Closed mitchellkrogza closed 5 years ago
I've now also tried on a fresh install of Debian using the Conda method and same problem with the Segmentation Fault. Has this repo been abandoned? I hope not.
Hi @mitchellkrogza ...the repo hasn't been abandoned. A lot of our past developments have been experimental in that we need to transition this repo to include Python 3 and Pyside 2 so there haven't been many changes to the Andover branch. I'm in transition between jobs/moving right now and haven't been able to respond to anything posted. It sounds like you're working with the Andover branch, which is our development branch. I'm not sure how much support we can provide for this over the next few weeks, but if you post some detailed information about the individual steps you took to create this error (numpy, pyside, SHARPpy versions), I can probably take a swing at it once I get settled.
Hi @wblumberg great glad to hear this repo will still move ahead. Good luck with the new job and move. I'll post my installation methods tomorrow but believe I am pulling off master branch, didn't look to see if you had other branches.
Should I try the HWT branch? Or stay away?
Hi, @mitchellkrogza. I don't have access to a Debian or Ubuntu box right now, and I'm unable to reproduce this on RHEL 6 (or Mac OS or Windows). Maybe I'll try spinning up a virtual machine later. My instinct is to say that the problem lies in the PySide binaries for those OSes, as Python errors will usually print a stack trace before crashing and/or burning.
Okay, I was able to reproduce this in a Debian VM. I'm not entirely sure what's going on, but it's something to do with the background threads. I've implemented what I think is a fix/workaround in the master-debian branch. Could you try that?
Thanks @tsupinie I pulled from master-debian
switched to that branch and ran python ./setup.py install
again. It still segfaults if selecting Southern Hemisphere before I choose anything else. If I first choose GFS > and select a profile then I can switch to Southern Hemisphere and select a point on the map. But then it's still giving me this.
I did download the Windows 7 64 Bit binary which works as suggested by @georgephillips1 but ultimately I would like this working in Linux. Let me know if there is anything else I can provide to help diagnose this.
Hi @tsupinie I tried this morning with a fresh CentOS 7 VM as it's the closest to RHEL. Here were my install steps and outcome which is much the same. Hope we can find what's causing these errors as the Windows 7 binary works ok.
I used the Conda method as follows:
conda install numpy
Solving environment: done
## Package Plan ##
environment location: /home/mitchellkrog/miniconda2
added / updated specs:
- numpy
The following packages will be downloaded:
package | build
---------------------------|-----------------
conda-4.5.9 | py27_0 1.0 MB
mkl_random-1.0.1 | py27h4414c95_1 361 KB
mkl-2018.0.3 | 1 198.7 MB
mkl_fft-1.0.4 | py27h4414c95_1 147 KB
intel-openmp-2018.0.3 | 0 705 KB
numpy-base-1.15.0 | py27h3dfced4_0 4.1 MB
libgfortran-ng-7.2.0 | hdf63c60_3 1.2 MB
blas-1.0 | mkl 6 KB
numpy-1.15.0 | py27h1b885b7_0 36 KB
------------------------------------------------------------
Total: 206.3 MB
Then
conda install -c conda-forge pyside=1.2.4
Solving environment: done
## Package Plan ##
environment location: /home/mitchellkrog/miniconda2
added / updated specs:
- pyside=1.2.4
The following packages will be downloaded:
package | build
---------------------------|-----------------
libxml2-2.9.8 | h422b904_2 1.8 MB conda-forge
libiconv-1.15 | h470a237_1 2.0 MB conda-forge
qt-4.8.7 | 7 31.4 MB conda-forge
pyside-1.2.4 | py27_8 6.7 MB conda-forge
harfbuzz-1.4.3 | 0 4.0 MB conda-forge
pixman-0.34.0 | 2 1.2 MB conda-forge
freetype-2.7 | 1 2.8 MB conda-forge
libxslt-1.1.32 | h88dbc4e_1 513 KB conda-forge
openssl-1.0.2o | h470a237_1 3.5 MB conda-forge
ca-certificates-2018.4.16 | 0 139 KB conda-forge
icu-58.2 | hfc679d8_0 22.8 MB conda-forge
cairo-1.14.6 | 4 1.2 MB conda-forge
graphite2-1.3.11 | hfc679d8_0 119 KB conda-forge
glib-2.51.4 | 0 6.0 MB conda-forge
pcre-8.41 | h470a237_2 244 KB conda-forge
certifi-2018.4.16 | py27_0 142 KB conda-forge
fontconfig-2.12.1 | 4 938 KB conda-forge
pango-1.40.4 | 0 999 KB conda-forge
xz-5.2.4 | h470a237_0 329 KB conda-forge
libpng-1.6.35 | ha92aebf_0 305 KB conda-forge
libtiff-4.0.9 | he6b73bb_1 521 KB conda-forge
jpeg-9c | h470a237_0 230 KB conda-forge
conda-4.5.9 | py27_0 634 KB conda-forge
gettext-0.19.8.1 | 0 6.3 MB conda-forge
------------------------------------------------------------
Total: 94.8 MB
Then cloned the repo and ran setup, output from setup.py (attached).
Then finally running ./full_gui.py and selecting my GFS ranges and generating profiles.
Oh, yeah, you'll need to downgrade NumPy. NumPy v1.12 works for me. I need to get on and fix bugs when using the latest versions of NumPy because it can't resist breaking SHARPpy whenever they do an update.
Oh awesome I will try revert to Numpy 1.12 tomorrow and will revert back. Thanks so much @tsupinie
Many thanks @tsupinie it works 100% with Numpy 1.12, now for me to start learning how to script with SharpPy 😂
@tsupinie Is there any simple and quick way via the command line to create the basic graph layout already created within the SharpPy gui for a certain station name? I currently have no need to change what is produced simply to try and automate it daily.
For others who may run into this issue, my steps are simply to
conda uninstall numpy
then
conda install numpy=1.12
cd SHARPpy/
python ./setup.py install
cd runsharp/
python ./full_gui.py
Tested on CentOS7 ✔️ Ubuntu 18.04 LTS ✔️ (see below) Debian 9 ✔️ (same install as below for Ubuntu)
@tsupinie Happy to also report I have this working on Ubuntu 18.04 LTS My install steps.
bash Miniconda2-latest-Linux-x86_64.sh
conda install -c conda-forge pyside=1.2.4
conda install numpy=1.12
git clone https://github.com/sharppy/SHARPpy.git
cd SHARPpy/
git pull origin master-debian
git checkout master-debian
python ./setup.py install
cd runsharp/
python ./full_gui.py
Initial 3-5 attempts to start full_gui.py
might give you NameError: global name 'socket' is not defined
keep trying to start it, it eventually fires up and generates the graphs needed.
Save Image Dialog Box is horribly messed up though. (on both Ubuntu and Debian)
Error message regarding socket
Traceback (most recent call last):
File "./full_gui.py", line 54, in doCrasher
ret = func(*args, **kwargs)
File "./full_gui.py", line 661, in createWindow
return Main()
File "./full_gui.py", line 537, in __init__
self.__initUI()
File "./full_gui.py", line 543, in __initUI
self.picker = Picker(self.config, parent=self)
File "./full_gui.py", line 110, in __init__
urls = data_source.pingURLs(self.data_sources)
File "/home/user/miniconda2/lib/python2.7/site-packages/SHARPpy-1.3.0-py2.7.egg/datasources/data_source.py", line 70, in pingURLs
urls[url] = _pingURL(url)
File "/home/user/miniconda2/lib/python2.7/site-packages/SHARPpy-1.3.0-py2.7.egg/datasources/data_source.py", line 54, in _pingURL
except socket.timeout as e:
NameError: global name 'socket' is not defined
On Ubuntu 18.04 LTS when I select Southern Hemisphere there is an immediate segmentation fault and
full_gui.py
exits