After installing ipython/jupyter cadquery crashes on macos under Python 3.6 #78
Comments
|
Not sure if this is related, but I see this error in the CQ-editor console when running it with the latest updates on MacOS. |
|
@bernhard-42 could you also paste a backtrace from the crash? |
|
Unfortunately, there is nothing more than the three lines above from "Fatal" to "trap: 6" However, I found out that when you reinstall cadquery with the It looks like the issue comes from python (whether it is from anaconda or from python-forge). Steps to reproduce:
We now have anaconda python and conda-forge python is removed. Result: cadquery crashes
Now we get conda-forge python again, and caqdquery works again. This is in line with conda-forge/gdcm-feedstock#8 (comment) What I don't know is whether the build process of conda-forge is different or version 3.6.6 vs. 3.6.8 (and I have no idea how to check it) |
|
@bernhard-42 could you generate a core dump and inspect it. Roughly like this: |
|
Here you are: |
|
could it have anything to do with tpaviot/pythonocc-core#541 ? |
|
@bernhard-42 thanks, yes it has to be it. I uploaded the latest build of pythonocc |
|
Well, now it doesn't work any more, even if you install python-cadquery again. Looks like another build of python 3.6.6 is in conda-forge (Oct 2018 instead July 2018). I now have to do a to get rid of python 3.6.6 and use python 3.6.5 Something is wrong in 3.6.6 |
|
@bernhard-42 well I assume it is this static linking thing. Could you try explicitly with the latest build (NB: an experimental build from "our" channel): |
|
Not too much of a success: |
|
@bernhard-42 I made a mistake. It is |
|
I looked at So the full install is:
A good sign is that conda does not try to install a python interpreter from conda-forge any more. Trying to install Even installing ipython afterwards keeps OCC working. I guess we now need a cadquery version working with occ 0.18.2 to finalize the test |
|
@bernhard-42 great! Indeed working OCC import proves it is correct solution (cf. the backtrace above - error occurs in pythonocc). You could install cadquery using pip for testing: |
|
works fine! btw. the syntax I had to use is |
|
I guess we can close the issue now? |
|
Great! Let's wait until this is in master branch.
…On Sat, Jan 19, 2019, 11:13 AM Bernhard ***@***.*** wrote:
I guess we can close the issue now?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#78 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ANVXYg8sdYDF9XoWCRQS48AVMlEE2aAjks5vEu_agaJpZM4ZplFB>
.
|
|
Well, now I wanted to work with the new version, and I can import cadquery (and this is what I tested before), however building models doesn't work: never comes back with a result. I gets stuck beyond interrupt (had to kill -9 it since Ctrl-C didn't work) |
|
Too bad. Are you able to run this in a debugger (like |
|
Without gdb, it fails in |
|
gdb not sufficiently code-signed on my mac, what so ever this means. No idea how to get that running. I'm a python, javascript and scala guy, gcc and gdb are strangers to me ;-) |
|
Maybe lldb then? I don't have Mac experience...
…On Sat, Jan 19, 2019, 12:04 PM Bernhard ***@***.*** wrote:
gdb not sufficiently code-signed on my mac, what so ever this means. No
idea how to get that running. I'm a python, javascript and scala guy, gcc
and gdb are strangers to me ;-)
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#78 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ANVXYlmdgEuBBXQ7_Er58U_gh4IFyWcoks5vEvusgaJpZM4ZplFB>
.
|
|
ok, If I interrupt the debugger after the fillet function has been called, it ends up in very different routines and modules. The python process goes to 100%.
The test script is and my "Ctrl-C cont" attempts start after "Now" is printed Does this help? |
|
This is the current threads call stack |
|
@bernhard-42 If you vary your fillet radius between 0.5 and 0.9 does it help at all? Also, what happens if you comment the print statements out and move the fillet call up to the box2 line? I ask because I get weird results in CQ-editor when including the print statements. |
|
Switched to Same result |
|
note, I increase the box by factor 10 to avoid fillet issues with too large values compared to the solid it is applied to |
|
I mean, I should have done However still fails. A few frames, copied after Ctrl-C and then continued |
|
This is the minimum failing example since works fine in the new cadquery version. Note: the fillet version works as expected with the conda forge cadquery version
|
|
Digging Deeper: delivers: So the call |
|
btw. the fillet function was taken from |
|
@bernhard-42 thanks, clear how to reproduce. For completeness could you paste a full backtrace (just one) from your lldb experiments? |
|
what different to the callstack above created with |
|
@bernhard-42 it is exactly what I meant, but would be good to have all frames to have a better picture. BTW: I requested free MacOS workers from CircleCI. If I understand correctly they allow ssh access for debugging builds. |
|
There is not much more. Here another one And another one after "cont" and Ctrl-C |
|
@bernhard-42 I think we sorted it out. Please look at PR #57 . The secret was different compilation options for |
|
#57 is now merged to master. All CI tests pass including those on Mac. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Somehow the installation of ipython kills cadquery:
I use miniconda as base with conda 4.5.2
The text was updated successfully, but these errors were encountered: