View | Annotate | Download (55.8 KB)
Ok, the oracle database backup never actually worked at all. We used the wrong delimiter, did not preserve correct sequence order, did not merge primary keys and uniques... i think no one ever tried to actually restore an oracle backup.
Fix all of this.
Fix Bug #48955 - Correctly extract column default value when exporting PostgreSQL tables.
Small change to r9765, Bug #47495
Fix bug #47495 - Move hardcoded language string to language file.
Fix Oracle database backup (Bug #46715)
Prevent incomplete backups stored if option "store and download" is selected and admin cancel download by removing the option. (Bug #20325)
Sort backups by date, newest first (Bug #14818)
Ok, story real database server info, as well as caching it Store it on installation too - allows us to check the db version used on installation and used currently to warn the user about incompatibilities
hmm
#24245
#22385
Hiding everywhere little buggers, eh?
#22385 - David told me to ci this
This commit was manufactured by cvs2svn to create branch 'phpBB-3_0_0'.
dumdidum... sorry. ;)
#13463 - tiny issue with get_tables()
#10785
#10715
I knew I forgot something :P
#10495
#9662
#9601
- PostgreSQL can now use schemas. During install, under database name, enter databasename.schema - phpversion() to PHP_VERSION
- Avoid creating backups that we can't restore :)
adding @package tag which is required for all classes.
forgot to actually commit it :P
- IE7 should now work too
- used wrong heuristic to measure if we were done, using a different one now.
- finally fixed this one, IE has a bug that makes puppies cry.
- some stupid bugs
- small mistake :-(
#8154
#8146 - came up with an alternate chunking algorithm, it does not require additional calls to seek (it reads exactly what it needs to and does not need to look at the same piece of data twice) and may or may not be more optimal than the previous chunking algorithm (may be faster, may use more memory). For this reason, it is only used when the stream is not seekable (bzip2 is the only format that is not seekable)
#8144
#7570 #7232 - pretty much rewrote backup/restore. whatever is not new code is refactored. Download now works properly with gzip on the fly, bzip2 is still very much not on the fly ( will to take suggestions on this ). Should never hit any memory issue. Loads of bug fixes, tested with every DB we support
#8084
- should fix some Firebird issues ( can't believe that nobody found this until now )
- fix some oddities (doubled spaces for example) - changed the way make_forum_select() is returning the forum list - now including skipped forums but being disabled. This should make identifying the correct forum much more easier. - Changed some permission namings based on suggestions by the community...
- fixing some bugs - removing utf8 characters from email files (has been discussed internally, you guys know why) - making sure some opendir calls are checked before calling readdir.
#6502
- remove some of my mess :P
- nobody found this bug :D
#6192 - i wanted to do this using streams but found myself unable to do so. if one could figure out how to do so, i would like to find out :D
#6164
#6130
#6118
#6010
- a bunch of fixes - added todos to the updater to make sure i do not forget. :)
#5466
#5122
#4758
- store sql_layer directly within the layer itself - new method sql_multi_insert to circumvent db-specific hacks (hopefully not introduced any parsing errors)
consistent acp layout regarding backlinks and messages.
- some bugfixes - using E_USER_WARNING if an error occurred within the ACP (sadly not able to use it as a default for trigger_error - it seems to be hardcoded in PHP)
some updates. Also adjusted the utf tools and normalizer more to our coding guidelines.
forgot to remove some debug code :P
If this does not fix it, I am not quite so sure it can be fixed as this output exactly models that of the official MySQL tools...
Backup files are now named backup_[timestamp].sql.*
OK...
This commit should increase the total number of BBCodes from 31 to 2040. Some things to watch out for:
Each database likes to deal with binary data in its own, special way. They are, quite frankly, too cool for school.
MySQL, MSSQL and Oracle all allow me to send in a default value for their binary column using a hex number. However, MSSQL forces me to send the specific data as a hex number and thus we must CAST it....
- Make sure that the table is empty before we start cramming things inside :D
- CAPTCHA: removed the extra slash
More backup stuff
- Made some things nicer for some of the DBs - Made postgreSQL work on non empty databases - Made SQLite ultra fast on restore
- Properly escaped (as far as I know) the profile data fields so that one may now use reserved words as column names
- Database: eh? meh. - CAPTCHA: A little easier to see which options map to which controls
- Dramatic speed-up in SQLite backup code - Wrote another patch around SQLite - SQLite has a lack of (among other things) solid definition of their tables. e.g. "foo" and foo are both valid col names... Database backup and profile creation are now both aware of such "features"...
bugs? bugs.
oops?
what? yeah... - turns out the backup issue was not a backup issue but a schema issue - let there be color
fixes...
- Some fixes
Another one bites the dust
assorted backup fixes
- Fixed a bug - PGSQL got updated, we now backup created domains
ok, this one is rather large... the most important change: re-introduce append_sid: old style continues to work, not a performance hog as it was in 2.0.x -> structure is different
apart from this, code cleanage, bug fixing, etc.
- eh? meh.
- Make it impossible to open/delete/download files that don't exist :P
eh? meh.
- Bug #1709
- Use specfic keys - Unborked mysqli :D - PGSQL now supports sequences! woo! - adsrc is not cool anymore, adbin is where it's all at...
- Tuples? Nah, - Install works w/ Oracle - Got the queries in the database acp to conform to the CS - profile acp now creates the proper alter column statement depending on the dbal
i can't seem to not commit this debug code :-(
- extrapolation is uncool, lets get the _real_name instead of guessing - fixed some bugs :P
- Olympus can now install every database, more work to come on getting the rest of the board to work
- d'oh!
- Firebird support for backing up - some bug fixes :P
- CS
- More bugfixes (Sorry!) - ODBC support
- Nobody wants my debug code :D
- An index can have multiple columns :P
- Olympus now properly works with MS SQL - DB backup system handles MS SQL - A few bug fixes to the backup system ;)
- Store and Download is now takes advantage of on-the-fly stuff correctly - Download and deletion of stored backups now allowed
ok, thought about this now for too long. I think the best solution circumventing memory consumption and not introducing "hacks" is to seperate module information (and probably more in 3.2 for installation/uninstallation) from the main code.
- We do On-the-fly file writting for everything in backups, should speed things up - Removed the ability to make archives for backups... Its one sql file, no need to use some heavy archive to store it... - Made it easier to use the restore feature - Flushed the data buffer at the end of every insert statement, this should make large tables (like posts and topics) faster to backup...
- a bunch of bugfixes. :P
- ok, we should try to write the sql data as early as possible (to prevent extensive memory consumption). What needs to be done now? Trying to compress in realtime if specified and it would be also nice if the filenames had a more meaningful naming. :)
- Free some results! - Nicer MySQLhandling (still needs work)
- Cleaned up some compress stuff - Replaced deprecated functions in the PostrgreSQL DBAL - Added an undefined constant during install - Oh, and we now have backups :D We currently work with all the MySQL flavors, PostgreSQL and SQLite...