Mailing-List: contact cygwin-developers-help AT cygwin DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-developers-subscribe AT cygwin DOT com>
List-Archive: <http://sources.redhat.com/ml/cygwin-developers/>
List-Post: <mailto:cygwin-developers AT cygwin DOT com>
List-Help: <mailto:cygwin-developers-help AT cygwin DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-developers-owner AT cygwin DOT com
Delivered-To: mailing list cygwin-developers AT cygwin DOT com
Message-ID: <3E4D1046.C2BEA2ED@ieee.org>
Date: Fri, 14 Feb 2003 10:50:30 -0500
From: "Pierre A. Humblet" <Pierre DOT Humblet AT ieee DOT org>
X-Accept-Language: en,pdf
MIME-Version: 1.0
To: cygwin-developers AT cygwin DOT com
Subject: Re: uh oh [Roland DOT Schwingel AT onevision DOT de: Re: bash broken with cygwin 
 1.3.20? - now working with 1.3.20]
References: <20030213144024 DOT GC27226 AT redhat DOT com> <3E4BB4EE DOT 6DF1C18B AT ieee DOT org> <20030214085422 DOT GB5822 AT cygbert DOT vinschen DOT de>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Corinna Vinschen wrote:
> 
> On Thu, Feb 13, 2003 at 10:08:30AM -0500, Pierre A. Humblet wrote:
> > Christopher Faylor wrote:
> > > Ok, what's causing these type of problems in 1.3.20?  There has been a
> > > few of these reported.
> >
> > > > Please post the output of 'ls -l ../gcc-3.2.1/configure.in'.
> > > -rwx------+   1 Administ Entwickl    55070 Jul  8  2002
> > > ../gcc-3.2.1-ov/configure.in
> >
> > Not sure how bash looks up the access mode.
> > I have some evidence that access() is broken but can't debug until tonight.
> 
> I can't reproduce it.  I've changed an executable to have the above
> permissions and the admins group as owner and I was able to call
> the executable with any of the shells, ash, bash and tcsh.

I haven't tried to reproduce it but there is definitely a bug in internal_getgroups
(introduced on 2003-02-04), which may explain it. Patch on the way.

Also looking at the logic in acl_access
      if ((!(flags & R_OK) || (acls[i].a_perm & S_IROTH))
	  && (!(flags & W_OK) || (acls[i].a_perm & S_IWOTH))
	  && (!(flags & X_OK) || (acls[i].a_perm & S_IXOTH)))
	return 0;
it looks like we only return OK if a single acl grants all the requested rights,
but we fail if say one acl grants R_OK and another one grants W_OK and we need
both.
I am thinking of redoing acl_access by using the MS AccessCheck() function.
This will take care of the issue above, make the call completely independent of 
the passwd and group files and better handle access_denied ACE's.
Any thought?

Pierre