libgcc_s.so.1 must be installed for pthread_cancel to work

[Red-Eyed]

How one can actually use this package?
I like flatpack but I still have to sit on snap --classic version of this app.

This package (Pycharm) is completely unusable in flatpak.

I'm a datascience and use a lot of packages: pytorch, tensorflow, numpy, pandas, scikit-learn, opencv, CUDA, etc .

When I want to run my project, I get errors with missing shared libraries that above packages depend on.

Maybe this application can be used for hello words, but I can't make it work for no "pure python" projects.

Is there any workaround for such type of applications (IDE like)
Or I should just give up and use snap --classic ?

PS: Yes, I've found similar issues and it seams that I should give up, but I just want to make it finally clear: can I use IDE in flatpak or not.
Just wonder, how does gnome.Builder works. I guess it uses something like "host-spawn" but didn't find it in the sources.

@TingPing what can you say about this?

Thanks!

[FakeShemp]

Though not the best solution, you can at least ignore venv's and use the system python and libraries.

If there are just some straggler libraries missing, I don't see a problem to adding them in the manifests. And above that, there is always the possibility of extensions.

[Red-Eyed]

Usage of system python is impossible, project dependencies could take up to 10 GB
different projects has it's own set of deps with different versions and cannot intersect.

And yes, if IDE dictates me how to program and organize project: get rid of that IDE.

I mean: fighting with your IDE is a bad idea.

[Red-Eyed]

Regarding extension:
IDE is just like text editor and debug visualizer.
So, python projects works fine w/o any IDE and IDE is optional.

Extensions are just more dependencies, that's awful.

[FakeShemp]

Libraries are dependencies. They need to be available somewhere.

[FakeShemp]

Do you know which package need libgcc_s.so.1 so that I can test it?

[Red-Eyed]

shapely
pip install shapely[vectorized]
version 1.7.0

[Red-Eyed]

I can actually send you traceback:

    from shapely import geometry
  File "/home/redeyed/miniconda3/envs/gen3/lib/python3.7/site-packages/shapely/geometry/__init__.py", line 4, in <module>
    from .base import CAP_STYLE, JOIN_STYLE
  File "/home/redeyed/miniconda3/envs/gen3/lib/python3.7/site-packages/shapely/geometry/base.py", line 18, in <module>
    from shapely.coords import CoordinateSequence
  File "/home/redeyed/miniconda3/envs/gen3/lib/python3.7/site-packages/shapely/coords.py", line 8, in <module>
    from shapely.geos import lgeos
  File "/home/redeyed/miniconda3/envs/gen3/lib/python3.7/site-packages/shapely/geos.py", line 92, in <module>
    free = load_dll('c', fallbacks=c_alt_paths).free
  File "/home/redeyed/miniconda3/envs/gen3/lib/python3.7/site-packages/shapely/geos.py", line 62, in load_dll
    libname, fallbacks or []))
OSError: Could not find lib c or load any of its variants ['libc.musl-x86_64.so.1'].

Process finished with exit code 134 (interrupted by signal 6: SIGABRT)

[FakeShemp]

Thanks :)

[Red-Eyed]

Frankly speaking, the question was not in this particular error. Because this error is just a consequence of sandboxed environment.

But more like: "how one could use flatpak version of PyCharm with access to all system env?"

[FakeShemp]

Yes, I understand that, but I'm curious about it. And I need some issue to start looking at the problem from.

[Red-Eyed]

Got it.

[FakeShemp]

Conda seem to be missing from this manifest. With a new virtualenv I didn't have the same issues as you.

I have no experience with Conda. Does it load libraries differently than venv?

I know there are issues with sandboxing the IDE, I've made issues about it myself, but I'm thinking that they might be extra bad with a Conda environment.

I'll look into adding Conda too the manifest as soon as I can. It should be included anyway.

[Red-Eyed]

Strange error
I've reproduced it and have minimal complete example:
I use python 3.7.4

pip install -U pyqt5
pip install -U "shapely[vectorized]"

# file test.py
from PyQt5 import Qt
from shapely import geometry

python test.py

[Red-Eyed]

I'll look into adding Conda too the manifest as soon as I can. It should be included anyway.

What do u mean by manifest?
U mean to include conda as extension?
If so, it's not required and will be redundant.

I used only pypi and didn't use conda package manager
used it only to install python and create venv

[FakeShemp]

The flatpak manifest. The thing that tells flatpak how to install this app.

Conda would be bundled with this flatpak app by default since it's available for use directly in PyCharm. Right now it just complain that the Conda executable isn't found.

[Red-Eyed]

Whoa, it seems that everything works fine, when I changed order of imports (first shapely than qt)
Even heavy project with 8GB deps works like a charm in flatpak PyCharm.

Taking in mind sandbox, I though that such an error is just one of many possible errors related to sandbox.

[Red-Eyed]

Right now it just complain that the Conda executable isn't found.

So, install conda locally (use this script) and don't have any problems with conda.

And yes, it needs to set python interpreter in PyCharm, then u set conda env and it will see conda executable

sh-5.0$ which conda
/home/redeyed/miniconda3/condabin/conda

like so:

[Red-Eyed]

So now I can finally get rid of snap, as problem is weakly related to flatpak and sandbox
It seems that from PyQt5 import Qt makes shapely "blind" and it cannot find libc

[Red-Eyed]

I'm closing this issue, as problem is not related to flatpak version of PyCharm.

Thanks to @FakeShemp for the help!