[PATCH 10 of 14] chg: calculate sockdirfd

[Yuya Nishihara]

On Wed, 13 Apr 2016 17:50:57 +0100, Jun Wu wrote:
> On 04/13/2016 04:07 PM, Yuya Nishihara wrote:
> > I still think we should avoid unnecessary path manipulation because we have
> > to be careful about pitfalls such as CHGSOCKNAME="/silly/basename/..".
> > That's why I prefer CHGSOCKDIR.  
> 
> I still prefer the flexibility. People using ".." should know what they are
> doing. I don't think it necessary to prevent people using developer-facing
> features from doing wrong. Things like "rm -rf ~" are not protected.

Then why do you check basename[0] == '\0' ?

I don't like being loose for processing paths because it tends to be a security
bug. I know that's okay right now, but can you be sure that basename = ".."
never ever trap someone who has to modify this function?

Also, I don't see how beneficial it is to allow putting all sockets into
a single directory.