X-Git-Url: https://git.sur5r.net/?a=blobdiff_plain;f=bacula%2Ftechnotes-2.3;h=9044407aba633744e3495005027d4e0f41bf47c7;hb=103b05239147ee5d7358a2ab5e607dcb7134aede;hp=febe9b21accb5bbe6b3f8bb3634cb8f730a4a876;hpb=ef028adf16477e96aed1956d2144c0e45aa9c4a8;p=bacula%2Fbacula diff --git a/bacula/technotes-2.3 b/bacula/technotes-2.3 index febe9b21ac..9044407aba 100644 --- a/bacula/technotes-2.3 +++ b/bacula/technotes-2.3 @@ -1,7 +1,188 @@ Technical notes on version 2.3 General: -0Sep07 +30Nov07 +kes This patch fixes bug #1012 where the job is canceled because + of Max Run Time exceeded when the job has not yet started. +29Nov07 +ebl Apply Richard Mortimer patch that fixes #1016 when there are more than + one "part" to write. Instead of each part being stored on disk, + written to DVD, then deleted from disk, all parts are written to disk, + and only the last is written and deleted from disk. +kes This patch fixes the status command to include the formating string for + JS_AttrInsterting. It fixes bug #1021. +kes This patch should fix the Mac OS X build problem on the latest + Darwin, where sys/types.h was not included correctly (apparently + due to a change in the Mac headers). The solution was to explicitly + ensure that it is defined for the ./configure. + This patch also has a rebuild of configure that includes the Debian + qt4 kludge (look for alternate qt4 names). + This patch fixes bug #1020. +kes This patch should ensure that queued messages to the console are + displayed as soon as possible. It should fix bug 1007, but has been + reported not to work. +14Nov07 +kes This patch fixes the infinite loop when trying to increase the + maximum number of volumes in a Pool during the add command. + This patch fixes bug #1008. + +Release Version 2.2.6 +09Nov07 +db Remove extra data on estimate command in joblist for bat. +04Nov07 +kes Cleanup of code in SD that sets the LastWritten date. This doesn't + really change anything, but is the beginning of doing it *right*. + To do it completely right the LastWritten date must be passed to + the SD, but it is not yet done. +kes Minor tweaks to some bat restore SQL. +kes This patch fixes bug #1003 where putting the message output from + a Verify job into the catalog results in a recursive loop. The problem + seems to show up only with postgresql (to be verified). +kes Correct %x to %p in postgresql debug statement. +02Nov07 +kes Fix bug #942 where lots of emails where generated when the heartbeat + interval was low and Bacula wanted a different tape from the one in + the drive. +25Oct07 +ebl Add new SD despooling attributes and Dir inserting attributes + job status codes in the catalog (Status table). +24Oct07 +ebl Use qmake-qt4 instead of qmake when available (debian system). +22Oct07 +kes Add a security warning to src/cats/make_catalog_backup.in indicating + that passing the password via the command line (arg 3) is insecure. + This responds to bug #990 (it doesn't fix it). +19Oct07 +kes Set default debug_level to zero. This corrects bugs #991 and #993. +kes Fix a crash in bat when it cannot connect to the Director for example + if the Director is not running. +kes Remove redundant INDEX (JobId) from File table in the trunk. + This fixes bug #992. +kes Make explicit exception to GPL in LICENSE to permit linking + the Win32 FD with Microsoft VSS code. +kes Apply the Solaris package patch from + Masopust, Christian +kes Apply the patch to optionally add a timestamp in the debug output + submitted by Mariusz Czulada . +18Oct07 +ebl Use PQerrorMessage instead of PQresultErrorMessage in postgresql + backend (sql_strerror). Fixes bug #989 +16Oct07 +kes Apply Michael Short's Win32 regress patch. + It applied with no problems. +kes Add PATH command to SunOS part of mtx-changer.in so that sed and + awk are on the path. +kes Redirect stderror to stdin on mt status command in mtx-changer.in + since the status is returned on stderr on Solaris. +kes Activate Close button on tray-monitor window. This fixes + bug #986. +kes Check for NULL item in parse_conf.c before calling it. In the + case of Device, there is no item and this caused a seg fault when + Device appears in the conf file. +kes Add a #define __CONFIG_H when including config.h in bacula.h. This + avoids pulling in config.h multiple times in some new code. +12Oct07 +kes Fix spurious warning message printed when creating a volume with + a LabelFormat. This fixes bug #976. + +Release Version 2.2.5 +09Oct07 +kes Add JobId in place of Job name in all the Jmsg() output. +kes Correct the APP_DESC for the Win32 storage service (remove an e). +07Oct07 +kes Make maxvol-test and new more strict maxvol2-test work. +kes Print JobId in all Jmsg() job output. +06Oct07 +kes Add new SD despooling attributes and Dir inserting attributes + job status codes. This fixes bug 961. +kes Fix Win32 drive display in bat version browser. Fixes bug #962. +05Oct07 +kes Fix doc issue. This fixes bug #963. +kes Fix listing performance problems in bat. Pointed out by + Chris Howells. +kes Remove old debug code. +kes Fix bat code that tests for Win32. This should fix bug #968 +kes Query 5 list wrong Vol after migration. This fixes bug #960 +04Oct07 +ebl Fix #969 where user can't change Replace option in restore menu. +5ASep07 +kes Save jcr in thread specific data (tsd) for each thread. +kes Make Dmsg() print JobId as -%u. +kes Make Jmsg, Emsg, and others automatically pickup the jobid + for the thread if it exists. +kes Remove old reservation debug jid code. +kes Remove get_jcr_from_tid() from OpenSSL and tls code. +ebl Add an option that permit to specify spool size in job + definition. (Applying patches/testing/spoolsize_per_job.patch). + You must upgrade SD and DIR at the same time. +dvl Add new configuration item --archivedir +dvl Adjust regression tests to use tmp not /tmp. This will allow + multiple concurrent runs of the regression tests. +28Sep07 +kes Fix race condition that drops final block written to volume. + This happens in rare cases with multiple simultaneous jobs + when the Volume fills. This fixes bug #964. May be related + to bug #935 and possibly #903. +kes Eliminate more strerror() and replace with bstrerror(). +kes Remove BSD getopt and replace with unrestricted IBM version. +kes Fix Win32 build for changed calling sequences. +kes More tweaks to the mount volume routines to get everything + right. Changed variable name from find to have_vol, which is + much easier to understand. +kes Enhance btimer debug code. +27Sep07 +kes Fix FD->SD authorization failure, which was due to spurious + wakeups from a pthread_cond_timedwait(). Simply check the + predicate before continuing. This fixes bug #953. +kes Add const char where needed (more to do). Remove temp debug code. +26Sep07 +kes Setup SQLite busy handler before doing queries. Fixes bug #967. +kes Don't overwrite error message in check_tables_version. + Fixes bug #966. +kes Unlock batch insert tables with same calls used elsewhere. +kes Allow starting 59 jobs a second. +kes Make SD session key more random. +kes Add retry for SQLite opening db if it fails. Probably helps + for bug #967. +kes Suppress extra error messages during batch insert failure. + Fixes bug #966. +kes Add jcr to timer packets so if killed message can be sent to job. +kes Add JobId to all Dmsg() output. +kes Put some FD auth code on dbglvl rather than fixed. +kes Return insert attributes error message in db msg buffer to avoid + false error messages. Helps fix bug #966. +kes Separate batch init error messages. Probably helps fix bug #966. +kes Make SD code that contructs attribute insert check string lengths + more carefully. +kes If Dir gets an error during inserting attributes, cancel SD. + This reduces unnecessary error messages. +24Sep07 +kes Correct search boolean for getting Volume info +ebl Cleanup batch insert code. Probably fixes bug #965. +kes Back out one small change to the reservation system (reserving a volume). +kes Rework how a Volume is mounted. It is now much more intelligent and + will always attempt to use any mounted volume if possible and reduces + calls to the Director asking about volumes. +23Sep07 +kes Turn off some code when batch insert not enabled. +kes Edit FD name in connect error messages. +kes Rework the reservation system to take into account that the Director + might give us a Volume that is different from the current one being + used, and to ensure that we don't exceed Maximum Volume Jobs. + This fixes (mostly) bug #947 ' Maximum Volume Jobs = 1 produces + fatal error with multiple jobs running' +kes Add more debug code in reservation system. +kes Implement maxvol-test to check bug #947. +22Sep07 +kes Add code to handle tray monitor separated from Win32 FD. +kes Fix display of Win32 tray monitor after reboot. Fixes bug #952. +20Sep07 +kes Add missed return in fixing verify bug -- trunk only. +ebl Fix a command parser issue causing a director segfault. +kes Fix ANSI tape labeling. Fix restoring ANSI labeled Volumes. + This fixes bug #954. +kes Increase the max block size to 4MB fixes bug #957. +kes ERABT if user sets min block size > max block size. Fixes bug #956. kes Apply 2.2.4-poll-mount fix, that resolves bug #908 where a tape is not properly mounted (recognized) during a poll. kes Apply 2.2.4-verify patch that resolves bug #958. A Verify catalog @@ -20,6 +201,8 @@ kes Implement restore callback that Dirk and I agreed on for bat restore. This fixes bug #928. Restore of backup of zero files fails. kes Fix SQL Catalog logging which broke with new db_escape_string calling sequence. + +Release Version 2.2.4 14Sep07 kes Fix bacula_cats.def for new db_escape_string(). kes Increase size of name string when FD making connection to SD. @@ -54,10 +237,10 @@ kes Note, you need GTK >= 2.10 to be able to link the Tray Monitor program. kes Move patches into patches directory. 11Sep07 -ebl Fix bug #946 about "bacula-dir -t" which doesn't works +ebl Fix bug #946 about 'bacula-dir -t' which doesn't works as expected. 09Sep07 -ebl Using "m" in bconsole will show messages like before, +ebl Using 'm' in bconsole will show messages like before, and not memory usage. kes Fix bug #935, and probably also bug #903 where files were not restored. MediaId was not properly set in JobMedia record after