permissions and ownership of /usr/src

[Arthur Chance]

On 12/05/2015 01:57, Michael Powell wrote:
> Lowell Gilbert wrote:
>
>> Polytropon <freebsd at edvax.de> writes:
>>
>>> On Sun, 10 May 2015 15:03:44 -0400, David Banning wrote:
>>>> Can anyone tell me what the permissions - ownership are/should
>>>> be for /usr/src ?
>>>
>>> /usr/src should belong to root:wheel and have rwxr-xr-x permissions
>>> (those are the defaults). Does mergemaster indicate an error for
>>> /usr/src itself or for a subdirectory thereof?
>
> I was wondering about this too, e.g. is the permission error in /var/tmp/
> when mergemaster is trying to create the temproot directory tree under
> /var/tmp, and reflecting back through mergemaster when it fails. Somehow, I
> don't really think it likely, as a time or two I've had a problem like this
> mergemaster was very explicit about what was wrong.

I have a note to myself in one of my machine's /etc/fstab that reads

# Note: whereas noexec,nosuid seems like a good idea on /tmp,
# make -C /usr/src installworld creates programs under /tmp that
# it uses to do the install, so fails with the secure options.

As I've switched to using freebsd-update lately this is irrelevant for 
me now, but maybe this is what's hitting the OP. ISTR that mergeemaster 
does something with make in /usr/src.

>> Or, if the bandwidth isn't a big deal, remove the whole tree and do a
>> fresh "svn checkout". Or remove the directory where the error is
>> happening, and do an "svn update".
>
> Just something to mention in case anyone is interested: When I cd to
> /usr/src and do rm -rf * it will delete everything except the hidden
> directory .svn. Operation not permitted. In order to delete this internal
> svn housekeeping directory one must first chflags -R noschg .svn
>
> Then one can rm -rf .svn normally, and /usr/src will be truly empty. If
> someone can blow away /usr/src without seeing the operation not permitted
> error on the .svn directory, something is wrong with the flags under
> /usr/src/.svn




-- 
Those who do not learn from computing history are doomed to
GOTO 1