Re: [rdiff-backup-users] More windows patches

[Josh Nisly]

Andrew Ferguson wrote:
> On Apr 13, 2008, at 6:24 PM, Josh Nisly wrote:

> > os.kill is not available on Windows, so there's no way from Python to 
> > determine whether a process is running, without using the PyWin32 
> > package. (We check to make sure that another rdiff-backup process 
> > isn't already running, so that we don't stomp on each other's data.) 
> > The rdiff-backup-kill.patch simply ignores the check. So here is my 
> > question: is it acceptable to import and use platform-specific 
> > modules? We would obviously need to handle it correctly if they 
> > weren't there. Seems like we do something similar for Mac OS X 
> > specific functionality, but I wanted to check.
>
> Yes, please use the PyWin32 package to implement the functionality. It 
> is very important that we make sure that only 1 rdiff-backup instance 
> is accessing a repository at a time.
>
> As a general rule, the goal here is to have absolute parity between 
> the posix and nt platforms. In cases where nt truly does not support 
> it (eg, symlinks, hardlinks), it's ok for the feature to be dropped 
> (which you did in the appropriate place: fs_abilities.py). The 
> os.kill() case is like popen2() -- rdiff-backup should use the 
> Windows-specific code as necessary.

Attached is a patch that checks for an already running process. It uses 
win32api to try to open the process. This is similar in principle to trying to 
send a signal to the process, i.e. os.kill().

Thanks,
JoshN

rdiff-backup-kill.tar
Description: Unix tar archive