[AG-TECH] [Patch] Big Linux fixup patch
csp at csperkins.org
Fri Nov 26 03:52:42 CST 2004
On 26 Nov 2004, at 01:02, Steve Smith wrote:
>> We have an AMD64 Shuttle cube on order, so this is great news!
> Yeah, cute aren't they. However the 64-bit port isn't perfect yet, I'm
> getting choppy sound off the network; I suspect there are other 32-bit
> centric assumptions in the Rat code I haven't tracked down yet.
Probably (although it did used to work on 64 bit Solaris/SPARC, so I'm
a little surprised you're finding so many problems - I guess the Linux
64 bit ABI is somehow different to the Solaris one).
>> Have you ascertained whether new patched rats are completely
>> compatible with current rats? I'm guessing mbus changes won't matter
>> because its used for communications between rat threads on the same
> That's correct. The mbus is trivial to test as it either works or it
> doesn't (and it didn't :)
>> On the other hand, RTP packets are being exchanged between all rats
>> in a particular session, so structure changes may be a problem, even
>> 32bit old <-> 32bit new.
> The structural changes are purely internal to let the compiler know
> it actually wants. I've tested it with unencrypted packets on the live
> AG and it seems to be working fine.
It looks fine - I merged it into the main RAT CVS.
> I haven't tested it in an encrypted session yet, but I'll see if I can
> have a look at that next week. However the encryption system within
> the current AG framework is of minimal value (or even negative value,
> as it gives a false sense of security where none exists) so it's not a
> high priority. Note I haven't touched the Rijndael code in Rat (yet).
The Rijndael code is broken. I have patches waiting to integrate which
make it pass the official AES test suites, at least. I'll try to merge
them over the next few days (into the UCL CVS, which has a lot of fixes
over the last release version, and that used in AG).
More information about the ag-tech