Commit Graph

107 Commits

Author SHA1 Message Date
Charles Kerr cb79214854 (trunk) remove trailing spaces 2009-08-10 20:04:08 +00:00
Eric Petit 67440b9286 Don't assume that errnos are positive (they are not on Haiku), and
pass them on through an additional parameter if needed
2009-07-09 18:14:33 +00:00
Charles Kerr 8e6eee683f (trunk) on non-mac platforms, try not building our own copy of libevent 1.4.x 2009-06-03 15:47:08 +00:00
Charles Kerr 74a4c2f2d5 fix last commit :/ 2009-05-15 12:28:12 +00:00
Charles Kerr 5bf4eb71ce (trunk libT) fix a minor bug in peer count tabulation 2009-05-15 12:19:27 +00:00
Charles Kerr 7599172836 (trunk libT) fix minor r8394 bug 2009-05-14 18:37:24 +00:00
Charles Kerr da42fe8a7c (trunk libT) #2035: Transmission causes wakeups by unnecessary polling. 2009-05-14 13:42:29 +00:00
Charles Kerr b2a6def953 (trunk libT) fix small dead code. found by clang static analyzer and reported by d0k 2009-05-13 19:04:14 +00:00
Erick Turnquist 2719372bc6 (trunk) #1497 Options to listen on specific network sockets 2009-04-15 21:05:58 +00:00
Charles Kerr a2b53f5cbd (trunk libT) more speed tweaks from valgrind 2009-04-12 14:52:17 +00:00
Erick Turnquist 2f8d56e77f (trunk libT) make IPv6's two IPv4-compatible address formats invalid peer addresses
IPv4 peers should only be seen from an IPv4-connected tracker.
2009-03-27 01:31:17 +00:00
Erick Turnquist dd7eb76718 (trunk libT) remove tr_suspectAddress() since all seems well 2009-03-27 01:14:25 +00:00
Erick Turnquist 72297dbdb4 (trunk libT) suppress ipv6 EHOSTUNREACH as well as ENETUNREACH 2009-02-08 22:52:47 +00:00
Erick Turnquist fc767a3f03 (trunk libT) don't attempt to connect to (half-complete) ipv6 link-local addresses 2009-02-08 22:45:30 +00:00
Erick Turnquist 61e505b8d1 (trunk libT) don't attempt to set V6ONLY on ipv4 sockets 2009-02-07 01:14:41 +00:00
Erick Turnquist 94b98030cb (trunk libT) stop using ipv4-mapped listener sockets. this should have the added effect of fixing #1789 2009-02-07 00:34:10 +00:00
Charles Kerr 9c9c94a81d (trunk libT) discard peers from the tracker or pex that have a port of 0 or a multicast address. 2009-01-17 23:14:35 +00:00
Charles Kerr d95a45b701 (trunk libT) fix IPv6 reported & solved by erdgeist. add debugging message to help track down any future problems. 2009-01-16 23:17:31 +00:00
Charles Kerr bc2b7fcb03 (trunk libT) minor tweaks to reduce libtransmission shear between trunk and 1.5x 2009-01-16 17:29:42 +00:00
Charles Kerr 2f6b56671f (trunk libT) win32 portability patch from Alexey 2009-01-11 17:02:04 +00:00
Charles Kerr c6a454d2e2 (trunk libT) win32 portability fixes, reported by Alexey 2009-01-10 02:22:13 +00:00
Erick Turnquist 5086cbe47c (trunk libT) use IN_MULTICAST macro 2009-01-02 21:45:55 +00:00
Charles Kerr 0cd572a0da (trunk libT) make tr_isAddress() and tr_isDirection() inline 2009-01-02 04:47:37 +00:00
Erick Turnquist e9f32d7523 (trunk libT) < Rolcol> For example: I have a lot of addresses in the message
log as "funny looking address" as something like ::ffff:192.168.1.1

These were coming from pex6. Apparently there is a client out there that sends
us IPv4 peers in mapped form. It's valid, but a little strange...
2009-01-01 00:34:44 +00:00
Erick Turnquist bd5463a37e (trunk libT) fix another endian oops 2009-01-01 00:23:09 +00:00
Erick Turnquist 37ac4b3fbf (trunk libT) fix endian oops in tr_suspectAddress() 2009-01-01 00:15:15 +00:00
Erick Turnquist 50d73a884f (trunk libT) Stop trying to connect to multicast addresses. Also, bump "funny
looking address" messages from tr_dbg to tr_err so I can get bug reports with
the message for once!
2008-12-31 21:25:52 +00:00
Erick Turnquist 38efb62100 add some debugging stuff to track down where some bogus addresses are coming from 2008-12-21 19:13:52 +00:00
Erick Turnquist a68764fcba (trunk libT) Don't log two "errors" we expect to see from time-to-time.
EAFNOSUPPORT: On OS X, socket() incorrectly throws EAFNOSUPPORT for certain
IP/Port combinations. There is nothing we can do about this.

ENETUNREACH: #1606 is evidence that some trackers return IPv6 peers when
contacted over IPv4. This code will silently ignore "network unreachable"
errors for IPv6 connections.
2008-12-20 08:51:32 +00:00
Erick Turnquist 72ee19dcca (trunk libT) Refactor ipv6 bind socket check
This will let the RPC server use the same test mechanism eventually. Networking code belongs in net.c anyway.
2008-12-18 05:55:22 +00:00
Erick Turnquist 2d427751d6 fix another bad assert *eyes charles* 2008-12-16 21:44:53 +00:00
Erick Turnquist 0a2c5efb65 Fix "ISO C90 forbids mixed declarations and code" 2008-12-16 21:35:14 +00:00
Erick Turnquist 86257e03f1 Fix tr_isAddress() 2008-12-16 21:28:06 +00:00
Charles Kerr de5228c415 (trunk libT) more runtime safety checks 2008-12-16 21:06:47 +00:00
Charles Kerr 748a9f1c92 (trunk libT) add some assertions to tr_compareAddresses() 2008-12-16 06:46:27 +00:00
Charles Kerr 54be88dec3 (trunk libT) add ipv6 support by jhujhiti. I think this is the largest user-contributed patch we've ever used... thanks jhujhiti :) 2008-12-15 00:17:08 +00:00
Mitchell Livingston 31482813b9 BeOS and Amiga support is, at least for now, a lost cause and hasn't worked for long over a year - removing from the source (if you for whatever unknown reason still need this, get it through SVN) 2008-12-02 03:57:01 +00:00
Charles Kerr 0d5fed6442 (libT) re-apply jhujhiti's IPv6 patch. This merges in my tr_port cleanup, so any new bugs are mine :/ 2008-12-02 03:41:58 +00:00
Charles Kerr a14621c363 (libT) janitorial: use the tr_port type whenever we have a `port' as a function argument or as a field in a struct. 2008-12-01 20:51:01 +00:00
Charles Kerr 7fa9813955 (libT) #252: revert r7195 (jhujhiti's IPv6 test patch). It seems to be the culprit in a `thrashing' issue reported by Waldorf and m1b in irc -- the behavior appeared between r7183 and r7187 -- so this commit is to trigger off a nightly build w/o the patch to test with in irc. 2008-12-01 20:21:06 +00:00
Charles Kerr a2ad4f6543 (libT) patch from jhujhiti to add ipv6 support. 2008-11-30 00:47:18 +00:00
Charles Kerr 40299a2284 (libT) silence the net.c warnings I introduced yesterday :/ 2008-11-29 21:34:06 +00:00
Charles Kerr 516ba0564a (libT) disable the so_sndbuf / so_rcvbuf code for now 2008-11-28 07:42:08 +00:00
Charles Kerr f44248a7d7 (libT) yet another stab at getting bandwidth management under control. this version may suck less than previous attempts. It also breaks the mac build until someone adds iobuf.[ch] to xcode... 2008-11-25 21:35:17 +00:00
Charles Kerr 8e6529ad8f (libT) I wonder if this will have any affect on the bug reported by Wizardling. 2008-11-24 21:59:00 +00:00
Charles Kerr be5e6774ad (libT) #1468: another stab at getting the peer transfer speeds both fast and a little more consistent. 2008-11-24 04:21:23 +00:00
Charles Kerr 853a67fbfc (libT) define peer connections' sockets' so_sndbuf size in the tr_session struct. 2008-11-21 16:32:55 +00:00
Charles Kerr 93b3dc5617 increase the socket's SNDBUF size 2008-11-16 21:16:14 +00:00
Charles Kerr 84dcaab5f4 remove dead code 2008-10-22 13:57:30 +00:00
Charles Kerr b5c2f14387 (libT) dead code removal 2008-10-14 20:44:41 +00:00