You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Ctrl-C not clearing prompt in terminal (with EPD?) #625
As reported by EPD users on the mailing list, Ctrl-C no longer works to clear a prompt in the terminal. One user has apparently tried this with a newer checkout from git, but sees the same problem. This has affected one Linux and one Windows user, but I can't replicate it on Linux.
Expected behaviour:
In [1]: cd something_wrong
KeyboardInterrupt
In [1]:
N.B. It seems Ctrl-C does still work to interrupt running commands.
The Pylab shortcut installed by EPD is somewhat intricate. I've tested how Ctrl-C behaves when starting with a plain "ipython". This is on Windows 7 64-bit, EPD 7.1 with Python 2.7.2 and IPython 0.11.dev (2011-07-26).
Ctrl-C does give the expected KeyboardInterrupt message, but in fact leaves you at a Windows cmd.exe command line. Strange things happen with the process tree in the task manager, and with the processing of console input. To reproduce:
1. Starting the Windows command line
Winkey-R (Run) > cmd
The process explorer now shows "explorer.exe > cmd.exe".
The process explorer now shows:
explorer.exe
cmd.exe
ipython.exe
python.exe
3. Press Ctrl-C
In [1]: ^C
C:\Users\jonvi>
KeyboardInterrupt
In [1]:
The process explorer now shows:
explorer.exe
cmd.exe
python.exe
...as if ipython.exe was killed, orphaning the python.exe process.
4. Stuck in cmd.exe
Although the prompt looks innocuous at first, input actually goes to cmd.exe:
In [1]: 2+2
2+2 is not recognized as an internal or external command, operable program or batch file.
C:\Users\jonvi>
5. exit
One might try to "exit" cmd.exe. This sometimes works (but see below), leaving the process explorer showing:
explorer.exe
python.exe
Pressing Enter again returns to IPython (python.exe, not ipython.exe) and stuff works normally.
In [2]: 2+2
Out[2]: 4
In [3]: exit
...exits normally.
However, sometimes the input processing of cmd.exe gets erratic:
C:\Users\jonvi>ipython
In [1]: ^C
C:\Users\jonvi>
KeyboardInterrupt
In [1]: 2+2
2+2 gjenkjennes ikke som en intern eller ekstern kommando,
kjørbart program eller satsvis fil.
C:\Users\jonvi>
// On pressing "e", everything back to the last "In [1]:" gets deleted.
In [1]: exit
xit is not recognized as an internal or external command, operable program or batch file.
C:\Users\jonvi>
// Yes, that's "xit" with no "e".
// Pressing "e" again deletes everything back to "In [1]:e".
// Finishing with "xit" gives this:
In [1]: eexit
xit is not recognized as an internal or external command, operable program or batch file.
// Pressing Enter twice gives this:
In [1]: ee
NameError: name 'ee' is not defined
I'm not sure what determines whether keyboard input gets mangled. For a while I thought it had to do with whether I typed anything before Ctrl-C, whether I entered anything before "exit", or whether I waited a little while (to put together this bug report) -- but unfortunately I seem unable to predict which of the two scenarios will happen.
On the Qt console, Ctrl-C doesn't do anything while you're entering code. Esc or Ctrl-G will clear the current cell. This is intentional.
As for the Windows thing, I have no idea. Does the same thing happen outside EPD, e.g. if you run python.exe C:\Path\to\ipython.py? Is ipython.exe something the setup process creates, or something Enthought produce?
Actually, Ctrl-C copies when you're entering code in the Qt console, and interrupts when you're executing code. We made Ctrl-C context sensitive as a compromise between the conflicting conventions in GUI and console applications.
On the Qt console, Ctrl-C doesn't do anything while you're entering code. Esc or Ctrl-G will clear the current cell. This is intentional.
Sounds reasonable, works as advertised.
As for the Windows thing, I have no idea. Does the same thing happen outside EPD, e.g. if you run python.exe C:\Path\to\ipython.py?
Huh. As it turns out, "python -m ipython" handles Ctrl-C just fine:
Microsoft Windows [Versjon 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. Med enerett.
C:\Users\jonvi>python -m ipython
Python 2.7.2 |EPD 7.1-1 (64-bit)| (default, Jul 3 2011, 15:34:33)
[MSC v.1500 64 bit (AMD64)]
IPython 0.11.dev -- An enhanced Interactive Python.
In [1]: 2+2
Out[1]: 4
In [2]: Pressing Ctrl-C now...
KeyboardInterrupt
In [2]: 2+2
Out[2]: 4
In [3]: ølkjsadfadfsø
KeyboardInterrupt
In [3]:
In [4]: exit
Works fine in Pylab mode too:
c:\Python27\Scripts>python -m ipython --pylab
Welcome to pylab, a matplotlib-based Python environment [backend: WXAgg].
In [1]: 2+2
Out[1]: 4
In [2]: Pressing Ctrl-C now...Nothing happens
File "", line 1
Pressing Ctrl-C now...Nothing happens
^
SyntaxError: invalid syntax
It does indeed seem that EPD can avoid the weird Ctrl-C behaviour just
by simplifying their shortcut. I'll post a message to epd-users@enthought.com pointing here.
Is ipython.exe something the setup process creates, or something Enthought produce?
I'm not sure, and unfortunately I have no convenient way to check. For
what it's worth, ipython.exe is in the python Scripts directory, which
is in my PATH.
c:\Python27\Scripts>dir ipython*
Innhold i c:\Python27\Scripts
2011-07-26 13:21 340 ipython-qtconsole-script.pyw
2011-07-26 13:21 7 168 ipython-qtconsole.exe
2011-07-26 13:21 327 ipython-script.py
2011-07-26 13:21 7 168 ipython.exe
EDIT: There is an ipython.bat too, but that one I wrote myself.
I have this problem on windows and am using standard ipython, 0.11 not enthought.
if i execute ipython from the windows dos like prompt it has this problem but does not when run from the icon i have pinned. I think the windows terminal also receives the interrupt and starts capturing input itself. My guess is that there is some call back to say "i got this one" that is not being made by ipython when it handles the signal.
I have cywin installed but this happens with and without cygwin enabled.
Thanks for the tips about using task manager for a fix. I hope to at least save some data!
I am happy to install from github and test.
The Windows bug of ctrl-C breaking the shell when run from cmd appears to be caused by setuptools when it creates the ipython.exe. If you do python C:\Python27\Scripts\ipython-script.py it seems to avoid the issue.
The problem on windows 7 64bit running ipython 0.11 python 2.7 happens when ipython is run directly by typing 'ipython' at the windows or cygwin prompt. Python installed on windows, not through cygwin setup.exe does not know about cygwin so you need to give it the windows path of the script file, but after ipython starts when run from cygwin you are able to use the gnu commands with bang ( !stat file,!ls, !egrep -v '^$|^ *#' myfile.py , etc)
With this issue being 5 years old, the lack of activity, and all the recent refactor I'll assume this is either fixed, or need a new issue to be reopened.
Thanks all, and closing to try to keep issue list small.
Activity
jonovik commentedon Jul 26, 2011
The Pylab shortcut installed by EPD is somewhat intricate. I've tested how Ctrl-C behaves when starting with a plain "ipython". This is on Windows 7 64-bit, EPD 7.1 with Python 2.7.2 and IPython 0.11.dev (2011-07-26).
Ctrl-C does give the expected KeyboardInterrupt message, but in fact leaves you at a Windows cmd.exe command line. Strange things happen with the process tree in the task manager, and with the processing of console input. To reproduce:
1. Starting the Windows command line
Winkey-R (Run) > cmd
The process explorer now shows "explorer.exe > cmd.exe".
2. Starting IPython
C:\temp\testing>ipython
Python 2.7.2 |EPD 7.1-1 (64-bit)| (default, Jul 3 2011, 15:34:33) [MSC v.1500 64 bit (AMD64)]
IPython 0.11.dev -- An enhanced Interactive Python.
The process explorer now shows:
explorer.exe
cmd.exe
ipython.exe
python.exe
3. Press Ctrl-C
In [1]: ^C
C:\Users\jonvi>
KeyboardInterrupt
In [1]:
The process explorer now shows:
explorer.exe
cmd.exe
python.exe
...as if ipython.exe was killed, orphaning the python.exe process.
4. Stuck in cmd.exe
Although the prompt looks innocuous at first, input actually goes to cmd.exe:
In [1]: 2+2
2+2 is not recognized as an internal or external command, operable program or batch file.
C:\Users\jonvi>
5. exit
One might try to "exit" cmd.exe. This sometimes works (but see below), leaving the process explorer showing:
explorer.exe
python.exe
Pressing Enter again returns to IPython (python.exe, not ipython.exe) and stuff works normally.
In [2]: 2+2
Out[2]: 4
In [3]: exit
...exits normally.
However, sometimes the input processing of cmd.exe gets erratic:
C:\Users\jonvi>ipython
In [1]: ^C
C:\Users\jonvi>
KeyboardInterrupt
In [1]: 2+2
2+2 gjenkjennes ikke som en intern eller ekstern kommando,
kjørbart program eller satsvis fil.
C:\Users\jonvi>
// On pressing "e", everything back to the last "In [1]:" gets deleted.
In [1]: exit
xit is not recognized as an internal or external command, operable program or batch file.
C:\Users\jonvi>
// Yes, that's "xit" with no "e".
// Pressing "e" again deletes everything back to "In [1]:e".
// Finishing with "xit" gives this:
In [1]: eexit
xit is not recognized as an internal or external command, operable program or batch file.
// Pressing Enter twice gives this:
In [1]: ee
NameError: name 'ee' is not defined
I'm not sure what determines whether keyboard input gets mangled. For a while I thought it had to do with whether I typed anything before Ctrl-C, whether I entered anything before "exit", or whether I waited a little while (to put together this bug report) -- but unfortunately I seem unable to predict which of the two scenarios will happen.
jonovik commentedon Jul 26, 2011
ipython qtconsole
Here Ctrl-C on the command line just gets ignored. It works fine for interrupting code, though.
In [1]: while True: pass
KeyboardInterrupt:
takluyver commentedon Jul 26, 2011
On the Qt console, Ctrl-C doesn't do anything while you're entering code. Esc or Ctrl-G will clear the current cell. This is intentional.
As for the Windows thing, I have no idea. Does the same thing happen outside EPD, e.g. if you run
python.exe C:\Path\to\ipython.py? Is ipython.exe something the setup process creates, or something Enthought produce?epatters commentedon Jul 26, 2011
Actually, Ctrl-C copies when you're entering code in the Qt console, and interrupts when you're executing code. We made Ctrl-C context sensitive as a compromise between the conflicting conventions in GUI and console applications.
jonovik commentedon Jul 26, 2011
On Tue, Jul 26, 2011 at 4:06 PM, takluyver
reply@reply.github.com
wrote:
Sounds reasonable, works as advertised.
Huh. As it turns out, "python -m ipython" handles Ctrl-C just fine:
Microsoft Windows [Versjon 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. Med enerett.
C:\Users\jonvi>python -m ipython
Python 2.7.2 |EPD 7.1-1 (64-bit)| (default, Jul 3 2011, 15:34:33)
[MSC v.1500 64 bit (AMD64)]
IPython 0.11.dev -- An enhanced Interactive Python.
In [1]: 2+2
Out[1]: 4
In [2]: Pressing Ctrl-C now...
KeyboardInterrupt
In [2]: 2+2
Out[2]: 4
In [3]: ølkjsadfadfsø
KeyboardInterrupt
In [3]:
In [4]: exit
Works fine in Pylab mode too:
c:\Python27\Scripts>python -m ipython --pylab
Welcome to pylab, a matplotlib-based Python environment [backend: WXAgg].
In [1]: 2+2
Out[1]: 4
In [2]: Pressing Ctrl-C now...Nothing happens
File "", line 1
Pressing Ctrl-C now...Nothing happens
^
SyntaxError: invalid syntax
It does indeed seem that EPD can avoid the weird Ctrl-C behaviour just
by simplifying their shortcut. I'll post a message to
epd-users@enthought.com pointing here.
I'm not sure, and unfortunately I have no convenient way to check. For
what it's worth, ipython.exe is in the python Scripts directory, which
is in my PATH.
c:\Python27\Scripts>dir ipython*
Innhold i c:\Python27\Scripts
2011-07-26 13:21 340 ipython-qtconsole-script.pyw
2011-07-26 13:21 7 168 ipython-qtconsole.exe
2011-07-26 13:21 327 ipython-script.py
2011-07-26 13:21 7 168 ipython.exe
EDIT: There is an ipython.bat too, but that one I wrote myself.
takluyver commentedon Jul 26, 2011
Thanks, Jon. For now I think it looks like the Ctrl-C behaviour is somehow related to EPD. Let us know what you hear from them.
jonovik commentedon Jul 26, 2011
Message posted here:
http://article.gmane.org/gmane.comp.python.epd.user/332
Whole thread:
http://news.gmane.org/find-root.php?message_id=%3cloom.20110727T130317%2d482%40post.gmane.org%3e
It seems you need to be subscribed to post; subscription info is here:
https://mail.enthought.com/mailman/listinfo/epd-users
ijstokes commentedon Sep 29, 2011
I'm using EPD 7.1-2 ipython on OS X, and CTRL-C doesn't work for me.
esotericbyte commentedon Nov 16, 2011
I have this problem on windows and am using standard ipython, 0.11 not enthought.
if i execute ipython from the windows dos like prompt it has this problem but does not when run from the icon i have pinned. I think the windows terminal also receives the interrupt and starts capturing input itself. My guess is that there is some call back to say "i got this one" that is not being made by ipython when it handles the signal.
I have cywin installed but this happens with and without cygwin enabled.
Thanks for the tips about using task manager for a fix. I hope to at least save some data!
I am happy to install from github and test.
minrk commentedon Nov 17, 2011
The Windows bug of ctrl-C breaking the shell when run from cmd appears to be caused by setuptools when it creates the ipython.exe. If you do
python C:\Python27\Scripts\ipython-script.pyit seems to avoid the issue.davidbrai commentedon Nov 18, 2011
thanks @minrk ! the ctrl-c bug was driving me crazy.
esotericbyte commentedon Nov 20, 2011
The problem on windows 7 64bit running ipython 0.11 python 2.7 happens when ipython is run directly by typing 'ipython' at the windows or cygwin prompt. Python installed on windows, not through cygwin setup.exe does not know about cygwin so you need to give it the windows path of the script file, but after ipython starts when run from cygwin you are able to use the gnu commands with bang ( !stat file,!ls, !egrep -v '^$|^ *#' myfile.py , etc)
Carreau commentedon Jul 21, 2016
With this issue being 5 years old, the lack of activity, and all the recent refactor I'll assume this is either fixed, or need a new issue to be reopened.
Thanks all, and closing to try to keep issue list small.
Thanks !