nscache6 - new command for IPv6 addresses cache
nsrecord supports IPv6 addresses
dnsauth should work with IPv6 servers and clients
+ caching is now more accurate
ACL handling corrected (affected all ACL-based operations like
allow/deny/banlimin/etc).
Network ranges like
10.0.0.0-10.0.0.1.64
are winally supported in ACLS.
Supports _real_ name resolution via parent proxy over TCP in nserver /
dnspr.
Can be used to create UDP->TCP name proxy, e.g.
nserver 127.0.0.1/tcp
nscache 65535
auth none
dnspr
auth iponly
tcppm 53 77.88.8.8 53
now, all names are resolved via TCP connection to 77.88.8.8
(dns.yandex.ru).
! avoid errno usage due to compatibility issues (and never use errno as
a lvalue). socks changed to use internal error codes instead.
! NOIPV6 compilation issues solved
! -lite version is switched to NOIPV6
There's no way to tell whether the architecture has its stdint.h, as
__UINT_FAST64_TYPE__ built in define didn't exist yet. Redefining the
types would break the build, and chances are we're not on some exotic
architecture where stdint would be missing, so let's assume it exists.
The buf gets copied to hostname and is present verbatim in logs looking
like pure random garbage. Moreover you can't really strlen() it.
This was likely some silly typo in inet_ntoa -> inet_ntop conversion for
ipv6, since it used to be correct before.
This used to look pretty random :(
Try to keep the last decimal digits of internal errors sensible for v5
error codes. For most cases we can't really do much better than using the
value of one.
One specific code paths where we can return something more useful is the
authentication, where we not communicate the underlying error code or
signal an ACL denial.
As for v4, there's not really much space to be too smart as only the code
of 91 is meaningful. Therefore we limit the last digit to zero and one now.
