Re: Subclass Gio.DBusProxy and UnicodeDecodeError under python 3

[infirit <infirit gmail com>]

Op 12/13/2016 om 04:31 PM schreef infirit:
> Op 12/12/2016 om 11:49 PM schreef Christoph Reiter:
> > On Sun, Dec 11, 2016 at 6:12 PM, infirit <infirit gmail com> wrote:
> > > Hi all,
> > >
> > > I subclassed Gio.DBusProxy and overridden the do_g_properties_changed
> > > virtual function. In there I emit my own signal that consumers of the
> > > proxy can connect to. I have created a minimal example of what I am
> > > doing which you can find on [1].
> > >
> > > The code from [1] runs perfectly fine under python2.7 with no
> > > UnicodeDecodeError at any point, the situation under Python 3.4 however
> > > is different. See below an example error. I narrowed it down to
> > > overriding do_g_properties_changed, without that the error goes away.
> > >
> > >   *connected function* /org/bluez/hci0/dev_7C_2F_80_3B_AF_AB RSSI -64
> > >   UnicodeDecodeError: 'utf-8' codec can't decode byte 0xb8 in position
> > > 0: invalid start byte
> > >
> > > The strange thing is that it does not generate a traceback so I am not
> > > entirely sure how to proceed with this. It also does not seem to result
> > > in issues however my gut tells me not to ignore it. Anyone know what is
> > > going on with this?
> > Under Python 2 text gets passed to Python as is (utf-8 encoded) while
> > under Python 3 things get decoded to unicode code points. I suspect
> > there is some invalid utf-8 involved which you should see in data
> > getting passed to do_g_properties_changed under py2.
> >
> > I've tested your code (with adjusting the device path) and it seems to
> > work fine under py3 here.
>
> The problem is printing both properties changed and invalidated
> properties is identical between the two python versions. The only
> difference is with Python3 at the end I see the error. I monitored the
> object from the command line (below) and I see all events back in my
> proxy output. However there is something strange with the invalidated
> properties, more on that down below.
>
> #gdbus -y -d "org.bluez" -o "/org/bluez/hci0/dev_7C_2F_80_3B_AF_AB"
> /org/bluez/hci0/dev_7C_2F_80_3B_AF_AB:
> org.freedesktop.DBus.Properties.PropertiesChanged ('org.bluez.Device1',
> {'RSSI': <int16 -61>}, @as [])
> /org/bluez/hci0/dev_7C_2F_80_3B_AF_AB:
> org.freedesktop.DBus.Properties.PropertiesChanged ('org.bluez.Device1',
> @a{sv} {}, ['RSSI'])
>
> Corresponding output from my code under python3
>
> *<class '__main__.MyProxy'> : changed*
> /org/bluez/hci0/dev_7C_2F_80_3B_AF_AB RSSI -61
> *connected function* /org/bluez/hci0/dev_7C_2F_80_3B_AF_AB RSSI -61
> *invalidated* <class 'str'> ''
> UnicodeDecodeError: 'utf-8' codec can't decode byte 0x90 in position 2:
> invalid start byte
>
> and python2
>
> ("*<class '__main__.MyProxy'> : changed*",
> '/org/bluez/hci0/dev_7C_2F_80_3B_AF_AB', 'RSSI', -61)
> ('*connected function*', '/org/bluez/hci0/dev_7C_2F_80_3B_AF_AB',
> 'RSSI', -61)
> ('*invalidated*', <type 'str'>, "''", '')
>
> Now on the invalidated properties. I personally have need for them but I
> did notice the virtual do_g_properties_changed is not handling them
> properly. Take the second line from gdbus, it has no changed properties
> and one invalidated property. Looking at what I get in
> do_g_properties_changed I noticed I always get an empty *string*
> (instead of a list). This looks like a bug as it is definitely on the
> bus but never shows up in do_g_properties_changed. When I connect
> directly (without using the virtual function) to g_properties_changed I
> do see the invalidated properties correctly as a *list* of strings.
>
> My guess is that the two are related but I am not sure where to go from
> here.

Should I open a bug report?

~infirit