Mail Archives: cygwin-developers/2002/09/30/13:39:52
"Christopher Faylor" <cgf AT redhat DOT com> wrote:
> I notice that the code in cygserver creates some objects with
> the default security rather than using something like &sec_none_nih.
> Is that intentional?
The security code in cygserver is much as I inherited it, except for
some bits that I've temporarily ripped out. My intention has been to
finish the whole System V IPC coding and then do the security as one
sweep afterwards (as most of the code will be common to all three
subsystems).
My impression about the existing security code is that a lot of stuff is
left wide open for the moment (i.e. for debugging purposes). For the
moment I'll fix the shared object creations to use the standard cygwin
approach as you suggest.
> I also didn't touch the many uses of \n terminated debug_printf's et
al,
> nor did I remove GetLastError from said calls, since I wasn't sure
when
> the code was supposed to run stand-alone.
I thought that the standalone *_printf's went through the smallprint too
(and so %E would work either way) but I now see that they don't. I'll
fix that and then use %E as appropriate.
I also thought I'd fixed all the trailing \n's, but obviously not if
there are still many of them abounding, so I'll take another sweep over
those.
> If it makes sense, I would appreciate it if someone (Conrad?) could
take
> a sweep over the code, use correct attributes in object (event, mutex,
> semaphore) creation, eliminate the \n from the printf output, and use
%E
> in place of GetLastError.
Watch this space etc.
// Conrad
- Raw text -