Re: [TESTERS needed] New POSIX permission handling

[Ismail Donmez <ismail@donmez.ws>]

Hi,


Corinna Vinschen-2 wrote
> Hi Ismail,
> 
> On Apr 12 16:25, Corinna Vinschen wrote:
>> On Apr 12 06:21, İsmail Dönmez wrote:
>> > Corinna Vinschen-2 wrote
>> > > On Apr 11 10:11, donmez wrote:
>> > >> Corinna Vinschen-2 wrote
>> > >> > I just applied a patch I'm working on for quite some time now.  As
>> I
>> > >> > outlined before on this list, the POSIX permission handling has
>> aged
>> > >> > considerably and, for historical reasons, did things differently
>> > >> > dependent on the calling function.  I took the time to reimplement
>> the
>> > >> > core functionality to handle all ACLs as strictly following POSIX
>> ACL
>> > >> > rules as possible.
>> > >> 
>> > >> I tested the updated package and at least quilt and mutt seems to
>> broken
>> > >> by
>> > >> the permission changes:
>> > >> [...]
>> > > No offense, but this is not overly helpful.  The problem is to learn
>> > > *why* this happens and how to fix it.  For that I'd need to know what
>> > > your permissions on /tmp look like (ls -l, getfacl, icacls). 
>> Creating
>> > > files in my /tmp (having an old-style ACL) with the following
>> > > permissions works as desired for me:
>> > 
>> > Hopefully this will shed some more light:
>> 
>> It does, thank you.  The problem is the dreaded "owner == group" problem
>> introduced with these weird Microsoft accounts.  I completely forgot
>> about this while implementing the new code.  It's pretty tricky to get
>> the Windows ACL right for this.  Additionally the ACLs already created
>> by setup are... borderline correct only.  Back to the drawing board...
> 
> I just applied a patch which is supposed to handle this owner==group
> scenario better.
> 
> In short, Cygwin will try to handle POSIX user and group permissions
> separately, even if owner == group.  This is basically a fake as far
> as the actual permissions of the account are concerned, but it allows
> applications still to chmod to different user and group perms.  It
> just *looks* different in the end.
> 
> The only restriction of this is that the POSIX user permissions are
> always changed so that the user perms are >= the group perms in this
> situation.  So this:
> 
>   chmod 460 foo
> 
> will be internally twisted into
> 
>   chmod 660 foo
> 
> 
> I uploaded new developer snapshots to https://cygwin.com/snapshots/
> and I'm just uploading a 2.0.0-0.5 test release with this change.

I tested the new test release (rm -rfd the cygwin installation and
re-installed from scratch just to be sure), my original testcase with quilt
and mutt works, BUT now when I run make install inside mutt source it
complains that /usr/bin/install cannot change permissions on the destination
executables.

Now this is a pretty vague error report but it might take some time before I
can post an easily reproducable error. Because running the same command
myself gives no error whatsoever. I'll try to come up with a better report.

Thanks a lot!




--
View this message in context: http://cygwin.1069669.n5.nabble.com/TESTERS-needed-New-POSIX-permission-handling-tp117406p117615.html
Sent from the Cygwin list mailing list archive at Nabble.com.

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple