* Adds note_severity and updates dbconfig. New SQL stuff too.
* whoops please don't hack into my database >:^(
* UI change, changed how it's stored in the DB, removed some queries when it returns, changed stuff to key.
* Update sql_message_system.dm
* this was not defined
* random indent
* wait how did this get here
* okay enough web edits I promise
* just kidding I got u
* Update common.css
* Added buttons, changed UI again, standardized the inputs, added severity for appearance bans, fed the dog
* forgot about the banning panel
* added an asset cache
* corrects asset datum var name
* converts to using key instead of ckey for user facing logs and ui
* more key_name for airlock wires
* futureproofing check for if key changes
* --onlyckeymatch script argument and fail/success counter
* fix
* Qdels all queries, adds sleep handling
* DB Core messages admins about undeleted queries
* Compile fixes. Adds missing set waitfor
* Remove world/New shennanigans. Add DBQuery/BlockingExecute()
* Less spammy notifications to admins about undeleted queries
* Increase dbcore fire time to 1 minute
* Upgrade undeleted query warning
* Better place of death
* Fix build
* Remove BlockingExecute, see BSQL PR for why
* Yep, missed that one.
* Psyche, that's the WRONG QUERY!!
By moving our logging to a DLL we see a drop in CPU/real time of 2-3 orders of magnitude. This is due to BYOND opening and closing file handles on every write, causing incredible amounts of unneeded overhead. The logging library also handles timestamps for us, further increasing performance gains.
This library will also allow for further offloading in the future, such as completely replacing file2text() and friends.
A pre-compiled DLL is bundled, but Linux users will have to compile manually. Directions can be found at the rust-g repo.
Log output is enhanced with millisecond time stamps:
[2018-04-01 15:56:23.522] blah blah blah
This includes runtimes as well, which benefit from the same timestamp improvements and no longer have hacky splitting code to add their own timestamps.
Log shutdown is handled in a dedicated proc called as late as possible, as rust-g integration expands this will be factored out into a generic native code shutdown proc.
* Cid randomizer check no longer triggers if the user's cid changes once then changes back
Normally, we only check for the randomizer if their cid changed from their last allowed connection. In some edge cases, somebody's cid can be different from a one time glitch, then change back on the next connection.
We now detect such cases and allow the connection.
* Update client_procs.dm
* Update client_procs.dm
Byond calls Stat() every 8 ticks with or without this code, it was literally doing nothing. This was pointless and none of this code is expensive enough to warrant throttling it anyways.
address is null in single user mode, but we didn't support that because address has been null because of a byond bug, but now that this is a config we don't have to care since its off in production.
* updates sqladmins with excluded flags and tempminning
* rank, editing and permission protections; json backup
* fixes
* reverted the wrong file
* fix slidecolor cascading and disabled switches not being sent in form
* removes debug uncommenting
* whoops this too
* commas fix + versions for changelog entry headers since 3.0
* actually account for can_edit_flags and use of @ with - or *
* fixes and rearrangement of admin > localhost > autoadmin precedence
* in case you want to not have autologin autoadmins
Disconnects from deleting the client are immediate, where are other commands are queued to some degree or another, browse is the worst one of all since it has to queue behind after connect resource sends as well as the asset cache or browse_rsc calls.
I'm finding hints that clients on lagger connections are getting disconnected before getting the reconnect command, and I was able to reproduce this myself on a cell phone connection while connecting to terry
