On Mon, Jan 05, 2009 at 03:23:15 -0800, David Caldwell wrote:
>> So, I'm still not 100% sure I understand here. Does this mean the only
>> thing you are really interested in is to be able to create patches with
>> _darcs in them? Explicitly relative patches and patches with ".." in
>> their paths, presumably would still actually be malicious in your eyes?
>
> Yes, in my particular use case I have patches with "_darcs" somewhere in
> the path name ("./t/darcs-old/_darcs"), though notably not in the first
> path component.Ok, thanks! > Speaking in general, I'm not sure why having "_darcs" anywhere other than > the first component of the path would be malicious, but maybe I'm just > not thinking deviously enough. > > I still think it's reasonable to reject "..". At first I was confused by why anybody would want to have explicitly relative or ".." paths and now I know that the answer is "we don't". > This seems like a pretty rare edge case so I'm ok with the way it is > now--I will just add the "--dont-restrict-paths" option when I get the > error. I don't intend to mess with those nested test repos much so I > doubt it will come up too often in my future (famous last words). If I > were heavily editing them all the time it might make sense to have a > whitelist feature like Florent suggested, or to change up the definition > of is_malicious_path like I suggested. It may be worth thinking about making the --dont-restrict-paths option a bit more conservative, so that instead of not restricting paths, it only restricts them minimally (i.e. it tightens the definition of is_malicious_path). Anyway, I've applied your patches. Thanks! -- Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow> PGP Key ID: 08AC04F9
signature.asc
Description: Digital signature
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
