Re: uid_t vs. gid_t

[Patrik Olsson]

On Fri, 2010-04-16 at 04:24 +0200, olafBuddenhagen@gmx.net wrote:
> Hi,
> 
> On Mon, Apr 12, 2010 at 09:11:52PM +0200, Patrik Olsson wrote:
> > On Sat, 2010-04-10 at 12:09 +0200, olafBuddenhagen@gmx.net wrote:
> > > On Fri, Apr 09, 2010 at 11:42:25AM +0200, Patrik Olsson wrote:
> 
> > > > I have found one thing that seems incorrect in
> > > > hurd.git/hurd/password.defs. See the parameter/field "group" in
> > > > "password_check_group"? Shouldn't it be of type gid_t and not
> > > > uid_t?
> [...]
> > > Still, this probably should be fixed. How about a patch? :-)
> > 
> > Well, if it ain't broken, then why fix it? :-)
> > 
> > I think that going through all those folders to change the type will
> > take a whole day. It's not really easy to make a script for it either
> > since not all uid_t should be changed, only some. Besides, I cannot
> > check if the software will still build after the changes (or if it
> > will break other packages that assume the old type).
> 
> Why can't you check it?...

Because I don't have the tools installed (yet). :-)

> I agree though that it's probably not worth the trouble, considering
> that it's really just a formalism...
> 
> I wonder whether it wouldn't be smarter just to abolish gid_t
> alltogether instead -- they are not likely to ever differ in technical
> terms, and there is probably no type checking worth talking of involved
> either...
> 

I thought about that too. Should it be wanted, I have attached a patch
that does this. It is *untested* and was made with:

$ git ls-files | xargs sed -i~ 's/\bgid_t\b/uid_t/g'

There are still some instances where gid_t-like types are used in
"./pfinet/linux-src/include/linux/". Like "ntfs_gid_t", "__kernel_gid_t"
and "vgid_t". See:

$ git ls-files | xargs grep gid_t

/Patrik

0001-Change-gid_t-to-uid_t.patch
Description: Text Data

signature.asc
Description: This is a digitally signed message part