2003-01-24 Version 1.29 released 22Jan03 - Added prefixlinks=yes/no Job record to specify applying the Where prefix to absolute soft links. Code is not yet passed to FD, because FD would then be incompatible with version 1.28. - Added skeleton of installation for Gentoo release for Patrick Naubert. - Add timer on open() for reading or writing a FIFO file. - Put btraceback and btraceback.gdb in sysbin dir (a bit of polution, but at least dumps will work). 20Jan03 - Added "append" all messages to a log to default bacula-dir.conf - Added WriteBootstrap to default bacula-dir.conf - Made smartall.c print "Out of memory" if malloc() fails. - Added pthread btimer routines. - Added timer to FIFO open statement - George was still having problems with VolUseDuration failing. On looking into it, one line of code subtracting 1900 was mysteriously missing from the source -- bizarre. 18Jan03 - Yesterday's version corrects all the problems I was previously having, and my production jobs are now completing properly. - Added a #define dev_cap() to test the capabilities bits. I just makes the code a bit shorter and a bit simpler. - Added phase 1 support for an output fifo device. The big difference here is that it is a STREAM device, which means that Bacula will only write to it and not read. Thus, Bacula assumes that the correct "Volume" is mounted and will construct a valid label (without needing the Volume to be prelabeled), and write to the device. - Added phase 1 support for input from fifo device (suppress re-read, add empty buffer flag). 17Jan03 - Improve printout of dbcheck with # files/path fixing. - Zap SD authorization code after use. - Added <> back to smtp (think about this some) - Doc 16Jan03 - Massive change to add jcr as the first argument to nearly every db_ call. This is because I was storing the jcr in the db structure, which will not work because everyone shares exactly the same structure. - More cleanups of error termination status in filed. - Found another bug in message.c where %s was missing in JmsgN. A lot of cleanup in message.c - Found places where filename listing was made (restore, verify vol) where the buffer could possibly overrun. - Chain include files on the end of the list so that the order will be correct. - Rewrote mtx-changer to output one slot per line terminated by a colon followed by an optional Volume Tag for the "list" command. Preparation for handling Volume Tags. - My production crash remains elusive. Adding debug code or running under the debugger eliminates the problem. I found a case in message.c where I was extending the message string by two characters to send it to the console. VERY BAD. Rewritten. - Started implementing fifo and program handlers (i.e. streams) in the Storage daemon. Lots to do. - Added a trace capability where trace statements are written to a file. Tmsg(). Hopefully this won't be used much. - Running lots of "production" saved by scheduling a few minutes after the current time -- slow process ... 15Jan03 - Removed <> on From and To in smtp.c as suggested by James MacLean. - Added code to suppress spurious error messages during cancel, but I was unable to eliminate all errors -- to be worked on later. - Up size of print buffer from 2000 to 5000. - My production run failed again. I'm beginnig to suspect hardware problems because running by hand or under the debugger always works -- we will see. - Added Update Volume VolFiles to reset correct tape files a bit dangerous. - Use the mysql_escape_string() rather than internal version. 14Jan03 - Cleanup handling of JobStatus by creating a subroutine. - Fix a number of minor things with JobStatus. - Print FD and SD JobStatus on backup report. - Add JCR to findlib -- so now FD stops normally when cancelled. - My production Director segment faulted during the second Job. I haven't been able to track it down. After rebuilding, all jobs finished correctly. - In investigating the duplicated Paths, I found that there were 10,552 of them -- only Path records. After checking the code, I do not see the reason, but I've enhanced the code to print the full path name. - I made a few improvements to tools/dbcheck. Mostly it provides a bit more feedback with verbose mode on when eliminating duplicate filenames or paths. 13Jan03 - During four simultaneous backups, the File daemon started detecting buffer corruptions. It turns out to be due to the fact that the smartall.c routines were not thread safe. They are now. - Based on input from James MacLean and team, I eliminated a number of places where printf could be recursively called by using "%s". - It turns out that because of an error of my understanding of mysql_escape_string(), the last argument was taken as the length of the string to be escaped rather than the maximum length. I've now corrected all my code -- Thanks James. 12Jan03 - George Motter reported problems with UseDurations, and it seems that there were a number of inconsistencies and problems with FirstWritten and LastWritten. Hopefully for the most part they are now corrected. Also fixed LabelDate if done through Console. - Try to chase down reasons why there would be buffer overruns. Added P & V around referencing last_fname for status. Rewrote find_one.c with MEMPOOL, but not yet tested enough to commit. 10Jan03 - Give extra margin to converting filenames from Unix to Windows in attribs.c of findlib -- dumb cygwin API doesn't provide for a length. - Added file:line traceback to size_of_pool_memory, check... and realloc in an attept to get closer to the memory overrun reported by James MacLean. 09Jan03 - Made yet another fix to quoted string -- paths! Thanks to Scott Medlock for reporting this. - Made | and < work. - Implemented FIFO reading/writing. To do so, simply explicitly mention the fifo file (named pipe) in the Include AND add the new option readfifo=yes 08Jan03 - Started implementing | and < on Include names. - Changed source to . in cats directory as requested by Andrew Kokarev. ============================================================================= 2003-01-05 Version 1.28b released - Corrected a typo of working_directory in bacula-dir.conf reported by James MacLean. - Fixed the fact that path and filenames in some cases were not being quoted before going into the database. Many thanks to James MacLean for reporting this. ============================================================================= 2003-01-05 Version 1.28a released - Corrected a missing quote in bacula-dir.conf reported by James MacLean ============================================================================= 2003-01-05 Version 1.28 released General: - Implemented Bare Metal Recovery for Linux and manual procedures for Solaris - Now using only a single technotes file kes-1.28 and will add to it as the development goes on. - Wrote a general purpose bi-directional pipe command. This replaces previous use of pipes as well as the run_program previously used. - Make BSRs stop if no more matches are possible. - Allow unliminted number of devices in Storage daemon. - Allow connections to Storage daemon before all devices are initialized. - Better documentation (and btape test command) on using fixed block tape drivers. Changes submitted this submission: 04Jan03 - Add cygreadline5.dll to Win32 release -- needed for console 03Jan03 - Add scripts make_catalog_backup and delete_catalog_backup that makes and deletes an ASCII copy of the catalog for backup. An example of how to use it is in the Type). - Added error messages for SQL errors in autopruning. 19Dec02 - Documented Bare Metal Recovery - Create new "rescue" directory containg the Bare Metal Recovery code. - Fiddle with SQL a bit for pruning as apparently the last InitCatalog was pruned after the expiration date eventhough it was the ONLY copy! I'm not sure this is fixed yet. 18Dec02 - Allow Director to pass a NUL where string to FD (fix in FD). - Fix installation mv of query.sql. - Make sure btraceback.gdb is not wiped on "make distclean" - Corrected a bug in mod of replace options pointed out by Dave Anderson. 16Dec02 - Started adding FileOptions ... - Fixed and incorrect print out of the number of files restored (Jarif). - Finally fixed EndBlock (and file address for Files) in catalog! - Added hostname to tape header as always planned. - Removed Level code (will not implement unless strong demand exists). - Tweaked bscan to print number of errors ignored before first SOS. - Enhanced btape "fill" to permit using one tape and to dump last block before writing and upon read back. - Make fsf_dev() return 0 on fail and 1 on success. - Use new db_get_job_volume_parameters() to enhance Write Bootstrap to contain more info (start/end file/block, file indexes). - Added --enable-static-fd, sd, and dir to configuration to enable making static versions of the daemons. 13Dec02 - The btape test program was indicating errors on Adrian's machine using the ATAPI (ide-scsi) tape drive. It turns out that this is a fixed block driver as a consequence, Bacula must be setup to write fixed blocks. btape was not always using the fixed blocks defined in the Bacula config, so that has been updated. It now works fine. A lot of tips added to the Bacula test command to help guide the user. - Documentation of the above significantly improved in the manual. 12Dec02 - Added code in watchdog to permit setting and clearing child timers. If the timer expires, the child process is killed. - Modified restore to handle differential jobs. - Added a new test to the btape "test" command 09Dec02 - More documentation of new features (week position in scheduler, bsr). - Re-read last block written on full tape to verify it. - Fix segmentation fault with btape fill command due to missing FileSet MD5. 07Dec02 - Created better SQL input editing routines str_to_int64 and str_to_uint64() - Pull Client from database in Console restore command. - Create a Unique list of Volumes to be mounted (previously had repeats). - Made many of the SQL searches better by using the ClientId rather than the name in the restore Console command. - Modified reading of a tape to include the VolFile info. Once the tape is past the specified file, reading stops. The BSR now includes both the VolFile and VolBlock. Currently VolBlock is not used. - Handle multiple volumes better by creating a real volume list with all parameters in it. - Display "At prompt waiting for input" in gnome console when at subcommand prompt. - Broke ascii to internal and internal to string editing routines out into new lib/edit.c file. 02Dec02 - Added a readme and an afs-bacula script to the examples directory that permits Bacula to backup an AFS filesystem. Thanks to Lucas Mingarro for the submission. - Added A Sun-desktop autoloader script and Device definition to the examples/devices subdirectory. Thanks to Lucas Mingarro for the submission. - If the WriteBootStrap fails, the job will now be marked in error. - Added a week position to the scheduler syntax that allows you to specify 1st, 2nd, 3rd, 4th, or 4th, or first, ... fifth as a week position specification in front of a day. So if you say 1st sun ... the scheduler will start only on the first sunday of the month. The day specification can also be a day range e.g. sun-fri. This code is untested. - Implemented bpipe.h and bpipe.c in src/lib, which defines a bi-directional pipe. This allows executing other programs and sending them information as well as getting info from them. - Replaced the previous pipe usage with bpipes in RunBeforeJob and RunAfterJob. - The mail program now uses bpipes rather than pipes, which means that any error output will appear in the job output (truly bi-directional). - Modified BSR to handle counts and to stop when no more matches are possible. This is untested. - Improved error messages in smtp. ============================================================================= 2002-12-12 Version 1.27d - This is a minor update that fixes a segmentation fault in btape as well as reduces non-important error messages in bscan. 2002-11-29 Version 1.27c - Yet another silly error duplicating a column name in the SQLite make tables. No code change, just a make file. 2002-11-29 Version 1.27b - Set DB version to 5 in DB make files. 2002-11-28 Version 1.27a - Use g++ instead of gcc for testfind in tools 2002-11-27 Version 1.27 (26Nov02) released 28 November 2002 General: from kes25Nov02 - Mostly Cygwin changes Changes submitted this submission: - Updated bdb_find.c to have new calling sequence for db_find_job_start_time - dird/catreq.c edit in Volume name in error messages to SD - attribs.c switch to using the cygwin API to convert from POSIX paths to Win32 paths. This now permits Win32 path specifications in Include statements (all combinations not yet tested). It also avoids the large number of errors seen by Lutz if a Win32 path is specified in the Dir. - Update testfind help and remove set_attribsEx debug code. - Build testfind as default. - Turn off debug message in stored/askdir.c General: from kes21Nov02 - Another change in the database. You MUST either re-initialize your database or use the appropriate ./alter_xxx_tables in the src/cats directory. Changes submitted this submission: - Additional documentation. - Added MaxVolFiles to the database (not yet implemented in code). - Increased the database version from 3 to 4. - Change VolMaxBytes to MaxVolBytes, which is much more descriptive. - Compressed unnecessary spaces out of a lot of SQL statements. - Changed many %d to %u where unsigned integers are used. - Added the Bacula version and build date to each backup output. This will help knowing what version of Bacula was used. - Implemented VolUseDuration MaxVolBytes, MaxVolFiles, and MaxVolJobs based on maximums set in the Volume (Media) record rather than in the resource. This means the values can be individually set on a Volume basis. - Allow commas separating Include options (this was a subtle bug). - Added maximum string length argments to a number of subroutines to prevent buffer overflows. Most notably was do_prompt(). - Replaced MANY occurrences of strcpy() with bstrncpy(), which guarantees both that the length is not exceeded and that the string is properly terminated. This has a risk of a certain destabilization -- as does the changes to the SQL noted above. - In cram-md5 routine use my_name if gethostname() returns an error. - Increase timeout from 2 minutes to 3 minutes in authorization code. - Check the full string including \n in authorization. - Throw away any response longer than MAXSTRING. - Added a number of additional error checks on subroutine return statuses. - Replaced as many lld's with edit_uint64 as I could find. 22Nov02: - Added MaxVolBytes to Pool record - had forgotten it. Updated DB version. 23Nov02 - Update manual to document new Pool/Volume attributes - Correct make_sqlite_tables (typo, plus missing value in Pool) - Fix bizarre behavior in gnome-console and console when auth fails. - Add Pool attribute query - Add a few more error messages in askdir.c - More strcpy() conversions to bstrncpy(). General: from kes18Nov02 - Did a number of cleanups of string copying to limit the length and prevent buffer overflows. Changes submitted this submission: - Added a mutex arount the gethostbyname() so that multiple simultaneous jobs get the correct address. - Added the MaxVolJobs to the Media alter tables script for MySQL (forgot it). - Changed arg to db_find_job_start_time to be POOLMEM for returning the string. - Add the new VolUseDuration and MaxVolJobs to all the db_ routines. - Use bstrncpy() in most places in the db_ routines to prevent a bad database from crashing Bacula (self protection). General: from kes13Nov02 - You MUST either re-initialize your databases or use the ./alter_mysql_tables or ./alter_sqlite_tables in the /src/cats directory to modify your database tables. - Major improvements to dbcheck including an interactive mode. Changes submitted this submission: - Implemented VolumeUseDuration and MaximumVolumeJobs which control when a tape can be marked Used. - New tape status "Used" means it was used and cannot be used any more. - Defined utime_t which is 64 bit epoch time in seconds. btime_t is 64 bit epoch time in microseconds. - Created alter_xxx_tables to add new columns (VolUseDuration, MaxVolJobs) to database. Incremented db version. - Changed a few subroutine names concerning dates to be more descriptive. - Fixed several places where the last filename was not stored in JCR in FD verify_vol.c and restore.c - Major update to dbcheck. Unfortunately I used subselects, which work perfectly fine in SQLite, but not at all in MySQL. I must now rework it for MySQL. What a pain! - Removed "Database found" from configure output. It was no longer used or valid. - Corrected doc --working-dir => --with-working-dir Thanks to Tuck for reporting this. - Added the database name to the error message for mismatched DB version. - Commented out GMP and CWEB from configure as they are not currently used. General: from kes09Nov02 - Converted the manual over to a new format written in wml. This gives navigation buttons on the top and bottom of each page as well as a standard page size. - Updated the main Web site, providing a menu bar to the left and much better organization and presentation of the information. - Added code to recognize a Volume written with a larger block size than specified. The code automatically adjusts. Changes submitted this submission: - Created a new scripts directory and moved most of the scripts previously in the main directory there. This cleans things up quite a bit. - Moved the randpass stuff into autoconf. - ensure that the generic make_bacula_tables and drop_bacula_tables are installed. - Added code to filed to allow it to be run from inetd. Just add the -i option. I have not tested it. General: from kes30Oct02 - Alex found a problem with GZIP compression -- fixed. - bacula stript fixes. - Segment fault in Director fixed. - Added openbsd to platforms Changes submitted this submission: - GZIP compression was broken for large files due to a variable that is changed during the compression. This was relatively easy to fix. However, there were a number of fixes required to make GZIP and SPARSE files work together. Found one more place in bextract where SPARSE_GZIP testing was missing. - During testing of the above, I ran into the restore problem of multiple FileSet records. Fixed by including the FileSetId and the MD5 on the selection string, then always using the FileSetId. It is a bit confusing for the user, but ... - Francis found a problem with bacula.in (fixed - thanks) - I found another problem with bacula.in, so now it REALLY should work on most systems. - Added the Replace options code for restores. Not tested. - Fixed an ugly stack overrun bug in reading the config file that has been in for a long time. I found this in testing on FreeBSD where the Director seg faulted. General: from kes29Oct02 - Major change to the bnet communications routines. Changes submitted this submission: - Modified the bacula script to use pid files, so this script should now work on more platforms. Updated devel_bacula as well. - Added the full GNU hostname on the configuration print output. - Added gettimeofday() to configure, and tweaked a few variables for OpenBSD. - Added a chapter to the manual on Porting Bacula to other platforms. - Documented in Tips how to use the WriteBootstrap record. - Modified bc_types.h to error if 64 bit types are not found. - Pass replace option for restore to the FD (no code to use it yet). - Modify the FD to pass back the GNU OS string as well as the DISTNAME and DISTVER. This needs to be put in the Client record in the catalog. - Major reworking of the bnet routines to eliminate the zero length which previously indicated a signal. This risks to create some subtle communications bugs. The changes now permit blank lines to be sent from the user to the Director. - Found and corrected a few more places where the Win32 attributes were not being recognized. - Permit spaces in the Where string (restore) by bashing/unbashing them. - Handle quotes correctly in the Console program (actually Dir ua code). Previously they were not handled in the middle of a string. - Corrected two error message (error reading file), which previously stated it was a network error. - Reworked the files in the binary Windows release. Removed unused programs and added a README with the copyright as suggested by John. Also added console.exe to the release. - Found and corrected an autoloader mount problem where the wrong tape was specified. Previously it looped, then gave up. Now it correctly detects the volume is not correct and zaps the Slot in the catalog. General: from kes23Oct02 - I have mainly worked on getting all the details of a Restore to work correctly (new tape format, support for Win32 attributes, ...) - Trademarked name Bacula. - Implement Bacula tape format 11 (1.0 Immortal). This format will be maintained forever. - Accept Any Volume is yes by default. This modifies Bacula's behavior when writing tapes. Changes submitted this submission: - Corrected a bug on FreeBSD where CFLAGS would get a "yes". I was unable to reproduce this, but a user confirmed the correction. It was a problem with detection of largefile support, which FreeBSD has by default. - Added a new "license" chapter in the manual. Re-licensed a number of library routines (bnet.c, hmac.c md5.c, cram-md5, ...) with LGPL so that they can be used in proprietary software to access Bacula if so desired. - Move Director's AutoChanger doc to correct location (in Storage resource). - Document why trademark (to protect compatibility). - Implement and turned on Bacula tape format 11 (also BB02). This format moves the VolSessionId and VolSessionTime from each record header into the Block header. This is MUCH more efficient when reading records as now whole blocks can be skipped. Also added JobStatus in End Of Session record, and added MD5 for FileSet, which is necessary to insure uniqueness. - Implement a new Bacula time format for btime_t. It is Epoch time in microseconds (i.e. base 1 Jan 1970 in microseconds). This replaces previous floating point times. - Added Win32 extended attributes. In doing so, I moved all attribute handling from src/lib into src/findlib. Added new streams for Win32 attributes and for GZIP Win32 attributes. - Modified "Accept Any Volume" so that it really permits any volume in the pool to be mounted. - Removed "Mount Anonymous Volumes" from Storage daemon config. - Implemented sparse files. You must add "sparse=yes" on the include line for it to be enabled. - Print "None" in backup summary rather than 0.0% if there is no compression enabled. - Improved error checking in daemon connection/authentication code to prevent garbage data from harming a Bacula daemon. - All daemon tools MUST have a config file. - Completely strip drive specification on Win32 if a Where prefix is specified. - Corrected DB info for writing to files. Now the File Address is stored in File-Block variables in the catalog. - All Storage daemon tools now use common code for acquiring/reading Volumes. - If a device is unmounted, report it even if the device is not open. This will help inform users who have BLOCKED Bacula by unmounting a drive. ============================================================================= 2002-10-12 Version 1.26 (10Oct02) General: from kes10Oct02 - Changed Job name conventions to avoid : which is an illegal character on Windows. Changes submitted this submission: - Added check for inet_pton to configure.in. If it does not exist (e.g. Windows) use inet_aton - Documentation on GZIP. - Another restore doc example. - Documented btape fill command. - Set default restore directory from /tmp to /tmp/bacula-restores - Add additional no find error messages to sql_get.c - Creating a bootstrap file on Windows failed because the Job name contained colons which are illegal on Windows. Replaced the colons with periods. This corrects the Windows restore problem. - Print number of files to be restored in "restore" command. - On Windows systems, if there is a prefix, completely eliminate any drive: at the beginning of a path. - Minor corrections to the tree routines to handle Windows files such as c:/ better. Previously it insisted on /c:/, which would then not be found in the catalog for a "dir" command. - Modified cd command in restore to try /c: if c: fails. - Add a new S_ISWIN32 bit to the st_mode word so that low level routines that create files do not print error messages. - More work must be done to restore all Windows files correctly. Currently the following bits are not handled: Archive Hidden System Also, all the dates are not properly restored. General: from kes09Oct02 - More documentation. - Implemented new fill command in btape that permits filling a tape and then reading it back to ensure that it works with Bacula. Changes submitted this submission: - Added ReadBytes to JCR, which contains Job bytes read by FD, JobBytes contains compressed output of FD. - Modified FD to pass back JobStatus, ReadBytes, JobBytes, ... for backup jobs. This is upward compatible. - Modified backup termination status report to contain the compression ratio 100 * (1 - compressed-bytes/uncompressed-bytes) This will always be zero if no software compression was done, or if you are using a version 1.25 or older FD. - Pickup Job termination status of FD in backup, so now the termination status represents the state of all three daemons. - Implemented new fill command in btape that permits filling a tape and then reading it back to ensure that it works with Bacula. General: From kes06Oct02 - Implemented first major cut of bscan -- program to scan a tape and recreate a Bacula catalog. - Fixed lseek() relative negative seek that prevented writing multiple jobs to a File volume. - Implemented BB02 tape format -- currently turned off for writing. Changes submitted this submission: - Updated README - Implemented new BB02 tape format, which moves the VolSessionId and VolSessionTime from the record headers into the Block header. It is currently turned off. - Implemented new btime time/date format on the tape. This is currently turned off. - Added JobStatus to EOS tape label. This is currently turned off. - Changed start_block, ... to StartBlock in JCR for uniformity. - Print a message telling which tables were dropped/made with ./drop_bacula_tables. - Return JobType and JobLevel in db_get_job_record(). - Implemented get_current_btime(), which returns btime_t. - Bump debug message max size to 2000 bytes. - Add btime_t serial/unserial routines. - Rework all tape label routines to integrate btime and other label format changes (currently turned off -- for writing). - Fix lseek() to always be positive -- block.c - Consolidate record, block, and label major #defines in block.h. - Major implementation of bscan. Some more minor tweaking will be necessary. E.g. add multiple records for multiple simultaneous Jobs. - Cleanup lseek() to always use off_t as second argument. - Add new "created" tape label (EOT_LABEL). This generates a callback to the read_records() routine when the final end of all tapes is reached allowing proper termination and updating of the media records. General: From kes25Sep02 - Added means to bind servers to specific address. - Documentation Changes submitted this submission: - Added DirAddress, FDAddress, and SDAddress records to the corresponding resources that allow the server to bind to a specific address rather than any address. This security improvement was suggested by a user -- thank you. - Eliminated deprecated "Address" record from all sample Storage resources. - Made quite a lot of improvements to the bscan program. Much more to do. Aside from details, it is able to recreate a database from which you can do a restore. - The s option is not accepted on all versions of ar, so replaced it by an explicit ranlib call. - Fixed a bug that caused the Director to crash if you rudely bring down the console program in the middle of an SQL command. - Fixed a bug (missing break) that caused scheduled Admin jobs to be listed as "Unknown type". - Removed old code from Storage daemon that used a separate FD port. General: From kes14Sep02 - Better key generation on non OpenSSL systems. - 64 bit file address support if available. - Implement autochanger for reading - Lots of cleanup of tape reading code. - Automatically create all Pool resources when Bacula starts. - Implement bscan.c - Implement autochanger use via the Console commands "add" and "label" - Begin implementation of a regression script - Write bootstrap after ever job Changes submitted this submission: - After noticing that the non OpenSSL random key generator was not good on Solaris, John wrote a very nice randpass generator. - Integrated John's makeSessionKey for generating the FD->SD authorization. - Statically link tools (doesn't work on Solaris). - Document how to debug Bacula (new chapter in the manual) - Remove unneeded printing of error message in sql_get.c - Free SD description config record (previous oversight) - Bash spaces when sending Director names, and unbash them when received. - Ensure no divide by zero in rate computation in backup.c - Implement WriteBootstrap in backup.c - Allow tape reading to request volume information for any volume. - Create all Pool resources at startup. - Show only Backup jobs in Restore listing of last 20 Jobs. - Handle 64 bit stat packets in restore. - Don't do shell expansion on store_dir in config if string starts with |. This is used for piping the program for the WriteBootstrap. - Ignore SIGHUP -- DéjàGnu was triggering this. - Implement autochanger for reading, many changes to acquire_device_for_read() - Implement callback for reading Volume. Implement in bextract, bls, and bscan. - Implement bscan - Add count to bsr. - Fix Volume in bsr. - Move autochanger code to new subroutine in mount.c ============================================================================= 2002-09-05 Version 1.25a (05Sep02) Released - Fix unitialized stack variable in bextract so it will always read the currently mounted tape. ============================================================================= 2002-09-04 Version 1.25 (01Sep02) Released General: - Added .cvsignore files in each directory to cut down on the CVS output when scanning directories and finding Makefiles and such which are not part of the CVS tree. Changes submitted this submission: - Cleaned up a lot of the error messages in Verify, including indenting. - Had to remove some of the "automatic" error message printing in the DB because they are in fact things that come up for Verify but not really errors. 2002-08-30 Version 1.25 (30Aug02) Beta From kes30Aug02 - Fixed a bug where only the first file was restored if it is hard linked. The other links were lost. - In some cases of restoring to alternate directories, softlinks were not properly restored. From kes28Aug02 General: - Bacula backups now run up to 12 times faster than version 1.24 - Bacula can run multiple simultaneous Jobs. - Had to turn off TRANSACTIONS in SQLite because it doesn't work with multiple simultaneous jobs (I'm working on this). - Added a better Job printout for Restore Jobs. - Added a save/restore Rate to backup and restore jobs. This does not yet subtract out operator wait time so it will be underestimated it the Job waits on the tape (mount/label). - Fixed a major bug caused by free()ing a buffer twice in the Restore code. This made the SD more or less useless after any restore Job. Arrggg! - DIR - SD protocol changed, both must be updated at the same time. - DIR - FD protocol changed in upward compatible way. Upgrade of FD not required, but recommended. Changes submitted this submission: - Added bell to "make" when errors are detected. - Changed default compile option from -g to -g -O2 - Additional documentation as usual -- much based on questions or feedback from users. Thanks. - Fixed example config files to use new syntax and to have a Restore Job. - Define ETIME to ETIMEOUT for Irix - Added JCR filed to DB structure permitting direct printing error messages from within the Database subroutines. Some error messages may be printed twice as a consequence. When I see them or someone reports them, I will remove the double. - Added JCR to the bsock structure. This allows direct printing of network errors from within the network code. - Made most "trivial" debug messages have level 100 or greater. This allows for easier debugging of new code using the range below 100. - Began replacing all Emsg() with Jmsg() using a NULL as the JCR in Jmsg() is the same as Emsg(). - Fixed several crashes in the Director because of malformed config files. - Added SpoolAttributes to Job resource. - Pass CatalogFiles to SD. If set to no, the attributes are dropped within the SD rather than being sent to the DIR. - Cleaned up a number of information/error messages in user interface. This includes eliminating the "pretty please" response in favor of "yes/no" to delete volumes and pools. - Cleaned up quite a number of uninitialized variables reported when going to -O2, most were harmless, but a couple could cause problems. - The cd command in restore was not working correctly (it didn't allow some legal cds). - Fixed a segmentation violation in the directory tree handling code in the restore command. - Handle a few error conditions in the restore command better. - Permit "Where" to be set to nul in modification of a run command. - Clarified the error message for Verify if an InitCatalog has not previously been done -- thanks Chuck. - Add MaximumConcurrentJobs to FD. - Added code to mem_pool to die if a buffer is released twice. - Lots of work done on SD for multiple simultaneous jobs. Split device.c into device.c, mount.c and acquire.c - Started writing new lock code for SD, but may back it out -- needs more thought. Current code works, but is too complicated. Maybe can simplify it. - Cleaned up the SD tools code quite a bit. Added bootstrap to all tools. - Ensure that tape session labels are not split across two blocks. This makes reading them back much easier. - Fixed another restore bug concerning tape labels on multi-volume saves. ============================================================================= 2002-08-14 Version 1.24 From kes12Aug02 - Made a new tools directory. - Moved smtp into the tools directory. - Created a dbcheck program in the tools directory that checks for certain database errors and if requested fixes them. - Put the database link flags on DB_LIBS so that it now only is on binaries that actually need it. - Document dbcheck and testfind programs. - Move testfind to tools directory. - Check for FileSet after getting Client name in Restore command. Modify all necessary SQL commands to accept FileSetId. - Add a db_get_fileset_record in cats directory. - Correct list last 20 files concatenation to be MySQL dependent. They aren't SQL standard. Use SQL standard || for other SQL programs. - Modify dir command in file selection to produce a long form listing of the file/directory name. - Add platforms/irix/*.in files to CVS - Make Console command scanner accept quoted value fields. - If after selecting Client to restore, there are multiple FileSets, ask user to select one. - After obtaining JobIds for restore, ensure that there is only one MediaType, and select a Storage resource to be used. From kes11Aug02 - Made restore actually work. - Made console and gconsole be configured (requested by Chuck) - Updated kernstodo - changed fs in Job record to fileset for clarity - The console run command was not properly picking up the command line arguments. Fixed! - Changed add/remove/rm to mark/unmark - Volume keyword VolumeName was wrong, changed to Volume (in restore bsr). - Pass jcr to parse_bsr, and if non-NULL, it will output error messages to Job stream rather than Emsg. - Modified lexical scanner to include caller context so that jcr can be stored in lex context allowing error messages to be properly routed. - Renamed the Job message chain to be jcr_msgs. This eliminated the previous confusion with the default value, now named jcr->messages. - Implemented multi-volume bsr records. - Implemented pm_strcat() and pm_strcpy() to cat/add to memory pool buffers in same way as strcat/strcpy, but expanding buffer. - Modified db_get_volume_names to handle arbitrarily long Volume list. From kes07Aug02 - If the Console program terminated during the printing of messages, the message file was left locked. This was corrected by using the new read-write lock code. - Implemented Pw() and Vw() for obtaining and releasing a write lock using the rwl_xxx routines. - Deleted some old pthread_mutex code from mysql.c - In making the Irix port, I previously corrected the base64 routine so they were not sensitive to the machine definition of char. This made them incompatible with previous versions invalidating MD5 signatures which are in base64 format. I have now corrected this, making the routines compatible with the previous version but also insensitive to the default definition of char. - Removed some unnecessary NPRT() usage in signal.c (pointed out by the Irix compiler). - Add back BNET_NONO for compatibility with older File daemons. - Implement runing a restore Job in the restore command. - Make installation of query.sql unconditional. To prevent you own file from being overwritten, give it a different name. - Made parse_command_args() in ua take command from ua->cmd, this simplifies it and allows usage in other places -- e.g. building arguments for created run command in restore. - Cleaned the naming of s_full_ctx to be s_jobids. - Write the bsr created in restore to "working-directory"/restore.bsr - Fix bugs in command line scanner in run command -- also some misplaced breaks! - Eliminate static cmd in ua_server.c - Eliminate segmentation fault when using the Run command in the Console -- forgot to init_msgs(). From kes06Aug02 - Did a good amount of documentation. - Added sql_cmds.c to CVS, forgot, pointed out by Chuck. - Added new platforms/irix directory to CVS - Turn off transaction debug code in sql.c - Reworked the scheduler so that it handles the same Job starting twice in the same hour. Previously it took one or the other. - Tightened up the "slop" in the timings that I programmed to ensure that the same job isn't run twice or that no job was missed. - Increase the sleep time. - Used job_type_to_str() in status so that Admin jobs are correctly reported. - Add internationalization message chars to a number of routines in util.c - Make the watchdog timer clear the mutex when it exits. - Some tweaks to the Restore command. It needs fixing when a Job spans two or more volumes. From kes05Aug02 - Initial cut at adding Irix File daemon support. - Lots of work on the restore command. - Take a stab at eliminating the "broken pipe message" when the console program prematurely terminates. - Reimplemented transactions for SQLite by keeping them all in the same thread. - quit in the Console program now waits for Bacula to quit. To immediately quit, you must enter .quit - Eliminated BNET_EOF -- this COULD have some negative consequences on interdaemon communications. - Define Job termination status for Admin jobs. - Write new bstrncpy() function that guarantees that strcpy is properly terminated. - Fixed command input in ua_input.c to accept quoted items as a single field. From kes01Aug02 - The changes that added indexes to the SQLite database were lost -- thanks to Chuck for pointing this out. Redone. - Fixed a problem quoting strings in SQLite. This showed up as database errors with files containing ' or " in the filename. - Implemente restore command -- still in progress. - Added indexes to all fields in SQLite that exist in MySQL. - Moved unused immortal files from src/lib to src/immortal. This will permit splitting depkgs. - Started work on adding embedded MySQL. Cannot get the tables to work for the moment. Please don't try to use this option. - Convert to using single quotes for enclosing all strings. - Implement correct quoting algorithm for SQLite strings. Is dequoting needed? - Started centralizing SQL statments in the Director in sql_cmds.c - Changed a number of M_ABORTs to M_ERROR_TERM so that program terminates but does not produce a dump. - Allow wild-card matches in bsr for Client name, and Job. - Change OflineOnUnmount=no as default. - Fixed improper printing of filenames containing a space in bls.c Thanks to Carlos for reporting this. =========================================================================== 2002-07-26 Release 1.23a - Fix segmentation fault is FD status. - Turn off TRANSACTIONs for SQLite. =========================================================================== 2002-07-23 Release 1.23 From kes22Jul02 - Updated Bacula to handle some minor differences in MySQL 4.0.2 (still works with older MySQLs). - Updated MySQL documentation to correct some minor errors. - Added kernstodo to CVS - Fixed create_mysql_database, which I recently accidently broke. - Disallow zero length volume names in Console program. - Create alter_mysql_tables and alter_sqlite_tables for conversion from 1.22 table format to 1.23 table format. From kes20Jul02 - Investigation of SQLite performace problems reported by Chuck. - Updated ChangeLog - Updated README - made root@localhost the default email address because many smtp servers reject root but accept root@localhost - Added an autochangers chapter to the manual. - Documented most of the new features in 1.23 - Modified the SQLite interface code to start a transaction at the beginning of a job and commit it at the end of the job. It also commits it after 10000 changes. Hopefully this will improve performance. - Set the default cache pages from 2000 to 10000 for SQLite (i.e. about 15Megs of memory) hoping to improve performance. - Terminate last filename referenced in FD prior to copy to avoid race problem with multiple threads and no locking. - Enhanced the status output to include the JobType - Print an error message if the email program terminates in error. - If your machine has an MTUNLOCK, do it before doing an OFFLINE to ensure that the door is unlocked. - Added code to stored/append.c to spool attributes. This is in a testing stage and must be explicitly enabled. - Added a no_attributes variable to stored/append.c that prevents the attributes from being sent to the Director. As yet, no way to turn it on. - Modified SQLite code so that after 10000 updates any transaction is committed, then restarted. This keeps it from using too much memory. - Set the default cache size to 10000 pages (previously 2000). - Fixed a segmentation fault on Sun due to no default value for the JobStatus. - Added the same indexes to SQLite that I exist in MySQL. This VASTLY reduces CPU usage for lots of inserts. - Added spooling to bnet_send(). From kes18Jul02 - The following two changes were prompted by questions/suggestions from A Morgan. - If you have "AlwaysOpen = no" in your SD Device resource, Bacula will free() the drive when it is finished with the Job. - If you have "Offline On Unmount = yes", in your SD Device resource, Bacula will offline (or eject) the tape prior to freeing it. - Added Maximum Open Wait to allow open() to wait if drive is busy. - Added RunBeforeJob and RunAfterJob records to Job records. This permits running an external program with the following editing codes: %% = % %c = Client's name %d = Director's name %i = JobId %e = Job Exit %j = Job %l = Job Level %n = Job name %t = Job type From kes17Jul02 - Added autochanger support to devices.c - Allow user to change Slot in the Volume record. - Implemented code in lib to run an external program (tape changer) - Implemented a changer script for mtx. - Currently the changer commands used are: loaded -- returns number of slot loaded or 0 load -- loads a specified slot unload -- unloads the device (returns casette to slot) - Other changer commands defined but not yet used: list -- returns list of slots containing a cassette slots -- returns total number of slots - Implemented ChangerCommand, specified in the SD Device resource, permits editing of: %% = % %a = archive device name %c = changer device name %f = Client's name %j = Job name %o = command %s = Slot base 0 %S = Slot base 1 %v = Volume name - Implemented MaximumChangerWait default 120 seconds. It is specified in the SD Device resource. From kes15Jul02 - Moved techlogs from main directory to be subdirectory of doc - Added code for strerror_r, and detection of gethostbyname_r(). - The protocol between the Director and the SD has changed. - Major rework of SD tape mounting to prepare for Changer commands. - Separated Update Media record and Create JobMedia record. These are done from the SD by calling the Director. Need separate Create JobMedia so that when a Job spans a volume, all other Jobs writing the same volume will also have JobMedia records created. - Added message to user indicating selection aborted if he enters . to a Console selection request. - Create a jobstatus_to_ascii() routine for use in status commands. This makes a single routine from three separate pieces of code. Updated the code to properly handle more (all) termination statuses. - Tried to fix the gnome-console to handle history a bit better. There are still some problems with focus not being properly set to the edit box after history replacement. - Removed the shutdown() from bnet_close() hoping to fix Console termination errors that are occassionally seen -- no luck. - Moved add_str() to lib/util and renamed it add_str_to_pool_mem() so that it can be used to edit Job termination codes and Changer command codes. - Reworked how the SD mounts tapes (in device.c) so that control passes through only a single routine. The logic is much simpler than previously, and now adding AutoChanger code is straight forward. - Made SD tape mounting much more fault tolerant -- more cases retry instead of terminating the Job. - Wrote code to edit_device_codes() for Changer commands. Not yet fully implemented. - Added a ChangerDevice directive to the Device resource. Still need to add ChangerCommand. From kes07Jul02 - This documents what I did while on vacation. - A fair amount of documentation. - Implemented Verify Volume to Catalog (needs renaming) - Modified the database to include new MarkId field as well as Counters. Database level is incremented from 1 to 2. - Added first cut of managing autochangers by implementing Slot in DB. - Fixed a good number of M_ERROR messages to be M_FATAL where appropriate. Also converted a number of Emsg() to Jmsg() permitting better Job error messages. - First cut at implementing a general packet passing mechanism for bsock to replace printf and sscanf for passing data. - Made make_xxx_tables and drop_xxx_tables and create_xxx_database for bdb as well as sqlite. With this each database has the same set of scripts (with the exception that MySQL has grant_privileges). - Modified Makefile to only install the database scripts for the database that you are using. - Updated the pdf script to include all the html files. - Did quite a lot of documentation. - Made Bacula compile correctly with DEBUG turned off. - Implemented Pmsg() macros that permit using the debug print routines even with DEBUG off. - Updated job_cancelled() macro to handle JS_FatalError. - Changed db_find_last_full_verify() to db_find_last_jobid(). This permits working with Verify Volume to Catalog. - Removed TRANSACTION code from incrementing NextId for SQLite. - Implemented quick and dirty (not fully functional code) so that console program does not die if ctl-Z is entered. - Cleaned up backup.c a bit adding a bail_out goto to ensure proper handling of errors with minimum code. - Added passing Slot back and forth to Storage daemon with Volume info. - Implemented Slot in Media record in DB. - Implemented first cut of Counter resource in dird_conf.c. A counter resource is part of the Pool resource. - Implemented necessary database record. However, the resource doesn't yet create a DB record. - Implemented onefs (One File System) and recurse options on Include resource. - Implemented autopruning of Verify and Restore Jobs. - Did a better job of calculating thread_concurrency() for Sun systems. This permits better multi-threading allowing all the threads to run at the same time. - Implemented a wait_for_storage_daemon_termination(), which is used by backup and verify volume. - Clarified a number of UA prompts, and add feedback where certain commands completed but printed nothing. (More work to be done here). - Added a pile of new code to implement Verify Volume to Catalog. It is sort of a hybrid of "restore" and "verify" in that it actually reads the tape, then compares it to the catalog. Need Verify Volume to Client. - Added a good number of checks for job_cancelled() this should permit quicker and more reliable cancelling of jobs. - Fixed one bug in restore where I used ofile instead of lname, which means that if the lname was too long, the memory was not properly reallocated. - Cleaned up find_one.c in findlib quite a bit. - Link smtp static so that it works even if the libraries are messed up. - If mail host is not found, retry with local host in smtp. - Added slot and stream to the bootstrap record processing. - Do offline on tape cartrige for unmount command. This can be turned off by the user in the config file. OffLineOnUnmount in Device resource. - Implemented AutoChanger = yes/no in Device resource. =========================================================================== 2002-07-07 Release 1.22 - All the basic Restore code to handle the boostrap file is now implemented with the exception of the Console Restore command. - Retained compatibility with previous Restore providing no bootstrap file is specified. However, the old code (JobId based) will go away in a future release. - Fixed a number of segmentation faults in the Console program due to printing of NULL pointers -- thanks to Chuck and his Sun for finding and reporting them -- mostly in ua_output.c. From kes25Jun02 - Null pointers now are printed as *None* if NPRT() is used. - At the request of Phil, all resources are now chained in in the order they appear in the config file. - Implement Bootstrap record in the Director's config file, which allows you to specify a predefined bootstrap file for Restore Jobs. - Allow Bootstrap on the Restore record. This is deprecated. - Implemented code to pass the Bootstrap file, if defined, to the File daemon, then from the File daemon to the Storage daemon. - Added ability to set Bootstrap file in the Console when running a Restore. - Allow scan_error() routine to be passed to lex_open_file(). This permits using lex in the Storage daemon without worrying that it will ABORT on a syntax error. More work is needed to direct the error messages correctly. - Improved error messages a bit when config syntax errors are found. - Tested and corrected some errors in match_bsr.c - Removed askdir.c from bls and bextract by defining dummy entry points. - bextract now has bootstrap pretty well integrated. - Changed bextract ABORTs into FATAL and exit(1) or ERROR_TERMINATE. More user friendly. - Eliminated at least one lld and replaced it by edit_uint64() - Eliminated a few more _PROTO()s. - Corrected a problem with the SQL for in query.sql for Listing Volumes to Restore All Files From kes22Jun02 - Additional documentation. - Reduce the time the jcr mutex is locked in lib/jcr.c - More null pointer printing fixes in lib/parse_conf.c (debug output only). - Added daemon filename to Kaboom message so I know what daemon died. - Fix big bad bug introduced in watchdog.c last update that caused a segmentation violation (forgot to check for NULL pointer). - Add a bit more info to bad returns from Stored when starting a job fails. From kes21Jun02 - Fixed scan_to_eol - Fixed backup scanning forgot one T_STRING -> T_UNQUOTED_STRING - changed all rwlock_t to brwlock_t to avoid conflicts with Solaris library. - Added NPRT() macro to allow printing of NULL pointers, which segment fault on Solaris. - Cast printing of pid_t to int for Solaris compiler. From kes20Jun02 - Added expect argment to lex_get_token(). - Added tree.c and tree.h to lib directory. From kes19Jun02 - Improvements to the bootstrap compiler - Implemented first cut of bootstrap code in the Storage daemon - Modified the authentication code to be more explicit during failures. - Added a few more debug message numbers with more arguments (Dmsg10, ...) - Added more and more precise error messages when authentication fails. - Implemented new "expect" codes in the lexical scanner that allows the caller to specify what he wants. - Added integer, positive integer, 64 bit integer, and integer ranges, and name scanning to lex. - Implemented new lex code in parse_conf.c, this reduces significantly the code. - Hopefully fixed a watchdog race bug that caused the watchdog to time out a line after 1 hour rather than 6 days. From kes15Jun02 - Implemented a Bootstrap file parser and matcher. - Implemented first cut of bootstrap code in bextract - Started clarifying different termination codes. - Added M_ERROR_TERM to immediately terminate the daemon -- no dump. Also changed a few termination codes as appropriate. - Fixed create_jobmedia_record() to include all the fields such as StartFile/EndFile, StartBlock/EndBlock. - Print user friendly message when query mode ends. - Fixed ls style print routines (bextract, bls, restore) to check buffer size, to print the link if any, and to used edit_uint64() instead of %lld. - Moved scan_to_eol() and s_err() to lex.c to avoid dragging in parse_conf.c for new bsr scanning code. - New files: bsr.h, parse_bsr.c and match_bsr.c. parse_bsr.c parses a Bootstrap file, and match_bsr.c matches a bsr against a tape. - use strcasecmp() in several places instead of lcase(). - Add first cut of bootstrap to bextract. - Added File Size limit -- writes EOF after limit exeeded. =========================================================================== 2002-06-12 Release 1.21 - Fixed several problems with filenames being truncated if they contain spaces. Thanks to the user that reported this. - N.B. To get the new queries, after the "make" and "make install" you must explicitly replace the existing query.sql by query.sql.new. query.sql is a user configuration file, so it is not overwritten. - Added GZIP compression of Included files. - Added additional fields to tape SOS record permitting better recovery with no database. - Be sure to remove any !terminate statements you may have in your config files. - MUST UPDATE ALL DAEMONS due to protocol change to handle new tape format. From kes11Jun02: - Updated examples directory with my most recent config files. - Modified the M_TERM meaning to mean that Bacula has terminated in error without producing a dump. Previously, it meant that Bacula terminate (with or without error). You should remove any !terminate from your messages resources. - Changed the order of the libraries in the link so that tcp wrappers link properly. Thanks Phil for reporting this. - A user reported that filenames containing spaces were truncated in bextract. This was indeed the case. They were also truncated in Restore as well as in Verify. They are, hopefully, all fixed now. - Made a few error conditions in Verify non-fatal. - Modified the Console "query" command to permit multiple SQL statements per command. - Implemented three new "query" options: List last 10 Full Backups for a Client List Volumes used by selected JobId List Volumes to Restore All Files To use them, simply type "query" to the Console program, and select the one you want. - Modified bextract to default to extracting all files (i.e. /). - Cleaned up the code in bextract a bit. From kes03Jun02: - Improve Verify command to include files examined. - Eliminate fcntl() locking of the console file and replaced it with a pthreads mutex. - Note Director - SD protocol is changed to handle new tape information (Fileset, JobType, JobLevel). - Create config.out that contains a summary of ./configure to see what you previously configured: "cat config.out" - Implemented GZIP compression. Added addition syntax to Include { } resource to permit setting of any GZIP compression level: compression=GZIP compression=GZIP1 ... compression=GZIP9 Level 1 is minimum compression and level 9 is maximum. Using just GZIP gives the default (level 6). - Enhanced the tape format to include the unique Job name, FileSetName, JobType, and JobLevel in the Start/End of Session records. The code detects that tape level difference and acts accordingly. You can add data in the new format to old tapes. - Fixed an incorrect display message in the prune command. - Began implementation of Verify Volume and Verify Data. - Cleaned up File daemon restore.c a bit ensuring that termination cleanup is done and all possible Emsgs are converted to Jmsgs. - Modify bls.c to use common setup routines. This reduced the big code duplication that creeped in when I added different listing options. Also, changing tapes is now handled uniformily in a subroutine. =========================================================================== 2002-05-27 Release 1.20 - Started documenting Catalog Pruning and Recycling. Added catmaintenance.html and recycling.html - Fixed the query command bug. - Implemented PID files to prevent two copies of Bacula from running at the same time. Be sure you --with-pid-dir points to a working directory. - Modified ASSERT() to cause a segmentation violation. This permits a core dump. - Worked on gnome-console. It paints better, it tells you if it is ready to accept a command or processing one, and it is almost entirely GTK+ (rather than GNOME, I'm removing GNOME a bit at a time). - Limit to 100 the number of tape labeling errors in a single Job. - Add some additional ASSERT()s to the database routines. I don't like ASSERTs but they do catch bugs. - Tested and debugged recycling of Volumes. Needed to change the Director-SD protocol slightly. - Removed the experimental SIGHUP in the Director as it is a potential source of errors. - Added Level, Pool, Messages, and Storage keyword/value scanning to the Schedule Run record. - Improved printing of Schedule resources with "show schedules" - Do not autoprune Jobs/Files if the Job errs. - Updated newvol() to pass back the updated Media record. - Fixed the "update pool" command, which was broken during a much earlier code reorganization. - Improved a few of the pruning messages. - Removed readline inclusions from gnome-console (hang over from console). - Make Jmsg use daemon message handler if no message handler is defined. This is always the case in the SD and FD. - Started removing __PROTO(). With C++ prototypes are manditory, so why keep around legacy code. - Added a good deal of message internationalization -- more to be done. - Move cached Path into mdb structure and use POOLMEM. - Fix verify bug introduced last update causing incorrect diffs by turning off autopruning. - Started adding code to distinguish Errors from Fatal Errors. - Errors are now counted in message.c for a Job. - More POOLMEM cleanups. - Implemented Automatic pruning. - Implemented Automatic recycling of Volumes. - Added PruneJobs/Files/Volumes=yes or no to Job resource. If set, it forces the specified pruning. - Modified rwlock to accept multiple nested write lock requests. - Modified the cats directory to use the new rwlock mechanism in place of P() and V(). - Defined db_lock(mdb) and db_unlock(mdb). They are in effect transaction locks. - Add CVS Id's to most files. - Made first cut at AutoPrune - Added new autoprune file that is called at the end of each Job. It then decides when and what to prune and calls the UA prune routines. - Modified Run (in Schedule resource) to permit overrides of Level, Pool, Storage, and Messages. (not tested). - Added more POOLMEM in place of char. - Implemented daemon pid files. This prevents multiple daemons from running simultaneously. You can run multiple daemons if they listen on different ports. - Documented the Schedule resource a bit better. - Began gently adding a POOLMEM type (perviously void). For the moment it is a #define to char, but long term it should be its own typedef. - Corrected a bit of spacing on the status bar of gnome-console - Added Admin and Archive Job types. - Implemented multiple Messages resources. Each daemon has its own message resource, which defaults if not defined. Jobs may now have different message handlers. - Cleaned up a bit of duplicate code in backup, verify, and restore by creating appropriate subroutines in job.c - Added Type, Client, FileSet, and Level resource record definitions to the Director. These can be used in place of the Backup = ..., Verify = ..., ... commands. This needs documenting. - Corrected an incorrect message in the prune code and added the number of Files pruned in the message. - Corrected several paint problems with gnome-console. - Added a bit of printout in signal handler during traceback. - Cleaned up most of sm_check() to be turned on/off by define in version.h =========================================================================== 2002-05-10 Release 1.19 - Allow the user to select a new period for pruning. - Lots of additions to the manual -- prune and purge commands documented. - Applied Phil's configure.in fix for --prefix, ... - Fixed bug found by Phil (patch supplied) in updating MD5 signatures (revert to 32 bit FileId, move "static" variables into JCR) (catreq.c fd-cmds.c). - Reverted to using INTEGER for FileId in make_sql_tables due to a bug in MySQL. - Change editing code to %d for FileId. - Remove sqlite in make distclean in cats directory. - Remove console.conf in console during make distclean. - Remove gnome-console.conf during make distclean. - Remove bacula-dir.conf during make distclean. - Set default level when using Console if none specified. Bug reported by Phil. - A simple . command from Console is ignored. - Change program named from filed to bacual-fd in winmain.cpp - Change default config file for Win32 in winmain.cpp - Free namebuf on early return from find_one.c. Bug reported by Phil. - Modify testfind.c to dump orphaned buffers. - Removed terabytes from parse_conf.c because of problems with older gcc compilers. - Turn off gnome options in gnome-console by constructing empty argv. - Fixes to Verify when only MD5 differs. - Insert 0 for MD5 as default rather than space. - Allow .messages "transparent" command while reading input in UserAgent server. - In dird/verify.c ensure that correct filename is printed if only the MD5 differs. Minor reindenting caused large diff. - Delete unused code from backup.c - In filed/verify.c ensure that same algorithm as backup.c is used to pass back MD5 signatures -- especially for directories and files that cannot be read. Change dummy filename from X to *MD5-id*. This dummy value should never be printed. - Make gnome-console poll Director every 5 seconds for output. This means that queued up messages are displayed at reasonable intervals. Delete some unused code hanging around from the tty console program. - Begin implementation of prune commands. - Add command line history to gnome-console. Not yet saved across sessions. - Fixed some broken URLs in the manual. - Added JobId type. - Wrote first cut of "prune files" and "prune jobs". - Added command line history to gnome-console 2500 lines max. - Widened store_time() and associated variables to 64 bits using new btime_t definition. - Removed GNOME about box and replaced it with a somewhat crude Gtk+ about box in the gnome-console. - Widened StartDay in the Job DB record to be 64 bits. - Changed edit_uint_ to edit_uint64_ everywhere. Much more descriptive of what is done. - Removed most llds and replaced them with %s and edit_uint64. This makes the code a bit easier to read for beginners. - Added a btime_t typedef which is 64 bits wide. - Added most of the code needed to do Purge and Prune of database. Not yet tested. Please do not use. - Additional config options for Console. - Started adding code for Automatic Recycling of Volumes. - Allow arbitrary length filenames in Verify code. - Fix incorrect filename in Verify code. - Significant enhancement to number scanning in config parser. - Requires initializing MySQL tables, or applying cats/alter.sql to modify Media table. - Modified Makefile to printer a much more visible message if the make ends in error. - Added the Purge value to the VolStatus in the Media table. This value is set if the Volume is purged. - Enhanced the db_delete_media_record() routine to delete all associated records in the database. - Modified Console to always write to the variable "output" rather than stdout. This will permit directing output to a file. - Enhanced the Console configuration file to have a Console resource. This will allow us to add an rcfile and history file. - Modified Find_next_vol in catreq.c to search for "Recycle" volumes if there are none marked Append. Also made Get_Vol_info return the VolStatus to the Storage daemon. - Allow upper/lower case match on job level names. - Added another warning message to the Console "delete media" command. - Corrected a number of FileId types from uint64_t to FileId_t in ua_retention.c. - In ua server close database before freeing JCR (because pointer is kept in ua and jcr structures). - parse all numbers as doubles. Do lots of checking for valid number formats. Allow numbers in scientific form (e.g. 1.5e+10). - Remove terabyte modifier as it doesn't work in some compilers. - Fix bug in handling some modifiers. - Added all necessary code for Recycle to Storage daemon. A Volume marked Recycle will now be overwritten. - Filled out the prune command a bit and did some testing. - Make console accept redirected input. - Altered the Table definitions to include Recycle, FileRetention, JobRetention, and AutoPrune. - Widened StartDay to 64 bits. - Use JobId_t in more places. - Added the new table fields to the database record definitions. - Changed Recycle from string to a binary quantity. - Added a Version table with a VersionId to detect. future changes in the database. This should prevent a Bacula from working with a database that is not in sync. - Modify Console to accept input from a file. This will permit the .read command and allow reading a .rc file. - Added new retention and recycle variables to the Director's configuration. - The UA subroutines or commands can now be called from core code because the output routines detect the absence of a UA socket and direct output to the Job. - Added a verbose flage to the ua packet to permit reduction of output while running a UA command (prune) from a Job. - Did a fair amount of work on the prune command. Prune Volume now works. - Purge Volume now works. - Made last changes for integrating prune and purge commands. - Add -ltermcap to CONS_LIBS when readline is configured. More work to be done to search for termcap. - Added cats/drop_sqlite_tables.in, which will delete the SQLite database. - Got CWEB working so that we can compile filesys.w and immortal.w - Modified depkgs to include cweb. - Note, CWEB is not yet used by the core Bacula code. - Made cats/alter.sql, which alters an old database to bring it up to the new format. This only works with MySQL since SQLite does not have the ALTER SQL command. - Changed the old StartDay field in the db to be JobTDate, which is the latest time/date in widened Unix time format that the Job ran. This value is used when doing pruning. - Added code in cats/sql.c to verify that the database internal version corresponds to the db version compiled in Bacula. It is set to 1 currently. - Lots of changes to cats to bring the SQL up to date with the new retention period changes. - Added Console command code to permit changing a Volume's retention period. - Removed old code that used date_encode() and replaced it with widened Unix time(). - Started modifying Message resource scanner so that we can have multiple message resources. Much more work to be done. - Moved scanning for time into new library routine string_to_btime(). =========================================================================== 2002-04-22 Release 1.18 - Applied Phil's configure.in fix for --prefix, ... - Fixed bug found by Phil (patch supplied) in updating MD5 signatures (revert to 32 bit FileId, move "static" variables into JCR) (catreq.c fd-cmds.c). - Reverted to using INTEGER for FileId in make_sql_tables due to a bug in MySQL. - Change editing code to %d for FileId. - Remove sqlite in make distclean in cats directory. - Remove console.conf in console during make distclean - Remove gnome-console.conf during make distclean - Remove bacula-dir.conf during make distclean - Set default level when using Console if none specified. Bug reported by Phil. - A simple . command from Console is ignored. - Change program named from filed to bacual-fd in winmain.cpp - Change default config file for Win32 in winmain.cpp - Free namebuf on early return from find_one.c. Bug reported by Phil. - Modify testfind.c to dump orphaned buffers. - Removed terabytes from parse_conf.c because of problems with older gcc compilers. - Turn off gnome options in gnome-console by constructing empty argv. =========================================================================== 2002-04-18 Release 1.17 Kern Sibbald - Ensure that platforms Makefiles are called for clean and distclean. - Hide a lot of Makefile messages with @ - Applied Phil's configure.in patch. Thanks. - Doc updates, many to clarify points brought up by Phil. - Applied Phil's make_mysql_tables.in patch. Thanks. - Made authenticate.c in dird test bnet_send status. - Make label a bit more friendly with extra info message. - Add level 200 to debug info in run command. - Remove old code from findlib/testfind.c so that it compiles. - Add two minute max wait to cram-md5 authentication. Hopefully, this will catch non-responding WinNT daemons. - Allow time and size modifiers to be upper or lower case. - Turn off old CYGWIN patch for pthread_cond_timedwait() bug that now causes File daemon hangs on WinNT. - Fix mount messages for restore. Previous updates to save broke the restore code. - By default write fixed block sizes rather than variable block sizes. My Sony DAT simply does not work (reread failures) with variable block sizes. - Data split across tapes was not being correctly restored (at least on DAT tapes). - Implement MaximumBlockSize and MinumBlockSize statements in Storage daemon so user can specify variable block sizes if desired. - Return error if block checksum error on read (previously ignored). - Fixed some error messages in bls, and in some cases exit(1) rather than ABORT. There remain ABORTs to be examined. - Made a local copy of tcpd.h and corrected the function prototype problem for C++. =========================================================================== 2002-04-14 First public release 1.16 Kern Sibbald - Many intervening changes/updates. 2001-09-29 Release 1.0 Kern Sibbald - Fixes to problems found at John's installation. - See techlogs/kes25Sep01 for details 2001-09-25 Release 1.0 Kern Sibbald - Helped John install at his site. He is backing up one Linux machine and one Win2000 machine. 2001-mm-dd Many releases - Released internally. 2000-04-09 Release 0.4 Kern Sibbald - Added a lot of Director configuration code. In part (a small part), the Director now uses the bacula.conf file to know what to do. There still remains a lot to do. - The file search code is now integrated into the MySQL database routines, and Bacula is now hooked into the MySQL database (if so configured). Thus all the database records are being created, though there still remains some work in getting all the details into the records (such as the Storage coordinates). - I've defined how Volume pools work (see manual). - I've defined error message handling though no code is yet written. - The following resources and records are implemented in the Director configuration file (bacula.conf): Director Name (passed to Storage daemon) Client Name, Address, Port Password (not yet used) Storage Name, Address, Port, Password, MediaType Job Name Type (not yet used) Backup (only one permitted) Client, FileSet Storage Schedule, Messages, Pool (not used) FileSet Name Include (options ignored) Exclude (no options) Schedule (not used) Name Messages (not used) Name Debug, syslog, mail, append (not used) Catalog Name Address, Port, Password (not used) Pool (not used) Name PoolType, MediaType (not used) - Added Storage configuration routines. Director Name, password (verified against those sent by Director) Storage Name, Address, Port (address, port must correspond to Director's values) Password, MediaType Device Name, MediaType (must correspond to those sent by Director) Archive Device (defines device name) 2000-03-10 Release 0.3 Kern Sibbald - Implemented new base64 encoding for attributes. This eliminates some of the error messages in the sprintfs on the Solaris due to different stat() sizes. - Implemented the first cut of the file search routines for the File daemon. The exclusion lists work including wild cards. There is a lot of work to be done, but the basic structure is now in place (wild cards do not yet work for the include). - Completed writing the memory pool code. Now I must implement it in the daemons. - Modified the bacula start/stop script so that it has a better chance of working on SGI and Solaris. - The catalog code has not been converted to the new file search code. 2000-03-06 Release 0.2 Kern Sibbald - Integrated John's fixes. - Made Makefiles self configuring when Makefile.in is changed. - Added a runit script in the top level that runs bacula - Added a "bacula" script in the top level directory that starts and stops the File daemon and the Storage daemon. 2000-03-03 Release 0.1 Kern Sibbald - Basic Director, File, and Storage daemons. Standalone Catalog services using MySQL 2000-01-22 Kern Sibbald * Setup basic file structure with ./configure