[GRASS-dev] [GRASS GIS] #3740: The GRASS session is "leaking" to the environment

[GRASS GIS]

#3740: The GRASS session is "leaking" to the environment
--------------------------+-------------------------
  Reporter:  pmav99       |      Owner:  grass-dev@…
      Type:  defect       |     Status:  new
  Priority:  normal       |  Milestone:
 Component:  Startup      |    Version:  svn-trunk
Resolution:               |   Keywords:
       CPU:  Unspecified  |   Platform:  Linux
--------------------------+-------------------------

Comment (by pmav99):

 Replying to [comment:6 mmetz]:
 > This is what happens for me: the `Quit GRASS GIS` button returns me to
 my own shell.

 I don't doubt you, but I am afraid this is not the behavior I am observing
 though. The complete setup on my end is the following:

 1. Create a fresh VM running ubuntu 18.04
 2. Check out trunk
 3. Compile

 Note: In order to avoid the overhead of running Xorg in the VM, I make the
 VM headless and connect via SSH. Whenever I need to run a gui application
 in the VM I launch an xpra server in the VM: https://xpra.org/

 In case someone is not familiar with xpra, it is more or less the
 equivalent of X11 forwarding on SSH. I can't say that xpra is definitely
 not interfering, but I don't think it is very likely. I will try to test
 with plain X11 forwarding to rule that out.

 So, when I want to e.g. launch the GRASS GUI, I start an instance of a
 terminal emulator (e.g. `terminator`) via xpra. Just to make sure that
 everything is clear, the terminal emulator is running on the VM; not on my
 laptop. It is being rendered on my laptop's Xorg though. But, at least in
 theory this should not be important.

 So from within this terminal emulator I launch GRASS using my script. As I
 said in a previous message, I did try to launch `grass77` directly from
 the emulator's shell (i.e. without the script) but this didn't change
 anything. Anyway, while the GRASS GUI is running, the relevant processes
 on the VM are:

 {{{
 pmav99    8770  2.7  2.3 1131716 192136 ?      Sl   12:02   0:04
 /usr/bin/python2 /usr/bin/xpra start --speaker=disabled --start-
 child=terminator --microphone=disabled --webcam=disabled --pulseaudio=no
 --env=XPR
 pmav99    8772  3.1  2.9 1196104 239844 ?      Ssl  12:02   0:04  \_
 /usr/lib/xorg/Xorg-for-Xpra-S8753 -noreset -novtswitch -nolisten tcp
 +extension GLX +extension RANDR +extension RENDER -auth /home/pmav99/.Xau
 pmav99    8896  0.3  0.9 1132248 76916 ?       Sl   12:02   0:00  \_
 /usr/bin/python /usr/bin/terminator
 pmav99    8921  0.4  0.0  54412  6684 pts/1    Ss   12:02   0:00      \_
 /bin/zsh
 pmav99    9094  0.0  0.0  15244  3116 pts/1    S    12:04   0:00
 \_ bash src2/grass/run_gui.sh
 pmav99    9098  0.1  0.1  35200 12344 pts/1    S    12:04   0:00
 \_ python /home/pmav99/src2/grass/bin.x86_64-pc-linux-gnu/grass77
 pmav99    9151  0.3  0.0  49808  5980 pts/1    S+   12:04   0:00
 \_ /bin/zsh
 pmav99    9152  4.2  2.0 1087160 169400 pts/1  Sl   12:04   0:01
 \_ python /home/pmav99/src2/grass/dist.x86_64-pc-linux-
 gnu/gui/wxpython/wxgui.py
 }}}

 As you can see, `grass77` starts two child processes ,9151 and 9152. Both
 of them are functioning as they should (the process states are S+ and Sl
 respectively). Now, when I click on `Quit GRASS GIS`, the GUI closes, but
 the shell remains. And this is how the process tree is looking:
 {{{
 pmav99    8770  1.4  2.2 1116340 183864 ?      Sl   12:02   0:05
 /usr/bin/python2 /usr/bin/xpra start --speaker=disabled --start-
 child=terminator --microphone=disabled --webcam=disabled --pulseaudio=no
 --env=XPR
 pmav99    8772  1.3  2.6 1089848 213520 ?      Ssl  12:02   0:05  \_
 /usr/lib/xorg/Xorg-for-Xpra-S8753 -noreset -novtswitch -nolisten tcp
 +extension GLX +extension RANDR +extension RENDER -auth /home/pmav99/.Xau
 pmav99    8896  0.1  0.9 1132248 76916 ?       Sl   12:02   0:00  \_
 /usr/bin/python /usr/bin/terminator
 pmav99    8921  0.1  0.0  54412  6684 pts/1    Ss   12:02   0:00      \_
 /bin/zsh
 pmav99    9094  0.0  0.0  15244  3116 pts/1    S    12:04   0:00
 \_ bash src2/grass/run_gui.sh
 pmav99    9098  0.0  0.1  35200 12344 pts/1    S    12:04   0:00
 \_ python /home/pmav99/src2/grass/bin.x86_64-pc-linux-gnu/grass77
 pmav99    9151  0.0  0.0  49808  5980 pts/1    S+   12:04   0:00
 \_ /bin/zsh
 pmav99    9152  0.8  0.0      0     0 pts/1    Z    12:04   0:02
 \_ [python] <defunct>
 }}}
 9151 is still running (state `S+`) while 9152 is a zombie process

 I would be happy to hear any suggestions to debug this

-- 
Ticket URL: <https://trac.osgeo.org/grass/ticket/3740#comment:7>
GRASS GIS <https://grass.osgeo.org>