Stephen Smalley wrote:
> On Wed, 2005-04-06 at 16:38 -0500, Darrel Goeddel wrote:
> 
>>Here is a version that I am happy with.  There were only a few minor fixes from
>>the past patch.  As before, this patch is relative to Joshua's
>>hierarchy-backport patch, and it should be applied when that patch is applied to
>>keep mls processing working in the policy compiler.  I haven't really looked
>>over Joshua's patch with a fine tooth comb, but it sure has been working nice
>>for me.  Anybody see anything wrong with the mls patch (or suggestions)?
> 
> 
> The original hierarchy patch also collapsed the identifier and
> user_identifier together, thereby allowing "-" to occur in any
> identifier.  As a result, if someone specifies s0-s9 in the policy
> without whitespace, it will be incorrectly interpreted as an attempt to
> specify a level named "s0-s9".  Further, nothing prevents someone from
> defining a level or category name that includes a "-" presently.
> Options are to revert the change from the original patch that collapsed
> identifier and user_identifier together (only adding "." to identifier,
> not "-") or to add further handling to the action routines to deal with
> it.
> 

That sure could lead to some confusion.  I am ok with either remedy.  Does
anyone have a good argument for one approach over the other?

> It also looks like you have double free error upon encountering an
> undefined sensitivity name in a user declaration.

Are you talking about the case where a unknown sensitivity is encountered and
ignored  (continue statement...).  If so, I think the correct behavior is to
return -1.  We should not be ignoring things that are not defined anyway.  I
also found that I was freeing the id before it was possibly used in the error 
case - I have fixed those.  If this is not the error you mentioned, could you
please elaborate (I apologize for not being able to see it).  Attached is a
version of the patch that addresses these issues (not the '-' issue yet).  Is
the bug still there?

> As a minor note, I suppose that the 'make mlsconvert' should also change
> MLS=y in the Makefile so that you don't have to do that as an extra
> step.

I can add that to the next version of this patch since I'm already hitting the
mlsconvert target.

-- 

Darrel