08.02.2014 06:31, Keith Packard wrote:
Egbert Eich <[email protected]> writes:
It would be nice if someone took a look at this issue. So far
noone has commented on this patch.
I'd prefer someone who actually understands IPV6 to review this.
I have looked, and I have some remarks.
First, there is an almost identical copy-paste between the two places
where entries are added to linklocal_list. Maybe it's worth making that
a function? And why the difference between IN6_IS_ADDR_LINKLOCAL in one
place and +IN6_IS_ADDR_LINKLOCAL in the other?
Second, that piece doesn't deal with realloc failures.
Third, the commit message does not reassure the readers that XDMCP over
link-local addresses works at all (I have not tried this) and is a
supportable configuration. If any of these conditions are false, then
let's just drop these addresses, like we do with IN6_IS_ADDR_LOOPBACK.
... OK, actually I have tested IPv6 UDP communication using link-local
addresses using netcat6. It doesn't work (-EINVAL on bind() that is used
to set the source address and port), so I will be very much surprised if
XDMCP works. If that requirement did not violate the list rules, I would
even require a tcpdump capture of a successful XDMCP session and some
X11 traffic along with the patch.
--
Alexander E. Patrakov
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
