- The receiver now sends keep-alive messages to the generator
when it is actively doing work and hasn't sent anything
recently. This ensures that the generator won't timeout
if the receiver is working hard.
- The perform_io() code has improved keep-alive participation.
- Allow the sender to send some keep-alive messages, which
ensures that if it is in a lull, it can probe the socket.
The receiving side also switches timeout handling from the receiver to
the generator, which obviates the need for the sender to send any
keep-alive messages at all (for protocol 31 and beyond). Given this
setup, all keep-alive messages are now sent as empty MSG_DATA messages,
with MSG_NOOP messages only being understood and (when necessary) acted
upon to forward a keep-alive event to an older receiver. This is both
safer and more compatible with older versions.
- The receiver notifies the generator if it is exiting with an error,
and then, if it is a server, waits around for the generator to die.
This ensures that the client side has time to read the error.
- The generator or sender will notifiy the other side of the transfer of
an error-exit value if protocol 31 is in effect. This will get rid of
some "connection unexpectedly closed" errors that are really expected
events due to a fatal exit on the other side.
Files-from data is now sent as multiplexed I/O so that it can mingle
with any messages (such as debug output). Requires protocol 31.
Protocol 31 no longer disables output verbosity in a couple instances
that used to cause protocol issues.
Got rid of MSG_* messages that have implied raw data that follows after
them. We instead send a negative index value as a part of the raw data
stream, which is guaranteed to be output together with the following
data. This only affects the (in-progress) protocol 31 and the (self-
contained) communication stream from the receiver to the generator.
Added --debug=IO and improved --debug=FLIST. Some --debug=IO output
requires --msgs2stderr to be used to see it (i.e. sending a message
about sending a message would send another message, ad infinitum).
- Improve function name: parse_rule -> parse_filter_str (to make the
similarity with parse_filter_file clearer, and better indicate that
it can parse multiple rules when FILTRULE_WORD_SPLIT is specified).
- In preparation for rule prefixes containing information beyond the
rflags, change the code to pass around a full "template" filter_rule
instead of just rflags. Callers of parse_filter_{str,file} that want
to specify only rflags can use rule_template(rflags) .
- Remove the MODIFIERS_* strings and instead hand-code the condition
under which each modifier is valid. This should make it easier to
see that the conditions are correct.
- Tighten up default modifiers on merge rules:
- Disallow "!" because it isn't useful.
- If the merge rule specifies a side via "s" or "r", the rules in the
file cannot also specify a side via "s", "r", "hide", etc.
[Patch was changed by Wayne a bit prior to application.]
Since the value is not needed, protocol 31 no longer sends it, while
older protocols are optimized so the sender just sends a valid rdev
value as efficiently as possible. The receiver no longer caches an
rdev value for special files, and the generator will always pass a 0
rdev value to do_mknod() for special files. Fixes bug #6280.
- Changed get_backup_name() to verify the backup path, and make any
missing directories. This avoids accidental use of a symlink as a dir
in a backup path, and gets rid of any other non-dirs that are in the
way. It also avoids the need for various operations to retry after
calling make_bak_dir(), simplifying several pices of code.
- Changed create_directory_path() to make_path(), giving it flags that
lets the caller decide if it should skip a leading slash or drop the
trailing filename.
- Mention when we create the backup directory, so the user is not caught
unaware when rsync uses a directory they didn't expect.
- Got rid of some dir-moving backup code that is not used.
- Added a little more backup-debug output.
- Backups do not interfere with an atomic update (when possible).
- Backing up a file will remove a directory that is in the way
and visa versa.
- Unify the backup-dir and non-backup-dir code in backup.c.
- Improved the backup tests a little bit.
- Mention how many files were created (protocol >= 29).
- Mention how many files were deleted (new in protocol 31).
- Follow the file-count, created-count, and deleted-count
with a break-out list of each count by type.
- Free a mergelist's parent_dirscanned filters the last time it is
popped or as soon as the filters are discarded due to the "n"
modifier. Aside from not leaking memory, this is needed to clean up
any mergelists defined during the parent_dirscan to avoid crashing by
trying to restore nonexistent state for them in pop_local_filters.
- Make push_local_filters save the current mergelist_cnt, and make
pop_local_filters assert that it has the correct number of mergelists
before restoring their state.
- Assert that mergelists get deactivated in strict LIFO order to catch
any glitches as soon as they happen. Free linked lists of filters in
reverse order to make that the case.
- Add a bunch of mergelist-related debug output (--debug=filter2).
- Renamed push_dir() to change_dir() and revised it a little so that it
can chdir() to a relative path without an intervening chdir() back to
the staring path.
- Renamed push_pathname() to change_pathname() and revised it to take
different args and to only call path_is_daemon_excluded() on a new
path (not a revisit of a file's already-checked path).
- Fixed change_pathname() to set the right pathname value when a chdir()
call fails.
- Set orig_dir once outside of the change_pathname() function.
- Got rid of pop_dir().
- Removed the CFN_KEEP_LEADING_DOT_DIR flag for clean_fname().
- Explicitly add an implied dot-dir to the transfer rather than keeping
a leading a "./" prefix as a part of a relative pathname.
- Added the CFN_KEEP_DOT_DIRS flag for clean_fname().
- Added the SP_KEEP_DOT_DIRS flag for sanitize_path().
- Call clean_fname() a couple more times.
the FLAG_SKIP_HLINK flag, which gets set on any hard-linked file that
the user wants to skip (e.g. via --ignore-existing, --append, etc.).
The code in hlink.c now deals with the skipped files instead of
triggering an assert() error.
- If a symlink/device/special-file changes its value without any
attribute changes, the itemized event no longer gets dropped.
- We put a 'c' into the checksum/change field now to indicate when
a symlink/device/special-file changes its value without changing
its type. This lets us properly interpret the --copy-links output
to know which items are getting copied without changes and which
are getting created with new content.
- Fixed the 'T' itemized output for a symlink when rsync tries to
set the right time but fails due to lack of OS/disk support.
