Strange Effect when using python-seabreeze in combination with gtk

[sdickreuter]

I'm not sure if this would better fit to the bugtracker of gtk, but maybe someone else using python-seabreeze can check if the problem exist on other computers too, and it's not a problem of my configuration.
When calling the following code with python 3.4, wrong values for the wavelength are returned:

from gi.repository import Gtk
import seabreeze.spectrometers as sb

devices = sb.list_devices()
s = sb.Spectrometer(devices[0])
print(s.intensities())
print(s.wavelengths())

When commenting out from gi.repository import Gtk, the right values are reported.
In both cases the intensities are correct.
I can't get my head around why gtk might interfere with seabreeze ...

[ap--]

This is very weird...
Does this happen for both backends?
Please provide the printed output.

[ap--]

Please do:

import seabreeze.spectrometers as sb
spec = sb.Spectrometer.from_serial_number()
for i in range(1, 5):
    print(spec.read_eeprom_slot(i))

with and without the Gtk import.

[sdickreuter]

I also testet the pyseabreeze backend and the problem happens only with the cseabreeze backend.
The scrambled wavelength vector looks like this:

[ -6.10800000e+03  -4.42500000e+03  -3.07200000e+03 ...,   6.57228806e+09
   6.59143373e+09   6.61061655e+09]

Running your code gives to following result:

AttributeError: 'Spectrometer' object has no attribute 'read_eeprom_slot'

[ap--]

Can you test if the newest commit on master fixes this problem?

[sdickreuter]

Unfortunately the result is the same:

[ -6.10800000e+03  -4.42500000e+03  -3.07200000e+03 ...,   6.57228806e+09
   6.59143373e+09   6.61061655e+09]

A workaround would be to import gtk after s = sb.Spectrometer(devices[0]), then everything works fine

[ap--]

Can you still reproduce this issue?
I tried, and I can't anymore.
I tried going back to old libseabreeze versions, python-seabreeze versions and python interpreters.

So my guess is that it might have been a gtk3 bug, that is fixed now? Since this is the only thing I didn't test.

[sdickreuter]

At the moment I use seabreeze under windows with qt instead of gtk which works fine.
I just tried to get the latest version running under linux to test with gtk, but unfortunately I get the following error:

`>>> import seabreeze.spectrometers as sb

Traceback (most recent call last):
File "/home/dunkelfeld/PyCharmProjects/python-seabreeze/seabreeze/backends.py", line 30, in _use_cseabreeze
import seabreeze.cseabreeze as sbb
File "/home/dunkelfeld/PyCharmProjects/python-seabreeze/seabreeze/cseabreeze/init.py", line 5, in from .wrapper import (SeaBreezeError,
ImportError: No module named 'seabreeze.cseabreeze.wrapper'
`

Now I'm a bit confused, do you know why this error occurs ?

[ap--]

Did you install the module properly?

This happens, if you install the module via python setup.py install and try to run it from the same path. In your case: /home/dunkelfeld/PyCharmProjects/python-seabreeze

Also you closed the issue. Does this mean you also can't reproduce the bug anymore?

[sdickreuter]

Thanks for the tip, that was actually the Problem, the new version works perfectly in a different directory.
Unfortunately the bug is still present and the wavelength vector still has the wrong values when importing gtk3 (version 3.18.7-1). Oddly enough, when importing gtk2 from gi.repository everything is fine.
Also sorry for closing the issue, I just clicked the wrong button.

[ap--]

Can you run the following code and post the output:

from gi.repository import Gtk, GObject

import seabreeze.cseabreeze as cseabreeze
import numpy
import sys

print ( sys.version )
print ( cseabreeze.__file__ )
print ( Gtk._version )
print ( GObject.pygobject_version )

device, = cseabreeze.wrapper.device_list_devices()
cseabreeze.wrapper.device_open(device)
specID, = cseabreeze.wrapper.device_get_spectrometer_feature_id(device)
speclen = cseabreeze.wrapper.spectrometer_get_formatted_spectrum_length(device, specID)

out = numpy.empty((speclen,), numpy.double)
cseabreeze.wrapper.spectrometer_get_wavelengths(device, specID, out)

print ( "First ten wavelengths:" )
print ( out[:10] )

Thanks

[ap--]

And also with the Gtk, GObject import after the sys import. :) Thanks.

[sdickreuter]

No problem,
Gtk first:

3.5.1 (default, Dec  7 2015, 12:58:09) 
[GCC 5.2.0]
/usr/lib/python3.5/site-packages/seabreeze-0.5.3-py3.5-linux-x86_64.egg/seabreeze/cseabreeze/__init__.py
3.0
(3, 18, 2)
1044
First ten wavelengths:
[ -6.10800000e+03  -4.42500000e+03  -3.07200000e+03  -2.01300000e+03
  -1.21200000e+03  -6.33000000e+02  -2.40000000e+02   3.00000000e+00
   1.32000000e+02   1.83000000e+02]

Gtk after sys:

3.5.1 (default, Dec  7 2015, 12:58:09) 
[GCC 5.2.0]
/usr/lib/python3.5/site-packages/seabreeze-0.5.3-py3.5-linux-x86_64.egg/seabreeze/cseabreeze/__init__.py
3.0
(3, 18, 2)
1044
First ten wavelengths:
[ -6.10800000e+03  -4.42500000e+03  -3.07200000e+03  -2.01300000e+03
  -1.21200000e+03  -6.33000000e+02  -2.40000000e+02   3.00000000e+00
   1.32000000e+02   1.83000000e+02]

So now the wrong wavelengths are returned regardless of the position of the gtk import.
That new behavior might be related to a package upgrade of gtk3 I recently installed.

Here is the output without importing gtk and gobject:

3.5.1 (default, Dec  7 2015, 12:58:09) 
[GCC 5.2.0]
/usr/lib/python3.5/site-packages/seabreeze-0.5.3-py3.5-linux-x86_64.egg/seabreeze/cseabreeze/__init__.py
1044
First ten wavelengths:
[ 184.34537107  185.15048402  185.95551917  186.76047652  187.56535608
  188.37015785  189.17488183  189.97952804  190.78409646  191.58858712]```

Maybe it is something distribution specific, which distribution did you use for testing ?

[ap--]

I'm closing this one, because it never got reported again... 🤷