|
| 1 | +Thu Dec 14 23:53:41 2017 NAKAMURA Usaku < [email protected]> |
| 2 | + |
| 3 | + * test/net/ftp/test_ftp.rb (process_port_or_eprt): merge a part of |
| 4 | + r56973 to pass the test introduced at previous commit. |
| 5 | + |
| 6 | +Thu Dec 14 22:55:05 2017 Shugo Maeda < [email protected]> |
| 7 | + |
| 8 | + Fix a command injection vulnerability in Net::FTP. |
| 9 | + |
| 10 | +Thu Dec 14 22:35:19 2017 Eric Wong < [email protected]> |
| 11 | + |
| 12 | + webrick: compile RE correctly for beginning and end match |
| 13 | + |
| 14 | + Using ^ and $ in regexps means we can accidentally get fooled |
| 15 | + by "%0a" in HTTP request paths being decoded to newline |
| 16 | + characters. Use \A and \z to match beginning and end-of-string |
| 17 | + respectively, instead. |
| 18 | + |
| 19 | + Thanks to mame and hsbt for reporting. |
| 20 | + |
| 21 | + * lib/webrick/httpserver.rb (MountTable#compile): |
| 22 | + use \A and \z instead of ^ and $ |
| 23 | + * lib/webrick/httpserver.rb (MountTable#normalize): use \z instead of $ |
| 24 | + * test/webrick/test_httpserver.rb (test_cntrl_in_path): new test |
| 25 | + |
| 26 | +Thu Dec 14 22:29:04 2017 Eric Wong < [email protected]> |
| 27 | + |
| 28 | + webrick: do not hang acceptor on slow TLS connections |
| 29 | + |
| 30 | + OpenSSL::SSL::SSLSocket#accept may block indefinitely on clients |
| 31 | + which negotiate the TCP connection, but fail (or are slow) to |
| 32 | + negotiate the subsequent TLS handshake. This prevents the |
| 33 | + multi-threaded WEBrick server from accepting other connections. |
| 34 | + |
| 35 | + Since the TLS handshake (via OpenSSL::SSL::SSLSocket#accept) |
| 36 | + consists of normal read/write traffic over TCP, handle it in the |
| 37 | + per-client thread, instead. |
| 38 | + |
| 39 | + Furthermore, using non-blocking accept() is useful for non-TLS |
| 40 | + sockets anyways because spurious wakeups are possible from |
| 41 | + select(2). |
| 42 | + |
| 43 | + * lib/webrick/server.rb (accept_client): use TCPServer#accept_nonblock |
| 44 | + and remove OpenSSL::SSL::SSLSocket#accept call |
| 45 | + * lib/webrick/server.rb (start_thread): call OpenSSL::SSL::SSLSocket#acc |
| 46 | +ept |
| 47 | + * test/webrick/test_ssl_server.rb (test_slow_connect): new test |
| 48 | + [ruby-core:83221] [Bug #14005] |
| 49 | + |
| 50 | + webrick: fix up r60172 |
| 51 | + |
| 52 | + By making the socket non-blocking in r60172, TLS/SSL negotiation |
| 53 | + via the SSL_accept function must handle non-blocking sockets |
| 54 | + properly and retry on SSL_ERROR_WANT_READ/SSL_ERROR_WANT_WRITE. |
| 55 | + OpenSSL::SSL::SSLSocket#accept cannot do that properly with a |
| 56 | + non-blocking socket, so it must use non-blocking logic of |
| 57 | + OpenSSL::SSL::SSLSocket#accept_nonblock. |
| 58 | + |
| 59 | + Thanks to MSP-Greg (Greg L) for finding this. |
| 60 | + |
| 61 | + * lib/webrick/server.rb (start_thread): use SSL_accept properly |
| 62 | + with non-blocking socket. |
| 63 | + [Bug #14013] [Bug #14005] |
| 64 | + |
| 65 | + webrick: fix up r60172 and revert r60189 |
| 66 | + |
| 67 | + Thanks to MSP-Greg (Greg L) for helping with this. |
| 68 | + |
| 69 | + * lib/webrick/server.rb (start_thread): ignore ECONNRESET, ECONNABORTED, |
| 70 | + EPROTO, and EINVAL on TLS negotiation errors the same way they |
| 71 | + were ignored before r60172 in the accept_client method of the |
| 72 | + main acceptor thread. |
| 73 | + [Bug #14013] [Bug #14005] |
| 74 | + |
| 75 | + webrick: fix up r60172 and r60208 |
| 76 | + |
| 77 | + Thanks to MSP-Greg (Greg L) for helping with this. |
| 78 | + |
| 79 | + * lib/webrick/server.rb (start_thread): fix non-local return |
| 80 | + introduced in r60208 |
| 81 | + |
| 82 | + webrick: fix up r60172 and r60210 |
| 83 | + |
| 84 | + Thanks to MSP-Greg (Greg L) for helping with this. |
| 85 | + |
| 86 | + * lib/webrick/server.rb (start_thread): properly fix non-local return |
| 87 | + introduced in r60208 and r60210 |
| 88 | + |
| 89 | +Thu Nov 30 23:37:08 2017 Nobuyoshi Nakada < [email protected]> |
| 90 | + |
| 91 | + parse.y: fix line in rescue |
| 92 | + |
| 93 | + * parse.y (set_line_body, primary): fix line number of bodystmt as the |
| 94 | + beginning of the block. [Bug #13181] |
| 95 | + |
| 96 | +Thu Nov 30 23:29:00 2017 SHIBATA Hiroshi < [email protected]> |
| 97 | + |
| 98 | + Merge rubygems-2.6.14 changes. |
| 99 | + |
| 100 | + It fixed http://blog.rubygems.org/2017/10/09/unsafe-object-deserialization-vulnerability.html |
| 101 | + |
| 102 | +Fri Sep 15 05:40:40 2017 URABE Shyouhei < [email protected]> |
| 103 | + |
| 104 | + fix --with-gmp (broken by r57490) |
| 105 | + |
| 106 | + Looking at the generated shell script (also the autoconf manual), it |
| 107 | + seems AC_SEARCH_LIBS() m4 macro does not define HAVE_LIBsomething C |
| 108 | + preprocessor macros, unlike AC_CHECK_LIB() which does define them. |
| 109 | + This previous change effectively killed building with GMP because |
| 110 | + building that mode depends on existence of HAVE_LIBGMP. [Bug #13402] |
| 111 | + |
1 | 112 | Thu Sep 14 20:25:55 2017 Yusuke Endoh < [email protected]> |
2 | 113 |
|
3 | 114 | lib/webrick/log.rb: sanitize any type of logs |
|
0 commit comments