Re: [hackers] [PATCH 1/3][sbase] chown: Ignore failure to find user/group name

From: Richard Ipsum <richardipsum_AT_fastmail.co.uk>
Date: Wed, 26 Jun 2019 00:07:10 +0100

On Tue, Jun 25, 2019 at 03:44:57PM -0700, Michael Forney wrote:
> On 2019-06-25, Richard Ipsum <richardipsum_AT_fastmail.co.uk> wrote:
> > This fixes an error on OpenBSD,
> >
> > % chown 1000:1000 yes.c
> > ./chown: getgrnam 1000: No such file or directory
>
> POSIX says that if the entry cannot be found, errno shall not be
> changed. Are you saying that OpenBSD doesn't behave this way?

Yeah when I run this on OpenBSD with uid:gid values I get ENOENT.

>
> > ---
> > chown.c | 18 ++----------------
> > 1 file changed, 2 insertions(+), 16 deletions(-)
> >
> > diff --git a/chown.c b/chown.c
> > index d7dc8e0..748ce97 100644
> > --- a/chown.c
> > +++ b/chown.c
> > _AT_@ -79,26 +79,12 @@ main(int argc, char *argv[])
> > *group++ = '\0';
> >
> > if (owner && *owner) {
> > - errno = 0;
> > pw = getpwnam(owner);
> > - if (pw) {
> > - uid = pw->pw_uid;
> > - } else {
> > - if (errno)
> > - eprintf("getpwnam %s:", owner);
> > - uid = estrtonum(owner, 0, UINT_MAX);
> > - }
> > + uid = pw ? pw->pw_uid : estrtonum(owner, 0, UINT_MAX);
>
> It looks like there is no error with getpwnam, only with getgrname, right?
>

Yes, but I would argue that realistically the only time this function errors is
when the entry doesn't exist, we don't want to exit in that case,
instead try parsing it as a numeric uid and if that fails then exit.

Thanks,
Richard
Received on Wed Jun 26 2019 - 01:07:10 CEST