| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
| |
where used in our overwrites:
- Check allocated memory and ensure it is released in our code
- Consolidate code controlling both an initial query and any retries following connection failures
- Perform additional checking to ensure a connection error was the cause
- Correctly capture any result sets after the max_allowed_packet setting was changed
- Check and cache the error strings and numbers for the intended queries, not any helper queries, and return them when requested.
|
| |
|
|
|
|
|
|
| |
and for resetting it (a bit more transparent info)
- now also check the delegate if queryGaveError: method is implemented
• while adding/updating a row to the db Table Content's pane now indicates that process better via spinning wheel
• after TRUNCATE TABLE reload that table content
|
| |
|
|
|
|
|
| |
- The correct result set is returned even if extra internal queries are used
- The correct error message is written even if extra internal queries are used
- If the mysql thread is altered as a result of a query - resetting the connection - the encoding details are now restored on the connection
|
| |
|
|
|
|
|
| |
- Add view and code to support drawing a connection status icon at the left of the titlebar
- Use for SSH tunnels; completes Issue #114.
- Remove reference to an image resource "j" in the DBView - fixes console log errors.
|
| |
|
|
|
|
|
| |
up queries further (1.5x faster for short socket queries).
Doing further research, it turns out that mysql_query should effectively lock the connection, and mysql_ping should not be performed while a query is active as mysql is specifically not thread-safe in this one regard. This fixes #293. However, it means that we will suffer from http://bugs.mysql.com/bug.php?id=9678 again; if we don't complete the framework integration before 0.9.6, we should roll back this patch and r831 to allow all connections to behave properly and avoid hangs.
|
| |
|
|
|
|
| |
If a query gave the error 2006 (server has gone away) try to increase max_allowed_packet - if the user has SUPER privileges - and redo the query. If max_allowed_packet was changed reset it to default after executing. In addition if error 2006 occurred and max_allowed_packet can't be set write an additional error message to the log "(Please check if query size < max_allowed_packet)".
- 10MB of data are no problem
- while trying to insert 67MB MySQL gave the error : ... X'(null)' plus malloc: mmap(size=141287424) failed (error code=12); can't allocate region
|
| |
|
|
|
|
|
| |
- (int) getMaxAllowedPacket;
- (BOOL) isMaxAllowedPacketEditable;
- (int) setMaxAllowedPacketTo:(int)newSize;
|
| |
|
|
|
|
|
|
|
|
|
|
| |
interface responsiveness.
- Rework [CMMCPConnection queryString:usingEncoding:] to no longer check the connection before every query. This removes a ping from the equation and therefore effectively halves connection lag.
- Move the actual query to a thread (note this is *not* the long-desired multithreading!) to still correctly cope with http://bugs.mysql.com/bug.php?id=9678 . As soon as we upgrade the MySQL client libraries this can be removed, the code simplified, and a further processing speedup will be seen.
- When testing a ~10,000 row CSV import (=10,000 queries), the following speedups were observed:
- Local server, socket connection: 1.2x speedup
- Remove server, 30ms ping: 1.85x speedup
Slower connections will observe the biggest benefit from this commit.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
- Implementation of a new SPSSHTunnel class, designed to closely integrate SSH tunnels within Sequel Pro.
- Integration of SPSSHTunnel - new connection methods using callbacks, and CMMCPConnection integration
- Keychain class upgrade to include the new SPSSHTunnel keychain password helper on the trusted access list for new passwords
- Keychain passwords are now held in memory/UI for only as long as necessary, increasing password security
- Updated interface to enable/add SSH tunnel functionality
- Remove old SSHTunnel class
- Addition of new target for the SSH Tunnel password assistant, addition as a dependency of the main target, and addition to build script to copy into resources directory
- Fix a keychain password deletion crash
|
| | |
|
| |
|
|
| |
- added methods to CMMCPConnection to determine server major,minor,release version
|
| |
|
|
|
|
|
| |
- Update a few pref checks which were using the old pref values still
- Resort preference defaults to match layout in prefs, restore old default values for some keys
- Restore missing defaults for new editor properties
|
| |
|
|
| |
ensure the connection encoding is similarly reset. This addresses Issue #79.
|
| |
|
|
| |
queries are sent as UTF8, if possible; this correctly preserves encoding data when working with files export from Sequel Pro or via mysqldump, and attempts to fall back to the current connection encodings for other files. This should resolve Issue #116.
|
| |
|
|
|
|
|
| |
- Improve connection error messages slightly.
- Avoid reading prefs for every keepalive check - now sets connection timeout and keepalive interval on connection setup
- Fix and extend connection checks to avoid showing extra errors when the connection has been closed
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Use method caches for oft-called functions, and support caching of chunks of the underlying string for string walking, resulting in an overall 1.3x-1.4x parsing speedup.
- Improve handling of multi-character comment starts (eg / or -) at the very end of strings
- When running splitString... methods return even empty strings for consistency.
- Update TableDump and TableData to match new usage
SPStringAddition changes:
- Add a formatter for time intervals.
CMMCPConnection changes:
- Add support for timing queries
CustomQuery and nib changes:
- Change the "Run Queries" button to "Run All".
- Add a "Run Current" button, which runs the query the text caret is currently positioned inside; if text is actually selected, this changes to "Run Selection". This addresses Issue #43.
- Amend the "rows affected" string to better reflect the actual number of rows altered by several queries, show the query count if > 1, and display the overall execution time of the queries. This addresses Issue #142.
- No longer execute blank strings as part of the custom query, preventing errors.
|
| |
|
|
| |
retrieval, and support for disabling keep-alive by setting the interval to 0.
|
| |
|
|
|
|
|
|
| |
Encoding" for clarity
- Add support for EUC-KR (I think, unable to test), hopefully fixing Issue #149
- Add workaround for UTF8 via Latin 1, supporting viewing and editing of data added using broken clients/upgraded old Wordpress/etc.
|
| |
|
|
| |
databases queries. This fixes Issue #176.
|
| |
|
|
| |
every sixty seconds (a value stored in prefs, though not exposed via the prefs interface yet) in a background thread, to further improve connectivity handling following r334. Pings are only run when necessary - ie active use of the program extends the next ping interval as appropriate, so general impact should be minimal.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
appropriately - offering to retry, reconnect, or disconnect. This fixes Issue #93, Issue #69, and Issue #77.
The gory details:
Previously, MCPKit was correctly running mysql_ping to ensure a connection still existed before running a query, and aborted the query if the connection was no longer active.However the code very rarely checked the response of this, so if a query failed subsequent queries would continue to be run and the program would end up checking non-existent results, throwing Cocoa exceptions and generally breaking. However, mysql_ping would also use the default timeout (30 seconds) for each check - when running the (previous to r333) 14 queries to switch tables, this resulted in a long hang before the program even broke.
To exacerbate the issue, certain situations triggered a bug present in mysql_ping in the old client binaries we're using (http://bugs.mysql.com/bug.php?id=9678), causing mysql_ping to never return despite the presence of a timeout, and so causing an indefinite hang.
This issue has been fixed by:
- Setting a new 10 second connection timeout for both new connections (Issue #69) and for mysql_pings. Once preferences have been redesigned we'll probably make this value editable.
- Enforce the 10 second timeout even if mysql_ping hangs by using interrupts.
- Wrap mysql_ping in a new method to do the above and also catch re-established connections without reporting false failures.
- When a connection has failed, prompt the user to Retry, Reconnect, or Disconnect. Reconnect uses the original details for the old connection to establish a new connection, also attempting to preserve the current encoding.
- Do not return control to the main loop until a connection has been reestablished (or disconnected) - this ensures the program is never in a broken state without having to rewrite all query usage.
Much of the above patches the MCPKit connection methods as necessary.
|
| | |
|
|
|
structure.
|
