zombie process generated by g.execute_shell_commands

[Xu Wang]

Dear Leo Developer,

I created one button in Leo, the corresponding script is:

c.save()

g.execute_shell_commands(['git -C /Users/wwww/leodata add austin.leo', 'git -C /Users/wwww/leodata commit -mupdates', '&git -C /Users/wwww/leodata push'])

Basically, it will save the changes and do a git add/commit/push.

However, I notice the script will leave zombie process (git). See below ps output.

Austin-MacBook-Pro:leodata wangxu$ ps -ef | grep git

  501 88822 20009   0  2:00AM ttys000    0:00.00 grep git

  501 88678 88353   0  1:57AM ttys001    0:00.00 (git)

  501 88683 88353   0  1:57AM ttys001    0:00.00 (git)

Do I need to do something extra in g.execute_shell_commands ?

BR,Austin

[Edward K. Ream]

Good question. Sometimes Leo appears to hang (on Windows) after running unit tests with g.execute_shell_commands. For me, a fix is to type a <return> in the console from which Leo is running.

I'm not sure what else can be done. The guts of g.execute_shell_commands is just:

proc = subprocess.Popen(command, shell=True)
if wait: proc.communicate()

You could try experimenting with shell=False, but I doubt that will work.

Edward

[Brian Theado]

On Sat, Feb 8, 2020 at 4:41 AM Edward K. Ream <edre...@gmail.com> wrote:

On Thu, Feb 6, 2020 at 1:08 PM Xu Wang <xu4...@gmail.com> wrote:

Dear Leo Developer,

I created one button in Leo, the corresponding script is:

c.save()

g.execute_shell_commands(['git -C /Users/wwww/leodata add austin.leo', 'git -C /Users/wwww/leodata commit -mupdates', '&git -C /Users/wwww/leodata push'])

Basically, it will save the changes and do a git add/commit/push.

[...] 

Do I need to do something extra in g.execute_shell_commands ?

Good question. Sometimes Leo appears to hang (on Windows) after running unit tests with g.execute_shell_commands. For me, a fix is to type a <return> in the console from which Leo is running.

I'm not sure what else can be done. The guts of g.execute_shell_commands is just:

proc = subprocess.Popen(command, shell=True)
if wait: proc.communicate()

You could try experimenting with shell=False, but I doubt that will work.

Here is a thread from last year on the same topic: https://groups.google.com/d/msg/leo-editor/1VOYPUJrNEM/ItZwstC0AwAJ. In my emails there, I was assuming that running short-lived processes in the background will inevitably lead to zombies and I so I shared a workaround of doing the git commits with a long-running background process.

But now I just noticed the subprocess oabjects have a poll method (https://docs.python.org/3/library/subprocess.html#subprocess.Popen.poll). Maybe if g.execute_shell_commands is called with 'not wait', then leo should save the proc object and at idle time make a call to proc.poll for it (and any other launched background tasks). Then at the next idle call proc.poll again. Keep doing that until proc.poll doesn't return None.

Brian

[Edward K. Ream]

On Sat, Feb 8, 2020 at 10:11 PM Brian Theado <brian....@gmail.com> wrote:

> I just noticed the subprocess oabjects have a poll method (https://docs.python.org/3/library/subprocess.html#subprocess.Popen.poll). Maybe if g.execute_shell_commands is called with 'not wait', then leo should save the proc object and at idle time make a call to proc.poll for it (and any other launched background tasks). Then at the next idle call proc.poll again. Keep doing that until proc.poll doesn't return None.

Thanks for this. I have just created #1489 for this topic. Python 3.5 and above has a new, high level, run method. Using this method might "just work".

Edward

[Edward K. Ream]

On Sunday, February 9, 2020 at 6:08:42 AM UTC-6, Edward K. Ream wrote:

> I have just created #1489 for this topic.

Rev 98530b0 in devel attempts a fix. I'm not sure it actually does anything useful. Please reopen if problems remain, or if you have other suggested fixes.

Edward

[Brian Theado]

This is working fine for me now. I don't see zombie processes for the exited background commands anymore. Nice work.

--
You received this message because you are subscribed to the Google Groups "leo-editor" group.
To unsubscribe from this group and stop receiving emails from it, send an email to leo-editor+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/leo-editor/e81a0a75-f12e-48ea-9bb3-0f81c447df95%40googlegroups.com.

[Xu Wang]

updated to devel 9b6f29d12..93abd72c6, the problem fixed on my MacBook too.

Thanks EKR.

Brian Theado <brian....@gmail.com> 于2020年2月12日周三 上午5:29写道：

To view this discussion on the web visit https://groups.google.com/d/msgid/leo-editor/CAO5X8CyAVyKbeiLVxOB1JO_H28%2BR8zcg3qnNRjT8o2bkuSNnBQ%40mail.gmail.com.

[Edward K. Ream]

On Tue, Feb 11, 2020 at 4:29 PM Brian Theado <brian....@gmail.com> wrote:

This is working fine for me now. I don't see zombie processes for the exited background commands anymore. Nice work.

Well, you were the one who made it work :-)

Edward

[Edward K. Ream]

On Tue, Feb 11, 2020 at 10:14 PM Xu Wang <xu4...@gmail.com> wrote:

updated to devel 9b6f29d12..93abd72c6, the problem fixed on my MacBook too.

Thanks EKR.

You're welcome. Thanks for the confirmation.

Edward

[Thomas Passin]

I believe that this happens when you accidentally cause an event in the Windows cmd terminal.  I'm referring to the terminal from which Leo is launched (i.e., with python -m leo.core.runLeo). It's not only during unit testing.  For me, the usual case is that I want to switch to the terminal to see more of the output.  I click the mouse in the terminal but accidentally drag it.  This hangs Leo.  I think it's something about stdin waiting for a keystroke.  That's why typing <ENTER> clears it - that terminates the readline on stdin.

Sometimes after I do this, Leo does a full hang, with the circle cursor and translucent overlay (Linux users: this is a Windows thing).  Usually this condition will time out after 5 - 10 seconds.

[Edward K. Ream]

On Mon, Mar 2, 2020 at 8:46 AM Thomas Passin <tbp1...@gmail.com> wrote:

I believe that this happens when you accidentally cause an event in the Windows cmd terminal.  I'm referring to the terminal from which Leo is launched (i.e., with python -m leo.core.runLeo). It's not only during unit testing.  For me, the usual case is that I want to switch to the terminal to see more of the output.  I click the mouse in the terminal but accidentally drag it.  This hangs Leo.  I think it's something about stdin waiting for a keystroke.  That's why typing <ENTER> clears it - that terminates the readline on stdin.

I agree, it's likely something waiting on stdin. Enter (in the console) clears it.

Edward

[Matt Wilkie]

... I click the mouse in the terminal but accidentally drag it.  This hangs Leo.  I think it's something about stdin waiting for a keystroke.  That's why typing <ENTER> clears it - that terminates the readline on stdin.

Yeah, this is a generic Windows Command Prompt window problem. I use the 3rd party shell ConEmu to avoid this, as well as adding a bunch of other goodies I've come to rely on. Microsoft has been putting a lot of work into Windows Terminal which I expect will probably ship with Windows at some point and address the same things.

That said, it's still good to have a solution that isn't telling people to install something else to fix things!

-matt

[Edward K. Ream]

Thanks for all these comments.

Edward