| 1 | NEWS for rsync 2.6.1 (UNRELEASED) |
| 2 | Protocol: 28 (changed) |
| 3 | Changes since 2.6.0: |
| 4 | |
| 5 | ENHANCEMENTS: |
| 6 | |
| 7 | * Lower memory use and more optimal transfer of data over |
| 8 | the socket (see the INTERNAL section for details). |
| 9 | |
| 10 | * The RSYNC_PROXY environment variable can now contain a |
| 11 | "USER:PASS@" prefix before the "HOST:PORT" information. |
| 12 | (Bardur Arantsson) |
| 13 | |
| 14 | * The --progress output now mentions how far along in the |
| 15 | transfer we are, including both a count of files transferred |
| 16 | and a percentage of the total file-count that we've processed. |
| 17 | It also shows better current-rate-of-transfer and remaining- |
| 18 | transfer-time values. |
| 19 | |
| 20 | * The configure script now accepts --with-rsyncd-conf=PATH |
| 21 | to override the default value of the /etc/rsyncd.conf file. |
| 22 | |
| 23 | * Added a couple extra diffs in the "patches" dir, removed the |
| 24 | ones that got applied, and rebuilt the rest. |
| 25 | |
| 26 | BUG FIXES: |
| 27 | |
| 28 | * When -x (--one-file-system) is combined with -L (--copy-links) |
| 29 | or --copy-unsafe-links, no symlinked files are skipped, even |
| 30 | if the referant file is on a different filesystem. |
| 31 | |
| 32 | * The --link-dest code now works properly for a non-root user |
| 33 | when (1) the UIDs of the source and destination differ and -o |
| 34 | was specified, or (2) when the group of the source can't be |
| 35 | used on the destination and -g was specified. |
| 36 | |
| 37 | * Fixed a bug in the handling of -H (hard-links) that might |
| 38 | cause the expanded PATH/NAME value of the current item to |
| 39 | get overwritten (due to an expanded-name caching bug). |
| 40 | |
| 41 | * We now reset the "new data has been sent" flag at the start |
| 42 | of each file we send. This makes sure that an interrupted |
| 43 | transfer with the --partial option set doesn't keep a shorter |
| 44 | temp file than the current basis file when no new data has been |
| 45 | transfered over the wire for that file. |
| 46 | |
| 47 | * Fixed a byte-order problem in --batch-mode on big-endian |
| 48 | machines. (Jay Fenlason) |
| 49 | |
| 50 | * Fixed configure bug when running "./configure --disable-ipv6". |
| 51 | |
| 52 | * Fixed "make test" bug when build dir is not the source dir. |
| 53 | |
| 54 | * When using --cvs-exclude, the exclude items we get from a |
| 55 | directory's .cvsignore file once again only affect that one |
| 56 | directory (and not all following directories too). |
| 57 | |
| 58 | * When specifying the USER@HOST: prefix for a file, the USER |
| 59 | part can now contain an '@', if needed (i.e. the last '@' |
| 60 | is used to find the HOST, not the first). |
| 61 | |
| 62 | * Fixed some bugs in the handling of group IDs for non-root |
| 63 | users: (1) It properly handles a group that the sender didn't |
| 64 | have a name for (it would previously skip changing the group |
| 65 | on any files in that group). (2) If --numeric-ids is used, |
| 66 | rsync no longer attempts to set groups that the user doesn't |
| 67 | have the permission to set. |
| 68 | |
| 69 | * Fixed the "refuse options" setting in the rsyncd.conf file. |
| 70 | |
| 71 | * Improved the -x (--one-file-system) flag's handling of any |
| 72 | mount-point directories we encounter. It is both more optimal |
| 73 | (in that it no longer does a useless scan of the contents of |
| 74 | the mount-point dirs) and also fixes a bug where a remaped |
| 75 | mount of the original filesystem could get discovered on a |
| 76 | disk we should be ignoring. |
| 77 | |
| 78 | * Rsync no longer discards a double-slash at the start of a filename |
| 79 | when trying to open the file. It also no longer constructs names |
| 80 | that start with a double slash (unless the user supplied them). |
| 81 | |
| 82 | * Fixed compilation problem on Tru64 Unix (having to do with |
| 83 | sockaddr.sa_len and sockaddr.sin_len). |
| 84 | |
| 85 | * Fixed a compatibility problem interacting with older rsync |
| 86 | versions that might send us an empty suffix without telling us |
| 87 | about the backup-dir. |
| 88 | |
| 89 | * The "hosts allow" option for a daemon-over-remote-shell |
| 90 | process now has improved support for IPv6 addresses and a fix |
| 91 | for systems that have a length field in their socket structs. |
| 92 | |
| 93 | * Fixed the ability to request an empty backup suffix when |
| 94 | sending files to an rsync daemon. |
| 95 | |
| 96 | INTERNAL: |
| 97 | |
| 98 | * Most of the I/O is now buffered, which results in a pretty |
| 99 | large speedup when running under MS Windows. (Craig Barratt) |
| 100 | |
| 101 | * Optimizations to the name-handling/comparing code have made |
| 102 | some significant reductions in user-CPU time for large file |
| 103 | sets. |
| 104 | |
| 105 | * Some variable-type cleanup that makes the code more consistent. |
| 106 | |
| 107 | * Reduced memory requirements of hard link preservation. |
| 108 | (J.W. Schultz) |
| 109 | |
| 110 | * Implemented a new algorithm for hard-link handling that speeds |
| 111 | up the code significantly. (J.W. Schultz and Wayne Davison) |
| 112 | |
| 113 | * The --hard-link option now uses the first existing file in the |
| 114 | group of linked files as the basis for the transfer. This |
| 115 | prevents the sub-optimal transfer of a file's data when a new |
| 116 | hardlink is added on the sending side and it sorts alphabetically |
| 117 | earlier in the list than the files that are already present on the |
| 118 | receiving side. |
| 119 | |
| 120 | * Dropped support for protocol versions less than 20 |
| 121 | (2.3.0 released 15 Mar 1999) and activated warnings for |
| 122 | protocols less than 25 (2.5.0 released 23 Aug 2001) |
| 123 | (Wayne Davison and J.W. Schultz, severally) |
| 124 | |
| 125 | * More optimal data transmission for --hard-links (protocol 28). |
| 126 | |
| 127 | * More optimal data transmission for --checksum (protocol 28). |
| 128 | |
| 129 | * Less memory is used when --checksum is specified. |
| 130 | |
| 131 | * Less memory is used in the file list (a per-file savings). |
| 132 | |
| 133 | * The generator is now better about not modifying the file list |
| 134 | during the transfer in order to avoid a copy-on-write memory |
| 135 | bifurcation (on systems where fork() uses shared memory). |
| 136 | Previously, rsync's shared memory would slowly become unshared, |
| 137 | resulting in real memory usage nearly doubling on the receiving |
| 138 | side by the end of the transfer. Now, as long as permissions |
| 139 | are being preserved, the shared memory should remain that way |
| 140 | for the entire transfer. |
| 141 | |
| 142 | * Changed hardlink info and file_struct + strings to use |
| 143 | allocation pools. This reduces memory use for large |
| 144 | filesets and permits freeing memory to the OS. (J.W. Schultz) |
| 145 | |
| 146 | * The 2 pipes used between the receiver and generator processes |
| 147 | (which are forked on the same machine) were reduced to 1 pipe |
| 148 | and the protocol improved so that (1) it is now impossible to |
| 149 | have the "redo" pipe fill up and hang rsync, and (2) trailing |
| 150 | messages from the receiver don't get lost on their way through |
| 151 | the generator over to the sender (the latter mainly affected |
| 152 | hard-link messages and verbose --stats output). |
| 153 | |
| 154 | * The reading & writing of the file-list in batch-mode is now |
| 155 | handled by the same code that sends & receives the list over |
| 156 | the wire. This makes it much easier to maintain. |
| 157 | |
| 158 | * Improved the internal uid/gid code to be more portable and |
| 159 | a little more optimized. |