Re: [Jailkit-dev] more jk_jailuser bugs

[Olivier Sessink]

Stephen Tallowitz wrote:
[..]

thanks for the research Stephen :)

saves me a lot of work (and thus more time for the parallel init program
I'm writing :) )

> My conclusion - there are three possibilities:
> - Try os.rename and warn the user if it fails, telling the user that 
> cross-device moving is not possible. Maybe along with the notice telling the 
> user to move using 'mv source JAILDIR/home'. I've tested GNU mv from 
> fileutils-4.0 on gentoo and it does preserve all information when moving 
> across devices. I suppose BSDs do the same thing?
> - Using some sort of exec() call from within python to call 'mv source 
> JAILDIR/home' (probably quite dangerous).
> - Implementing move() in jailkit - more time-consuming and will probably need 
> some investigation into other python library functions. Of course many ideas 
> of the shutil library could be used to ease the process of implementing such 
> a method.
> 
> Personally, I'd opt for the first possibility. I wouldn't need loads of 
> moving around anyway. I don't know about requirements of other jailkit users, 
> though.
> This would leave the path open, if the python devs fix the shutil library 
> (they do mention in the source that improvements will have to be made). It 
> also depends on how soon you want to realease the next version of jailkit, of 
> course.

opt 1 and then tell the user to move the files him/herself.. hmm I'm now
wondering what the problems of system('mv source jail/source') are.. do
you have a clear view on the possible problems?

regards,
        Olivier