Wayne Davison [Mon, 21 Jul 2008 03:08:08 +0000 (20:08 -0700)]
Added a debug-helping option, --msgs2stderr, than should help all
messages to be seen in a situation where rsync is dying (as long
as stderr is a viable output method for the remote rsync).
Wayne Davison [Sat, 19 Jul 2008 15:49:53 +0000 (08:49 -0700)]
Changed the POOL_QALIGN flag to POOL_NO_QALIGN, reversing the setting
(making pools aligned by default). Added the missing code to make the
documented behavior of pool_free() with a NULL addr work. Updated the
pool_alloc.3 manpage.
Wayne Davison [Fri, 11 Jul 2008 16:34:28 +0000 (09:34 -0700)]
If the user specifies --protocol=29, rsync will avoid sending an -e
option to the server (which is only useful for protocols 30 and above
anyway). This gives the user an easy way to talk to a restricted
server that has overly restrictive option-checking.
Wayne Davison [Sat, 5 Jul 2008 07:21:05 +0000 (00:21 -0700)]
The --progress output now leaves the cursor at the end of the line
(instead of the start) in order to be extra sure that an error won't
overwrite it. We also ensure that the progress option can't be enabled
on the server side.
Wayne Davison [Tue, 24 Jun 2008 06:05:07 +0000 (23:05 -0700)]
Some deny-rsync fixes:
- Fixed messages longer than 63 chars.
- Don't require the presence of a "bc" program.
- Append a newline to the message to make the script easier to call.
- Make extra sure the message outputs without escape interpretation.
- Stick around long enough for a client to reliably get the message.
Wayne Davison [Mon, 9 Jun 2008 04:07:30 +0000 (21:07 -0700)]
A couple more xattr improvements:
- Made the XSTATE_* defines avoid using 0.
- Call !XATTR_ABBREV() in recv_xattr_request().
- Improved the "internal abbrev" error message.
- Fixed the potential for a directory time glitch in xattrs.diff.
Wayne Davison [Mon, 9 Jun 2008 03:26:22 +0000 (20:26 -0700)]
Fixed an "Internal abbrev error" when dealing with an xattr value
that is unchanged on an early file, and changed on a later file.
Added 2 new test cases to ensure this stays fixed.
Wayne Davison [Wed, 4 Jun 2008 14:13:22 +0000 (07:13 -0700)]
Improved the proto.h target so that a build in a different dir from the
srcdir will ensure that the builddir has a copy of the proto.h file when
the Makefile found it to be out of date. This prevents the repeated
building of all the targets when the srcdir's proto.h file is accurate,
but older than the newest .c file.
Wayne Davison [Sat, 31 May 2008 21:41:20 +0000 (14:41 -0700)]
Have send_file_list() check is_excluded() (but only on non-dot-dirs)
and then call send_file_name() with NO_FILTERS. This gets rid of
the need for a FLAG_DOTDIR_NAME flag (used only by make_file()).
Wayne Davison [Sat, 31 May 2008 16:08:13 +0000 (09:08 -0700)]
Make sure the generator doesn't try to send a negative checksum count to
the sender (which would cause it to die with a cryptic error). Instead,
warn the user when the file's size is too large for checksum processing.
Wayne Davison [Tue, 6 May 2008 17:39:19 +0000 (10:39 -0700)]
Added options to tls.c to allow us to ask for mtime and ownership info
on symlinks. The testsuite will now pass these options to tls if rsync
is configured to affect such attributes on symlinks.
Wayne Davison [Tue, 29 Apr 2008 04:17:36 +0000 (21:17 -0700)]
- Changed prev_name into a buffer so that there is no chance that its
contents can be overwritten by other calls to f_name().
- Changed an hlink assert into a check that provides more debug info.
Wayne Davison [Sat, 19 Apr 2008 02:55:08 +0000 (19:55 -0700)]
Reorder the filenames to touch to try to avoid a weird error on Solaris
5.8. Also, use lsh in one of the runs in order to try a hard-link run
that uses a (pretend) remote shell.
Wayne Davison [Fri, 18 Apr 2008 14:20:30 +0000 (07:20 -0700)]
Added a --no-cd option to support/lsh so that the script can be used by
the testsuite. Improved the home-directory-changing code and added an
error message when "localhost" is not the hostname specified. Use the
updated script in the testsuite instead of creating a pretend-ssh script
in a couple spots.
Wayne Davison [Tue, 15 Apr 2008 15:39:59 +0000 (08:39 -0700)]
When running in --progress mode with a progress message active, the
client now outputs a newline prior to an error message, which avoids
overwriting the active file's last progress line.