSL users don't have to keep doing this dance to install packages from a PPA.

[mrsiamen]

this fix doesn't work for my Surface 3...
I can still see the message
Invalid argument (bundled/zeromq/src/tcp_address.cpp:171)

Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name Version Architecture Description
+++-========================-=================-=================-=====================================================
ii libzmq3:amd64 4.0.4+dfsg-2wsl1 amd64 lightweight messaging kernel (shared library)

[aseering]

@mrsiamen -- it sounds like you have the new package but it's not getting used for some reason. Is this a clean install of WSL? Did you try installing libzmq3 yourself before using the PPA?

Could you post the results of the following commands?:

ldd /usr/lib/python2.7/dist-packages/zmq/backend/cython/utils.x86_64-linux-gnu.so | grep zmq
md5sum /usr/lib/x86_64-linux-gnu/libzmq.so.3.1.0

For reference, on my machine, I see:

$ ldd /usr/lib/python2.7/dist-packages/zmq/backend/cython/utils.x86_64-linux-gnu.so | grep zmq
        libzmq.so.3 => /usr/lib/x86_64-linux-gnu/libzmq.so.3 (0x00007f72cc2a0000)
$ md5sum /usr/lib/x86_64-linux-gnu/libzmq.so.3
87b97738b28a38a8e89796ecd41862b0  /usr/lib/x86_64-linux-gnu/libzmq.so.3.1.0

[pulinagrawal]

@aseering

on dpkg on libzmq3 I do have wsl1 at the end

ldd gives
ldd: /usr/lib/python2.7/dist-packages/zmq/backend/cython/utils.x86_64-linux-gnu.so: No such file or directory

md5sum is the same as your output

[aseering]

@pulinagrawal -- Sounds like you're not using Ubuntu's iPython package? This fix only fixes Ubuntu's libzmq package. If you're using an iPython that pulls in libzmq from elsewhere, they will need to apply the fix as well.

To be specific, I'm testing as follows:

• Uninstall and fully remove any third-party ipython versions that you may have installed
• Run the command sudo apt-get install ipython
• Once ipython has successfully installed, run ipython notebook --no-browser. It should print out a URL. Open a browser in Windows; navigate to that URL.

If you want to track this down further, you could run a command like

find /usr $HOME -type d -name zmq | grep -v IPython

to try to figure out the actual path to the Python zmq module that you're running. Then update the first command above with the appropriate path and re-run ldd; that should tell you where the third-party libzmq that you're using lives.

Another approach, just a guess: Are you getting ipython through Anaconda? I don't personally use Anaconda on Linux (I do on Mac/Windows) because it tends to cause this sort of confusion, and because there's less need for it -- apt-get install and pip install work well on their own.

[mrsiamen]

@aseering Thanks for keep tracking on this:). looks like I'm facing the same issue as pulinagrawal does.
Will try to confirm the path.

[mrsiamen]

I try to install the fixed libzmq3 before installing ipython notebook and it works now.
the kernel doesn't crash anymore. thank you guy for helping on this :)

libzmq.so.3 => /usr/lib/x86_64-linux-gnu/libzmq.so.3 (0x00007fb3722e0000)
87b97738b28a38a8e89796ecd41862b0 /usr/lib/x86_64-linux-gnu/libzmq.so.3.1.0

[tanguita]

@aseering Could you possibly package this for conda??? unfortunately I have to use a python distribution that is through conda (specific not available packages). Or prehaps tell me how to link it?

[pulinagrawal]

@aseering @mrsiamen

Thank You for the help!! It is fixed now.

I uninstalled ipython and jupyter. I removed the zmq folder from dist-packages.
I installed libzmq3 then installed ipython. It did not run yet because I guess removing zmq removed other stuff as well. So, I installed python-zmq. Now it is running.

I might have broken other stuff by just deleting the zmq folder. So I guess better method would have been to use uninstall on python-zmq. But I did not know the name of the package then.

But it works for now.

[aseering]

@tanguita , sorry for the delay -- I believe Anaconda manages their own packages so they would have to incorporate this patch themselves. Feel free to go to Anaconda's bug tracker and file a bug there. If they have questions, I'd be happy to help them with the process.

If you want to get the source code for the patched libzmq, first run the commands above to get the updated package; then run

apt-get source libzmq

This will download the source package from my PPA. You can "cd" into it, read zmq's documentation, and see how to build it.

Alternatively, you can just install the binary package and take the pre-built libzmq that it installs, and try replacing Anaconda's libzmq with Ubuntu's. But that will only work if the two are the same or similar versions; I haven't tested it myself.

[jzuhone]

@aseering, This is not strictly true--one can create a package for Anaconda and upload it to one's own conda channel:

https://conda.anaconda.org/

Then others can install it in their conda distribution.

[tanguita]

This is what I thought but do not know how to do it..

On 21/07/16 11:25, John ZuHone wrote:

@aseering https://github.com/aseering, This is not strictly
true--one can create a package for Anaconda and upload it to one's own
conda channel:

https://conda.anaconda.org/

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#185 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AID11O3GdlLali0cfqZHhW1dxThfSg9mks5qX49tgaJpZM4IG4pi.

[aseering]

Ah, thanks @jzuhone , I didn't realize people can create their own packages. I'll look into that, but I'm a bit busy these days and I'm not a big Anaconda user so I'd have a bunch to learn...

The actual code change to libzmq is quite small; you can see it here. If anyone here knows more about Anaconda packaging, feel free to apply that patch and add it to a channel.

[aseering]

Hi @arthurhjorth -- what arguments are you running pip with?

sudo -H will not, by itself, cause things to be installed to your home directory. sudo (by default) runs commands as the root user rather than as your normal user account; root has elevated privileges to install things systemwide. sudo -H says, even though you're running as the root user, set the $HOME environment variable to point at your regular user's home directory. This affects programs that try to write to "the current user's home directory".

apt-get always installs software systemwide. (It's the Ubuntu package manager; it installs packages for your Ubuntu system.) Therefore, by default it must be run as root, and the value of the $HOME environment variable (and therefore the -H flag to sudo) doesn't affect it.

pip (and pip3) can install either systemwide or just for your current user account. By default, it decides where to install based on whether you run pip with the --user flag: With that flag, it will install to your home directory; without that flag, it will install systemwide. If you install systemwide, by default it will need to run as root. If you install to your home directory, you should not need to install as root -- this is very useful for users on shared machines who don't have sudo access.

There are a few weird edge cases to consider:

• If you run sudo pip install --user <some package>: sudo means "run the following command as root". pip install --user means "install to the current user's home directory. But the current user is root. So this will install the package to root's home directory, and it will only be available if you run as root. Which is probably not useful, unless you really know what you are doing.
• If you run sudo -H pip install --user <some package>: This is the same as above, except, you're telling pip that root's home directory is your current directory. That's actually bad: It means that pip will install the package to your home directory, but your user's package repository will be corrupted / updated to be owned by root. By default, this means that you can run packages, but the next time you try to install using pip without sudo, you won't be allowed to do so. You can fix this by running sudo chown -R $USER ~/.local.
• What happens if you install via both apt-get and pip? Some packages will conflict. In this case, apt-get, which is a general-purpose package manager and has only a minimal understanding of Python in particular, will simply error out and say (I'm paraphrasing here) "go fix your package!" pip is a little more clever; it will try to resolve the dependency in favor of whatever version it is installing. However, most packages won't conflict. This is because Python has a "search path": When you go to import a package in a Python script, first, Python checks for packages installed to your user account's home directory (where pip install --user puts them): $HOME/.local/lib/python2.7. If it doesn't find a package there, next it looks in /usr/local/lib/python2.7 (or similar for Python 3), which is where sudo pip install puts them. If it still doesn't find the package, then it looks in /usr/lib/python2.7, which is where apt-get installs its packages. If it still doesn't find the package, only then does it error out. So if you install a package multiple times using different tools or commands, Python will pick whichever it sees first. This can cause a lot of confusion -- for example, if you install using pip, then you decide you want the official Ubuntu version so you install using apt-get, you'll keep getting the pip version because it's first in the search path. Also, if you have a broken version of a package early in the search path that won't uninstall for whatever reason, installing a correct package may not fix it if the correct one is installed later in the search path.

All of the above assumes that you're using the standard version of Python packaged for Ubuntu Linux (which is what WSL uses). If you're running on CentOS, or Mac, or on Ubuntu using an alternative version of Python such as Anaconda or Linuxbrew, or if you compile and install Python yourself from source, then at minimum the number and location of paths in Python's search path will change; also, some of those installations will reconfigure your system to not require root access to install Python packages systemwide.

[gencee]

@jzuhone Hey man thks for the fix. I tried to fix this for days now and your package made it work like a charm! I appreciate it!

[spoonsso]

Sorry, found some issues with the above. Try:
conda install -c jzuhone zeromq=4.1.dev0
instead

Aye, also just saved by @jzuhone. For anyone working from an anaconda install, his package totally works. Thanks!

[jzuhone]

Hi @gencee and @spoonsso,

Thanks for the appreciation--but my package should be totally unnecessary now if you update to the latest Windows 10 Fast Ring update, as Microsoft has fixed the underlying issue.

[fedorzh]

Hey, @aseering I'm getting
E: Unable to locate package libzmq3
for some reason. Is it because I use Ubuntu 16.04?
W: The repository 'http://ppa.launchpad.net/aseering/wsl/ubuntu xenial Release' does not have a Release file.
Any ideas?

[fedorzh]

I reinstalled 14.04 from scratch, but still getting
Invalid argument (bundled/zeromq/src/tcp_address.cpp:199)
after using fix from @aseering

[aseering]

Hi @fedorzh -- for what it's worth, an alternative fix is bundled into the latest Insider build. So if you're willing to run an Insider build (either Fast or Slow Ring is new enough), then you no longer need my fix. (Also, Insider builds have native official support for Ubuntu 16.04, as well as many other WSL fixes and improvements.)

If you are staying with regular Windows, though -- how did you install iPython Notebook? My fix only works if you installed it via sudo apt-get install ipython-notebook. If you have ever installed it any other way, you must fully uninstall it (which can sometimes be tricky), then add my repository, then reinstall it via apt-get.

If you need to use a version of Notebook that's installed from some other source, others on this thread (for example, @jzuhone ) have created packages for other packaging systems.

[fedorzh]

@aseering
Yes, I tried to upgrade to 16.04 to see if it helps with other issues, but it only becomes more unstable, so decided to go back to 14.04. I can't move to Insider builds unfortunately due to company policy

I install jupyter with
sudo pip3 install jupyter.
Would it work if you provide host to get, compile your fix manually, and install the compiled version (in this case could you please provide some direction)?

[aseering]

@fedorzh -- hm... You could try installing the libzmq-dev package through apt-get; then uninstalling jupyter through pip and reinstalling it as sudo pip3 install --no-binary :all: jupyter. I may have the syntax wrong -- basically, you want to force pip to compile jupyter notebook from source, rather than using pip's precompiled binary. I believe its build scripts will notice that your system already has libzmq, and will use the (fixed) system version rather than the (broken) version bundled in pip's upstream repository.

[aseering]

Also -- if you're installing anything from source on Ubuntu (including the command that I recommended above), first you want to run sudo apt-get install build-essential. It installs a compiler and a bunch of standard build tools that are usually needed to build things from source.

[nimasasa]

Hi,
Does anybody know how we can use matplotlib (inline) on Jupyter? My kernel dies when I am using this package. I have also tried Xming and it didn't work.

Nima

[sunilmut]

@nimasasa - Thanks for your post.

My kernel dies...

Which kernel are you referring to? Is your setup on native Linux or WSL/bash?

[nimasasa]

@sunilmut I ma using WSL. I know it does not support graphical applications officially. But I was able to use Xming. Now, my problem is I can have anaconda and Jupyter on my chrome to run my codes, but when I am trying to plot something, my python kernel dies.

[sunilmut]

@nimasasa - Thanks for confirming. I was worried that you might be referring to Windows kernel dying :). I would recommend starting a new issue/post for this, using the template. A strace would be useful too.

[ggada]

Thanks @jzuhone!

conda install -c jzuhone zeromq=dev fixed this for me. The apt-get one didn't work probably because I am using conda environment.

Not going for the fast ring because I've faced BSOD's before.. ;)

[ericweijieli]

Thanks @jzuhone!
→ conda install -c jzuhone zeromq=4.1.dev0 works for me!

[ahsteven]

How to get tensorflow gpu support? I have cuda and tensorflow-gpu installed on Windows but not on Windows bash. Any suggestions on using the GPU with GYM on windows Bash.

[Yangff]

@ahsteven it's impossible to run tensorflow with gpu on WSL.

[WSLUser]

@Yangff Did you try using VcxSrv to run it with 17063? (I'm unable to go Insiders to test myself). You can now run VSCode from WSL so I wouldn't be surprised if other GUI dev tools are able to work now too. It will be interesting to see what now works and what still doesn't and what is preventing them at this point from running.

[ahsteven]

No, I have not. I plan to setup a dual boot soon so I will just leave it for now.

…

On Wed, Dec 20, 2017 at 1:39 PM, DarthSpock ***@***.***> wrote: @Yangff <https://github.com/yangff> Did you try using VcSrv to run it with 17063? (I'm unable to go Insiders to test myself). You can now run VSCode from WSL so I wouldn't be surprised if other GUI dev tools are able to work now too. It will be interesting to see what now works and what still doesn't and what is preventing them at this point from running. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#185 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJsMv4O3uSnDRRU-VEPSw3PjIFubJEDSks5tCVRggaJpZM4IG4pi> .

[phobson]

Despite @jzuhone's herculean efforts on his build of zeromq, my WSL/jupyter installation isn't working with matplotlib.

For example, a notebook consisting simply of:

from matplotlib import pyplot
pyplot.plot([1, 2, 3], [3, 1, 2], 'ko')

Kills the kernel.

Is this still expected on non-Insider builds of Windows 10? Should bug my IT department to let me into the Insider program?

Output of `conda list`

paul@POR-PHOBSON10:/mnt/c/Users/phobson$ conda list
# packages in environment at /home/paul/miniconda:
#
_nb_ext_conf              0.4.0                    py35_1
anaconda-client           1.6.6            py35h6b90917_0
asn1crypto                0.23.0           py35h4ab26a5_0
beautifulsoup4            4.6.0            py35h442a8c9_1
bleach                    2.1.1            py35hd5e75dd_0
ca-certificates           2017.08.26           h1d4fec5_0
certifi                   2017.11.5        py35h9749603_0
cffi                      1.11.2           py35hc7b2db7_0
chardet                   3.0.4            py35hb6e9ddf_1
clyent                    1.2.2            py35h491ffcb_1
conda                     4.4.7                    py35_0
conda-build               3.2.1                    py35_0
conda-env                 2.6.0                h36134e3_1
conda-verify              2.0.0            py35he912c7b_0
cryptography              2.1.4            py35hbeb2da1_0
cycler                    0.10.0           py35hc4d5149_0
dbus                      1.10.22              h3b5a359_0
decorator                 4.1.2            py35h3a268aa_0
entrypoints               0.2.3            py35h48174a2_2
expat                     2.2.5                he0dffb1_0
filelock                  2.0.13           py35h4160c53_0
fontconfig                2.12.4               h88586e7_1
freetype                  2.8                  hab7d2ae_1
glib                      2.53.6               h5d9569c_2
glob2                     0.6              py35hff66265_0
gmp                       6.1.2                h6c8ec71_1
gst-plugins-base          1.12.2               he3457e5_0
gstreamer                 1.12.2               h4f93127_0
html5lib                  1.0.1            py35h2f9c1c0_0
icu                       58.2                 h9c2bf20_1
idna                      2.6              py35h8605a33_1
intel-openmp              2018.0.0             hc7b2577_8
ipykernel                 4.7.0            py35h2f9c1c0_0
ipython                   6.2.1            py35hd850d2a_1
ipython_genutils          0.2.0            py35hc9e07d0_0
ipywidgets                7.0.5            py35h8147dc1_0
jedi                      0.11.0                   py35_2
jinja2                    2.10             py35h480ab6d_0
jpeg                      9b                   h024ee3a_2
jsonschema                2.6.0            py35h4395190_0
jupyter                   1.0.0            py35hd38625c_0
jupyter_client            5.1.0            py35h2bff583_0
jupyter_console           5.2.0            py35h4044a63_1
jupyter_core              4.4.0            py35ha89e94b_0
libedit                   3.1                  heed3624_0
libffi                    3.2.1                hd88cf55_4
libgcc                    7.2.0                h69d50b8_2
libgcc-ng                 7.2.0                h7cc24e2_2
libgfortran-ng            7.2.0                h9f7466a_2
libiconv                  1.15                 h63c8f33_5
libpng                    1.6.32               hbd3595f_4
libsodium                 1.0.15               hf101ebd_0
libstdcxx-ng              7.2.0                h7a57d05_2
libxcb                    1.12                 hcd93eb1_4
libxml2                   2.9.4                h2e8b1d7_6
markupsafe                1.0              py35h4f4fcf6_1
matplotlib                2.1.1            py35ha26af80_0
mistune                   0.8.1            py35h9251d8c_0
mkl                       2018.0.1             h19d6760_4
nb_anacondacloud          1.4.0                    py35_0
nb_conda                  2.2.1            py35hccc8299_0
nb_conda_kernels          2.1.0                    py35_0
nbconvert                 5.3.1            py35hc5194e3_0
nbformat                  4.4.0            py35h12e6e07_0
nbpresent                 3.0.2            py35h9c03491_1
ncurses                   6.0                  h9df7e31_2
notebook                  5.2.2            py35he644770_0
numpy                     1.14.0           py35h3dfced4_0
openssl                   1.0.2n               hb7f436b_0
packaging                 16.8             py35h2260b46_1
pandoc                    1.19.2.1             hea2e7c5_1
pandocfilters             1.4.2            py35h1565a15_1
parso                     0.1.1            py35h1b200a3_0
patchelf                  0.9                  hf79760b_2
path.py                   10.5             py35h39c98c1_0
pcre                      8.41                 hc27e229_1
pexpect                   4.3.0            py35hf410859_0
pickleshare               0.7.4            py35hd57304d_0
pip                       9.0.1            py35h7e7da9d_4
pkginfo                   1.4.1            py35hb526c55_1
prompt_toolkit            1.0.15           py35hc09de7a_0
ptyprocess                0.5.2            py35h38ce0a3_0
pyasn1                    0.4.2            py35h8f845f4_0
pycosat                   0.6.3            py35h6b6bb97_0
pycparser                 2.18             py35h61b3040_1
pycrypto                  2.6.1            py35h9852db7_1
pygments                  2.2.0            py35h0f41973_0
pyopenssl                 17.5.0           py35h4f8b8c8_0
pyparsing                 2.2.0            py35h041ed72_1
pyqt                      5.6.0            py35h0e41ada_5
pysocks                   1.6.7            py35h6aefbb0_1
python                    3.5.4               h417fded_24
python-dateutil           2.6.1            py35h90d5b31_1
pytz                      2017.3           py35hb13c558_0
pyyaml                    3.12             py35h46ef4ae_1
pyzmq                     16.0.2                   py35_0
qt                        5.6.2               h974d657_12
qtconsole                 4.3.1            py35h4626a06_0
readline                  7.0                  ha6073c6_4
requests                  2.18.4           py35hb9e6ad1_1
ruamel_yaml               0.11.14          py35h8e2c16b_2
setuptools                36.5.0           py35ha8c1747_0
simplegeneric             0.8.1            py35h2ec4104_0
sip                       4.18.1           py35h9eaea60_2
six                       1.11.0           py35h423b573_1
sqlite                    3.20.1               hb898158_2
terminado                 0.6              py35hce234ed_0
testpath                  0.3.1            py35had42eaf_0
tk                        8.6.7                hc745277_3
tornado                   4.5.2            py35hf879e1d_0
traitlets                 4.3.2            py35ha522a97_0
urllib3                   1.22             py35h2ab6e29_0
wcwidth                   0.1.7            py35hcd08066_0
webencodings              0.5.1            py35hb6cf162_1
wheel                     0.30.0           py35hd3883cf_1
widgetsnbextension        3.0.8            py35h84cb72a_0
xz                        5.2.3                h55aa19d_2
yaml                      0.1.7                had09818_2
zeromq                    4.1.dev0                      0    jzuhone
zlib                      1.2.11               ha838bed_2

[sunilmut]

@phobson - Can you clarify what you mean by Kilss the kernel.? Are you getting a blue (or green) screen of death? i.e. Is your Windows crashing? If so, can you please follow the instructions for BSOD here?

Also, which version of Windows are you running. From your WSL bash prompt, output of cmd.exe /c ver

[phobson]

@sunilmut

Sorry I wasn't clear. I meant that trying to plot in a jupyter notebook with matplotlib was killing the ipython kernel. The windows/WSL/linux kernels were fine.

I was just digging back into this and working with my IT department. I'm not sure which if these incantations did the trick, but we upgraded my windows build to 1709 and then uninstalled/reinstalled WSL/Ubuntu via lxrun and and it worked!

🤷‍♂️

[sunilmut]

@phobson - Thanks for the details. Good to know your issue was resolved. There is significant investment going into WSL. So, staying on the latest Windows release will always give you the most mileage, at least that's what we aim for.

[therealkenc]

The OP here was:

Invalid argument (bundled/zeromq/src/tcp_address.cpp:171)

Which was:

    //  Get the addresses.
    ifaddrs *ifa = NULL;
    const int rc = getifaddrs (&ifa);
    errno_assert (rc == 0);
    zmq_assert (ifa != NULL);

Which is #308 addressed in 15007 and Creators Update 15063.