diff options
author | Steve Huston <shuston@riverace.com> | 2007-02-22 16:48:44 +0000 |
---|---|---|
committer | Steve Huston <shuston@riverace.com> | 2007-02-22 16:48:44 +0000 |
commit | be3bd76766a72d0445d9d7cb561ca60e7df831be (patch) | |
tree | cf98c7352f6940839b6ee11a7fb21f576faabdeb | |
parent | 9c3579fa02a0b9ea98df6a01ba74a9072847a8cc (diff) | |
download | ATCD-be3bd76766a72d0445d9d7cb561ca60e7df831be.tar.gz |
ChangeLogTag:Thu Feb 22 16:45:30 UTC 2007 Steve Huston <shuston@riverace.com>
-rw-r--r-- | ACE/ChangeLog | 9 | ||||
-rw-r--r-- | ACE/tests/MT_SOCK_Test.cpp | 10 |
2 files changed, 15 insertions, 4 deletions
diff --git a/ACE/ChangeLog b/ACE/ChangeLog index ce697e59013..ee18d7a7472 100644 --- a/ACE/ChangeLog +++ b/ACE/ChangeLog @@ -1,3 +1,12 @@ +Thu Feb 22 16:45:30 UTC 2007 Steve Huston <shuston@riverace.com> + + * tests/MT_SOCK_Test.cpp: Document the actual cause of what's been + observed for some time with some Windows systems exhibiting a + "delayed rejection" of connections in excess of the listener's + backlog. Thank you to Aleksandar Vukajlovic <vukajlo at finsoft + dot co dot yu> for citing the full cause with its supporting + documentation and explanation. + Thu Feb 22 15:59:12 UTC 2007 Steve Huston <shuston@riverace.com> * ace/config-win32-common.h: Change the comments around ACE_HAS_PENTIUM diff --git a/ACE/tests/MT_SOCK_Test.cpp b/ACE/tests/MT_SOCK_Test.cpp index 11220d36e29..adfd3dc58ee 100644 --- a/ACE/tests/MT_SOCK_Test.cpp +++ b/ACE/tests/MT_SOCK_Test.cpp @@ -101,10 +101,12 @@ client (void *arg) { // This is, I believe, more of an issue with WinXP-64 _server_ // side, but we can trap it here since we know we're connecting - // to localhost. It appears, though I haven't found documentation - // stating, that WinXP-64 will appear to accept connections at the - // TCP level past the listen backlog but if data arrives before the - // actual application-level accept() occurs, the connection is reset. + // to localhost. Some Windows versions will appear to accept + // connections at the TCP level past the listen backlog but if + // data arrives before the actual application-level accept() occurs, + // the connection is reset. This is caused when we trip the Windows + // SYN attack prevention (http://technet2.microsoft.com/WindowsServer/ + // en/library/910c8482-e5e5-4e2c-9ea4-11301ddfc4661033.mspx?mfr=true) // So, if we get a reset on the first send, don't flag the error - // just note it and act like the connection was refused. if (c == ACE_ALPHABET && errno == ECONNRESET) // First byte sent |