1 Technical notes on version 2.1
5 kes Implement a BSOCK send() method that writes the whole record
6 in a single write() request rather than in two.
7 ebl add update volume=xxx recyclepool=yyyy
8 volume takes recyclepool from pool when using update volume=xxx from pool
11 kes Fix encryption deblocking bug, which caused some restored files
12 to be truncated. This fixes bug #763.
13 kes Add FD event sequence order prepared by Eric -- for RunScripts.
14 kes Fix 12am/pm bug as reported in bug #782.
16 kes Apply Eric's ClientRunScriptAfter patch to 2.0 and 2.1.
17 ebl Use btime_t instead of uint64_t in media patch.
19 kes Optimize the use of the database a bit in the Status dir command.
20 Only open it when needed, ensure that if any previous database
21 was opened, it is closed.
23 kes Correct VolCatBytes bug from media patch.
24 kes Apply patch from bug #612 by Rudolf Cejka to speedup pruning
25 in the case of orphaned records.
26 kes Modify dbcheck to handle orphaned JobMedia, Path, Filename,
27 and File records in 300K chunks to be more efficient. This
28 idea came from Juan Luis Frances (if I remember right).
29 kes Apply Eric's scratch patch that moves a purged Volume to
30 the RecyclePool. Question: how is RecyclePool set? what
31 happens to the ScratchPool?
32 kes Apply Eric's media patch that collects read/write media
33 times as well as byte counts. This patch requires a
34 simultaneous upgrade of the DIR and SD. Note, there
35 should be some way to turn off timing. I'm not sure
36 times are in Bacula units.
37 kes Apply Eric's batch-insert patch.
39 kes Update projects list.
41 kes Implement Include/Exclude file lists in FD as dlists using
42 the new dlistString class. This permits efficiently having
43 large include/exclude lists and completes project Item 25.
44 kes Fix dird/ua_cmds.c so that a cancel command checks if the
45 console is authorized to cancel the job. This fixes bug
47 kes Modify SD so that the VolCatJobs medium record is updated
48 at the beginning of a Job rather than the end. This
49 fixes bug #775 where exceeding MaxVolJobs caused jobs
51 kes Added a mutex around getting and setting Volume information
52 so that multiple simultaneous jobs will single thread.
54 Switch to using Subversion
55 kes Remove src/pygtk-console/ from configure
57 kes Delete src/lib/btree.c from win32 build, then add rblist.c
59 kes Apply patch supplied that corrects debug print
60 in canceling jobs for max run time. Supplied as
61 part of bug #621, which was previously fixed.
63 kes Test on job_canceled() and sd_msg_thread_done inside
64 loop starting the message thread to avoid a race condition.
66 kes Remove rl_catch_signal from console.c as it conflicted
67 with the header definition. Fixes bug #765.
68 ebl Fixes bug #766 about RunsOnClient = Yes
70 kes Clarify some stored/acquire messages to indicate if the
71 problem is with read or append.
73 kes Fix memory leak with storage ids in cats/sql_get.c
74 kes Terminate watchdog earlier to avoid reference to released
75 memory -- reported by Jason Austin.
76 kes Move closing the database from jobq.c to the director daemon
77 termination routine. This fixes memory leaks for shadow jobs
78 (i.e. migration jobs).
79 kes Free up the unique jobid chain items in migrate.c. This fixes
80 a memory leak problem.
81 kes Convert some ugly looking for statements to use foreach_alist
82 in findlib/find.c. This will facilitate converting the structures
83 to use dlist (for large include/exclude lists).
84 kes Fix a bug in the btree.c and btree.h routines, then rename them
85 rblist and add them to be built in src/lib. Include some new
86 methods written by Rudolf Cejka that make the code more readable
87 (hides some of the ugly casting).
88 kes Add set_next() and set_prev() methods which make the code much more
89 readable. Also add a new dlistString class that facilitates storing
90 strings in dlists. To be used in the large include/exclude lists.
91 kes Make some trivial modifications to lib/tree.h that use rblist
92 rather than dlist for storing the tree links. This was suggested
93 by Rudolf Cejka. The result of this is that the restore tree now
94 uses red-black binary trees rather than simple linked lists. This
95 should give rather dramatic speed improvements for directories
96 contining large numbers of directories/files (more than 10000).
98 kes Fix maxruntime bug #621.
100 ebl Implement the include JobID in spool file name project.
101 kes Reorder projects file in order determined by Jan 2007 vote.
102 kes Implement item #12 on project list -- quick release of FD by
104 kes Fix open of SQLite3 db where user does not have write permission
105 so that DIR does not crash. Fixes bug #761.
107 kes Add back missing @ in Win32 Makefile.
108 kes Change 'Device not configured to autolabel' from INFO
110 kes Fix jobq.c to restart failed jobs when Reschedule Times = 0
111 which should restart an indefinite # of times.
112 kes Fix configure --help to print --with-mysql[=DIR]. Same for
113 other DIR specifications.
115 rn Fix msvc build problems. Fix bacula.dll exports for msvc build.
116 Fix path quoting in SQLite scripts.
117 Fix problems with SHGetFolderPath.
119 kes Create patches/2.0.1-restart.patch. Fixes bug #755.
121 kes Fix Job restart on error bug that promotes an Inc to a Full
122 backup. This should fix bug #755.
123 kes Add qt-console and first cut of code.
127 kes Fix Bacula->Documentation link on Win32 to point to index.html
128 instead of bacula.html. Fixed bug #750.
129 kes Return JobId in db_get_job_record() when JobId==0. This should
131 kes Do not release source pointers when restarting a failed job.
133 kes Add dynamic dll entry point for SHGetFolderPath to Win32 code.
134 This *should* fix bug #747.
135 kes Modify winbacula.nsi to substitute with g bin_dir_cmd. Should fix
138 kes Modify USTORE constructor to set an empty store_source string,
139 and don't copy the store_source string in a cancel. Hopefully
140 this will fix Arno's seg fault, bug #744.
141 kes Add back code to disable conio in configure. Fixes bug #743.
142 kes Correct the Options scanner in the FD to correctly handle
143 SHA1 option, which was eating the next option. Reported by
145 kes Add code to indicate when the SD is spooling, spool wait, and
146 despooling as requested by Alan Brown.
148 Version 2.0.0 released: 4 January 2007