On 01/22/2010 09:44 AM, Daniel Veillard wrote:
On Thu, Jan 21, 2010 at 11:00:58AM -0600, Jamie Strandboge wrote:
> On Thu, 2010-01-21 at 11:33 -0500, Chris Lalancette wrote:
>> Signed-off-by: Chris Lalancette <clalance(a)redhat.com>
>> ---
>> .x-sc_prohibit_readlink | 2 ++
>> cfg.mk | 5 +++++
>> 2 files changed, 7 insertions(+), 0 deletions(-)
>> create mode 100644 .x-sc_prohibit_readlink
>
> This breaks AppArmor (see why in my response to the AppArmor change).
> Readlink() can be used safely, so perhaps the check can be done such
> that if using readlink, you must check the return code. Or simply warn
> if using readlink.
>
> virFileResolveLink() behaves substantially differently than readlink()
> and deprecating readlink() without adjusting virFileResolveLink() is
> IMHO unwise (while AppArmor is the only thing affected atm, it seems at
> least possible that new future code may need/want to readlink() things
> in /proc (eg /proc/self/exe)).
>
> Jamie
I understand this as being resolved by gnulib implementation, in which
case ACK to the make check addition,
Right, exactly. Thanks, pushed now.
--
Chris Lalancette