pje, can you please add the patch to the trunk?

Thanks,
Guy

This patch passes the unit tests

At 04:40 PM 1/21/2011 +0000, Guy Rozendorn wrote:
>I can make a refined patch if you need.

If you send me one that passes the test suite, I'll have it in the 
development snapshot versions for you ASAP.  If I have to do the 
testing and debugging myself, it'll be a lot longer before I get around to it.

you're right, that was the intention: translating it to:
"%s %s" % (ptr, ' '.join(newargs))

I can make a refined patch if you need.

At 10:07 AM 1/21/2011 +0000, Guy Rozendorn wrote:
>Attached is a patch draft, just wrote it w/o compiling and making 
>sure it really works.

Most of it looks straightforward, but I'm a little confused by the 
part that turns argv back into a command line.  It looks like it's 
just gluing the arguments together with no spaces in between.

The idea is to set an internal control event handler, rather than using the default one (that kills the cli).
the new handler should pass the ctrl-c event to the python child pid, by using GenerateConsoleCtrlEvent.
the new handler is set by using SetConsoleCtrlHandler.

so you need a control handler function, and to set it up.
you also need the child pid (to send the ctrl-c to), but spawnv doesn't return it. so you also need to move to CreateProcessA and WaitForSingleObject.
in order to use CreateProcess, you need to convert ptr and newchars into one big char*.

Attached is a patch draft, just wrote it w/o compiling and making sure it really works.

Can you supply a patch?  From what I can tell, SetConsoleCtrlHandler is inherited by the child process, so wouldn't that make Python ignore it too?

Thanks.

the cli should use SetConsoleCtrlHandler to ignore CTRL-C and pass it to its child process