- Release Notes for Bacula 2.2.7
+ Release Notes for Bacula 2.4.3
- Bacula code: Total files = 520 Total lines = 195,550 (*.h *.c *.in)
- 82 new files, 41,221 new lines of code, 208,380 lines of change from 2.0.3
+ Bacula code: Total files = 500 Total lines = 188,945 (*.h *.c *.in)
+
+================== Warning !!!!!! ==========================
+As of Version 2.4.0, the Win32 FD should not descend into
+any reparse point directory or reparse mount points unless the
+specified directory is explicitly mentioned at the top level of
+FileSet (as with Unix). A file that is linked to another file will
+be backed up, much like Unix hardlinked files. If you want
+reparse directories backed up, you must explicitly include them
+in your FileSet.
+==============================================================
+
+Release 2.4.3
+ This is a bug fix to version 2.4.2. All daemons are compatible
+ with other 2.4.x versions.
+
+Bugs Fixed:
+ 1159, 1162. 1149, 1128, 1143, 1140, 1129
+
+Change Summary:
+- Apply 2.4.2-cancel-non-running-jobs.patch to be able to cancel
+ a non created job faster.
+- This should correct bug #1159 where Migration does not properly
+ respect the Migration Low Bytes directive.
+- Fix typo in configure.in that breaks --with-db-password option
+- This code should fix the race condition that leads to a Director
+ crash at job end time when the job list is updated. This was reported
+ in bug #1162.
+- Remove all double quotes from SQLite creating script and
+ replace by single quotes as suggested by John Huttley.
+- Fix SQL case problem that may cause the failure of DiskToCatalog
+ in bug #1149.
+- Copy missing storage name into edit buffer. As far as I can tell
+ this never caused a bug.
+- Remove catalog dependency for bcopy
+- Modify catalog scripts to have an easier packaging integration,
+ using default variables.
+ Make difference between SQLite3 and SQLite in db_get_type()
+- Attempt to fix bug #1128 InChanger flag cleared during Migration
+ job when reading from one autochanger and writing to another.
+- Add more information to SD acquire.c INFO messages.
+- Fix NULL Volume name error when reading and the drive must
+ be switched.
+- Fix a Verify InitCatalog problem where in certain cases
+ a garbage filename may be entered in the verification database. This
+ fixes bug #1143.
+- Add space after version before date in bat about dialog.
+- Fix seg fault in Dir during estimate command with no level value
+ given. This fixes bug #1140.
+- Add message to migration job when the target job is already migrated.
+ This closes bug #1129.
+
+
+==============================================================
+
+Release 2.4.2
+ This is an important bug fix release since version 2.4.1.
+
+Bugs Fixed:
+ 1034, 1125, 1118, 1124, 1107, 1129, 1126
+
+Change Summary:
+26Jul08
+- When a migration job actually runs, re-check the Job record
+ and skip if the job is already migrated. This should
+ significantly reduce the problems with bug #1129.
+- Break the do_swapping into do_unload, do_swapping, and
+ do_load. It is much more logical that way.
+- Implement a set_dcr_from_vol subroutine in acquire.c for
+ reading volumes. This allows the dcr to be refreshed after being
+ zapped when the wrong volume is mounted. This should fix bug #1126 -
+- During multiple tape restore, bacula
+ does not ask for physical tape change, but rereads same tape
+- Apply patch submitted for bug #1107 with a small modification.
+ This fixes a bug where bcopy copied too many records.
+- Make some tweaks to bsmtp based on patch submitted in bug #1124.
+ This fixes bug #1124.
+- Make the default bat restore Pool be Any. This fixes bug #1118.
+- Fix bug #1125 which reports a SD crash at the end of a tape
+ during restore. It was a debug statement that should not be
+ in a release.
+- Fix #1034 by setting big timeout on mysql connections
+- Do partial integration of the Win32 bat build created by Eric.
+- Ensure that SD tried to mount a volume not in an autochanger
+ at least once before asking for operator intervention.
+- Fix a pthreads bug in the Win32 pthreads emulation code affects only
+ version 2.5.x and above.
+
+
+=================================================================
+
+Release 2.4.1:
+
+Bugs Fixed:
+ 1103, 1100, 1105, 1094, 1109, 1106, 1101, 1102, 1112
+ 1117, 1116, 1097.
+
+Change summary:
+- Clear in-use bit on vol when unused while swapping.
+- Remove sleep at end of ./bacula
+- Clear hash packet for hardlinked files correctly. This corrects
+ a long standing bug where hardlinked files selected individually
+ would occasionally not be restored.
+- Apply Eric's patch to ensure that autoprune does not return
+ a Volume not in the autochanger unless requested to do so.
+- Correct bat restore display of multiple drives. This fixes
+ bug #1117.
+- Do find_a_volume() each time an unload is done. This
+ fixes the failure of maxvol2-test.
+- Use Qmsg() in job.c watchdog callback.
+- Release main control rwlock if ABORTing.
+- Make watchdog connect timeout queue messages rather than sending
+ directly to avoid lock conflicts with the real thread.
+- Remove const char that causes problems with Python, which has
+ older 'incorrect' headers.
+- Add const char in dbd.c to avoid compiler warnings.
+- Fix mtx-changer to detect both versions of Ubuntu (Debian)
+ mt. This fixes bug #1116.
+- Fix failure of 2drive-concurrent-test. Always read label;
+ clear_unload() only after drive is defined in acquire_for_read().
+- Generally clean up the manual tape loading code. The main
+ conceptial change is that when a volume is marked to be unloaded,
+ its volume name is retained, and it is only marked as unloaded
+ when either the autoloader says it is unloaded or another tape
+ is read on that drive.
+- Fix a recycling problem with two autochangers reported in bug
+ #1106. This may also (unlikely) fix a second recycling bug as
+ reported in #1103.
+- Fix a SD lock volumes deadlock problem reported in bug #1100.
+- Fix format problem in bscan output reported in bug #1105.
+- Copy more data when restarting a job so that run
+ overrides are kept. This should fix bug #1094.
+- Backport updates to jobq.c where possible.
+- Fix bug where SD did not ask operator if the device could not
+ be opened. Reported by Eric.
+- Add dbuser to DIR conf file (replaces user).
+- Add --with-db-password to ./configure
+- Fix regress to handle db_password.
+- Fix a couple of pedantic compiler warnings.
+- Fix bug when starting a restore on a storage that doesn't
+ have the requested MediaType.
+- More work on word alignments -- eliminate all ints from
+ the configuration routines.
+- Fix bug reported by Scott Barninger where the bacula script
+ refers to scripts in the wrong directory. Needed to meet the
+ requirements of recent FHS location changes.
+- Fix word alignment problem on non-Intel 64 bit machines
+ such as Solaris. This should fix bug #1097 -- bus error in SD.
+- Add missing win32/patches/wxWidgets.patch to 2.4 repository.
+
+=====================
+
+Release 2.4.0
This Director and Storage daemon must be upgraded at the same time,
but they should be compatible with all 2.0.x File daemons, unless you
use some of the new features that affect the FD. In other words, you should
not have to upgrade all your File daemons when you upgrade. There is
-no database upgrade needed from version 2.0.x to 2.2.0.
+no database upgrade needed from version 2.0.x to 2.2.0 nor from version
+2.2.x to version 2.4.0
+
+This release consists mainly of a number of bug fixes since 2.2.x
+but most importantly, a rewrite of significant portions of the
+Storage daemon reservation and autochanger code. This rewritten
+code is simpler, more stable, and works much better, particularly
+with multiple drive autochangers than the prior code did. That
+said, there are still a number of issues to be resolved. Note,
+the new code has not been tested in manual volume mount situations.
+
+Bugs fixed:
+ 1095, 1086, 1091, 1070, 1041, 1083, 1078, 1066, 1068, 1038
+ 1062, 1065, 1046, 1047, 1031
+
+New features:
+- You can now build bat without the QWT libraries. Use
+ --without-qwt as a ./configure option.
+
+Release Version 2.4.0
+- Update version to 2.4.0 to reflect the magnitude of the SD changes.
+- Make sure to clear JS_WaitMedia when operator mounts a Volume.
+ This fixes bug #1095.
+- Add create_postgresql_database.sql to Win32 installer. Should
+ fix problem reported by Juilio Monteiro.
+- Get correct slot when auto unloading a device. This fixes bug
+ #1086.
+- Fix possible seg fault if SQL error.
+- Fix renaming a Volume, which used the wrong Volume name.
+- If operator has rewind tape, print warning, release tape and
+ try once more. If tape is positioned somewhere, something went
+ wrong, so mark the tape in error and try once more. Previously
+ this error was fatal, now it produces an error message.
+- Ensure correct volume name displayed during restore
+- Fix a few more Coverity reported problems.
+- Fix #1091 about bad output in estimate command.
+- Modify autochanger locking to attempt to avoid race
+ conditions.
+- Make sure device not busy before doing label command.
+- Display open() errors except when polling. Previously too
+ many were suppressed. This should fix bug #1070.
+- Fix Win32 reparse points. Bacula will not recurse into any
+ reparse point directory, including mount points, unless the
+ directory is explicitly mentioned at the top level (same as
+ with Unix). A file that is linked to another file will be
+ backed up -- much as Unix does for hardlinked files.
+ This fixes bug #1041.
+- Remove StorageId test when pruning and recycling (Eric's changes).
+- Fix buffer overruns detected by Coverity.
+- Implement --without-qwt
+- Fix layout of restore tree dialog
+- Add configure bat QWT libraries, so that bat can be built
+ with or without the QWT libaries.
+- Implement regression that explicitly tests swapping a Volume
+ from one drive to another.
+- Enhance disk-changer to detect most error conditions.
+- Fix SD code so that it properly swaps a Volume between drives.
+ This fixes bug #1083.
+- Prevent a Volume that is being swapped from being freed from
+ the volume list. This will most likely fix, at least partially,
+ bug #1083.
+- Fix strippath so that it does not get a buffer overrun and crash FD.
+ This fixes bug #1078.
+- Make inability to change owner/group when creating a dir only a
+ warning rather than an error.
+- Fix SQL query in migration code
+- Fix bat seg fault at termination.
+- Add Bacula generated version to bat about box.
+- Backport development stream SD reservation system changes.
+- When wrong volume is mounted during read, unload_autochanger.
+- Stop searching for Volumes in SD askdir if DIR returns the
+ same volume name twice in a row.
+- Rework class structures for VOLRES, DCR, and DEVICE to make
+ the method names a bit more logical, and for more logically
+ handling the responsibilities.
+- Remove redundant code in terminating the scheduler that just
+ causes a seg fault in many cases.
+- Improve algorithm for detecting pre-reserved volume and
+ swapping volumes.
+- Prevent volume from being released while being swapped.
+- Refactor parts of stored/mount.c
+- Add sanity checks for VolWriteTime and VolReadTime
+- Take care of bad clock changes while computing VolWriteTime
+ and VolReadTime. This should fix or limit #1066
+- Correct error string numbers in dird/catreq.c
+- Restructure reserving, acquiring, and mounting volumes. Calls to
+ autochanger are deferred for mount.c -- simplifes the code.
+- Do not prune any running job. It just fails the job.
+- Lock the volumes when changing dev->reserved_device and marking
+ the volume unused otherwise the device can get reserved by
+ another job before the volume is released, thus blocking it.
+- Correct some SD catalog request error messages.
+- Turn off code in read_record that causes a seg fault in the SD
+ when reading past an EOS_LABEL.
+- Turn off unloading the autochanger in reserve.c as it just
+ causes problems.
+- Lock volumes when unreserving a device.
+- Do not mark volume unused when recycling.
+- Fix bug #1068 fixes a SD crash when using Virtual autochanger.
+- Generate correct JobMedia records during spooling/despooling when
+ running concurrent jobs. Thanks to Tom Ivar Helbekkmo
+ <tih@hamartun.priv.no> for excellent analysis and testing.
+- *Massive* commit of all fixes and back ports to create 2.2.9
+ Mostly SD reservations enhancements.
+- Fix Win32 FD backup/restore memory leak due to improper termination
+ of BackupRead/Write. This fixes bug# 1038.
+- Fixes a StorageId problem with volumes after migration.
+- Implements autostart scripts for Debian
+- Fix mtx-changer.in for broken Debian mt program.
+- Apply doc fix from bug #1062.
+- Resolve crash and improper restore wx-console Win32 restore
+ GUI. Fixes bug #1065.
+- Update Win32 wxWidgets to latest version.
+- Attempt to do correct handling of Win32 mount points. Should
+ fix bug #1046.
+- Add const to AIX prototype for initgroups()
+- Fix to the JobMedia patch, which introduced a new problem.
+- Fix creating first JobMedia record during Migration to include
+ proper index. This caused slow restores of migrated jobs.
+- Fix bug #1047, do not strip paths on symbolic links.
+- Set catalog backup database and user name from values specified on
+ the ./configure line.
+- Apply patch to correct bug #1031, about wrong pool source information
+ in job report.
+
+
+=========================================================================
+
+Version 2.2.8 is a bug fix to version 2.2.7:
+- It fixes bugs: 1036, 1033(doc), 1028, 1040, 1045(doc), 1030,
+ 1042(partially), and possibly bugs 1018 and 1032.
+ Bugs 1018 and 1032 involve problems with multiple-drive autochangers,
+ and are difficult to reproduce.
+ See the ChangeLog for more details.
-Version 2.2.7 has several new features and several importatn bug fixes
+Version 2.2.7 has several new features and several important bug fixes
since version 2.2.6:
- It fixes bugs: 1009, 1022, 1024, 1019, 1012, 1021, 1020
1007, 1008. For more details, please see the technotes-2.1 file.
by skipping |. Fixes bug #936.
- Implement patch supplied by Landon to fix bug #944 where using
TLS with bconsole uses 99+% of the CPU.
-- Fix bug #946 about "bacula-dir -t" which doesn't works
+- Fix bug #946 about 'bacula-dir -t' which doesn't works
as expected.
-- Using "m" in bconsole will show messages as in prior versions
+- Using 'm' in bconsole will show messages as in prior versions
and not memory usage.
- Note, you need GTK >= 2.10 to be able to link the Tray Monitor
2.0.0 on, and probably versions prior to that. A detailed analysis
of the bug can be found at:
- http://www.bacula.org/downloads/bug-395.txt
-
+ http://www.bacula.org/downloads/bug-395.txt
+
- Add ./configure search in qwt-qt4 for qwt package
- Apply Martin Simmons patch that should turn off the new API usage
when batch insert is turned off allowing building on older
the main changes since 2.2.0 are:
- Fixed bugs #921, 930, and 923.
- More portable timezone offset in bsmtp thaks to Attila Fülöp.
-- Fix several bat issues.
+- Fix several bat issues.
- Use GTK+ calls instead of egg files in tray-monitor.
- Win32 service routines rewritten
- Updated projects file
New directives:
- Heartbeat Interval (bconsole.conf)
- Heartbeat Interval (bacula-dir.conf in Director, Client, Storage)
-- TLS Allowed CN (bacula-dir.conf in Client)
-- Regex Where (bacula-dir.conf in Job)
-- Strip Prefix (bacula-dir.conf in Job)
-- Add Prefix (bacula-dir.conf in Job)
-- Add Suffex (bacula-dir.conf in Job)
-- Recycle Pool (bacula-dir.conf in Pool)
-- FailJobOnError (bacula-dir.conf in RunScript)
+- TLS Allowed CN (bacula-dir.conf in Client)
+- Regex Where (bacula-dir.conf in Job)
+- Strip Prefix (bacula-dir.conf in Job)
+- Add Prefix (bacula-dir.conf in Job)
+- Add Suffex (bacula-dir.conf in Job)
+- Recycle Pool (bacula-dir.conf in Pool)
+- FailJobOnError (bacula-dir.conf in RunScript)
- CheckFileChanges (bacula-dir.conf in FileSet)
-- StripPath (bacula-dir.conf in FileSet)
+- StripPath (bacula-dir.conf in FileSet)
Other features or bug fixes:
- Fixed bugs: 916, 910, 917, 914, 906, 907,
- Different locking in reservations and despooling systems,
which means more micro-locking and less macro-locking, which
should give a lot more concurrency at the expense of slightly
- (<0.1%) more overhead due to more locking/unlocking, but
+ (<0.1%) more overhead due to more locking/unlocking, but
concurrent jobs should run much faster.
- Additional drive reservation algorithm that should solve a lot of
the problems experienced with multiple drive autochangers.
projects / bacula / bacula / blobdiff