summaryrefslogtreecommitdiff
path: root/memcached.c
Commit message (Collapse)AuthorAgeFilesLines
...
* Fix enum reference on the wire.Dustin Sallings2009-02-021-1/+2
| | | | | | | | By defining 'eno' as an int, the compiler couldn't validate that a valid value was going to be sent into it. In the current code, an invalid value couldn't make it in in practice, but if we're going to use enums, we should use them consistently.
* Remove stats malloc and stats mapsTrond Norbye2009-01-291-49/+0
|
* Return all errors for binary commands except not_found for getk.Dustin Sallings2009-01-271-47/+43
| | | | | | | | | | | | | | | | | | The theory is that we want quiet commands to be quiet in *normal* cases. A set that succeeds does not need to return anything. A quiet get has two normal cases: 1) Got a value (must return it!) 2) not_found (quiet here) All other errors are potentially interesting, so they will also be allowed through. Client authors can use this information to create bath commands of various types and not lose any useful status. A multi-set can be as efficient as a multi-get and still know which sets worked and which failed (even in the case of CAS).
* Do hash expansion in it's own threadTrond Norbye2009-01-271-1/+8
| | | | | | | Previously we tried to migrate one bucket over to the new hash table before we started a new command for a client, and we tried to lock the cache in order to determine if we should move an item or not. This resulted in extra contention on an already hot mutex...
* Fix whitespace buildbreak..Trond Norbye2009-01-221-2/+2
|
* The binary protocol does not implement noreply for the storage commandsTrond Norbye2009-01-221-33/+80
| | | | See http://code.google.com/p/memcached/issues/detail?id=11
* Don't leave stale data on failed set attempt.Dustin Sallings2009-01-211-0/+11
| | | | | | | | | | On a set that fails to allocate memory, remove any existing value for the key. Presumably, if a user was attempting to overwrite a cache entry for whateer reason, that should at the very least be considered an invalidation. This catches the binary protocl up to the text wrt 5f54ede879939a160168fabbc8cfc462f28e0004.
* Build fixes for FreeBSD 7.Dustin Sallings2009-01-201-1/+1
| | | | FreeBSD doesn't like %lu used for rel_time_t.
* Fix cast warning for sprintf(3), which is treated as an error thus causing ↵Toru Maesaka2009-01-091-1/+1
| | | | | | compile error in gcc 4.2.4 (Ubuntu 4.2.4-1ubuntu3) Signed-off-by: Dustin Sallings <dustin@spy.net>
* Fixed spelling of INADDR_ANYVictor Kirkebo2009-01-091-1/+1
|
* Realign command input buffer (binary protocol)Trond Norbye2009-01-031-0/+15
|
* keys are not zero-terminated: fix binary stats "reset" subcommandTrond Norbye2009-01-031-1/+1
|
* Added -C to remove the cas_id from the data structures to save spaceTrond Norbye2009-01-031-31/+40
|
* Fix for binary stats with subcommand.Dustin Sallings2009-01-031-5/+7
| | | | | | | | The subcommand is not necessarily a null terminated string (in the case of the binary protocol, it certainly isn't). The subcommand would not be recognized due to strcmp() failing. Tested with both text and binary protocols.
* Fix for stats opaque issue pointed out at the hackathon and removed some ↵Toru Maesaka2009-01-031-96/+151
| | | | wasteful function calls (more to come).
* Added support for detailed stats mode over the binary protocol.Toru Maesaka2009-01-031-2/+58
|
* Always use bundled daemonize function, fixed warnings on MacOSXTrond Norbye2009-01-031-8/+8
|
* Add -Wall to gcc compilationTrond Norbye2009-01-031-14/+24
|
* Removed compiler warnings and increased warning levelTrond Norbye2009-01-031-20/+11
|
* The slabber no longer needs a is_binary like flag for stats due to ↵Toru Maesaka2009-01-031-6/+6
| | | | abstraction by the callback.
* Fixed a memory corruption bug in stats.Toru Maesaka2009-01-031-1/+10
|
* Add DTrace probes to the binary protocolTrond Norbye2009-01-031-39/+79
|
* A callback function is used for stats request over ascii protocol as well now.Toru Maesaka2009-01-031-4/+25
|
* Changed the argument ordering for stats callback to something more common.Toru Maesaka2009-01-031-5/+5
|
* Report various listen failures, Updated Changelog/CONTRIBUTORS and Protocol ↵dormando2009-01-031-1/+9
| | | | Document.
* Remove managed instance mode (Dormando) - We should come back and do this ↵dormando2009-01-031-133/+1
| | | | right, probably built on top of the binary protocol and SE stuff.
* Warning messages added for setsockopt() and IPv6 patch for broken stack.Brian Aker2009-01-031-5/+22
|
* Fixed indent issues.Toru Maesaka2009-01-031-31/+32
|
* UDP/TCP can be disabled by setting their port to zero.Brian Aker2009-01-031-4/+4
|
* stats_prefix_* should not assume that all keys are zero-terminatedTrond Norbye2009-01-031-6/+6
|
* Prevent starvation by busy threads. This gives a configurable limit to how ↵dormando2009-01-031-2/+18
| | | | many commands to process before the next connection is checked.
* Changelog update for 1.2.6 release and Patch originally from the Facebook ↵dormando2009-01-031-3/+3
| | | | crew, cleaned up and split by dormando.
* Stats via ASCII protocol now queries the engine for non server keys.Toru Maesaka2009-01-031-23/+15
|
* Fix for duplicate stat entries.Toru Maesaka2009-01-031-4/+0
|
* GET related stats increment fix based on HIROSE Masaaki's patch.Toru Maesaka2009-01-031-0/+10
|
* Improved Stats error handling to discriminate between ENOMEM and ENOENT.Toru Maesaka2009-01-031-3/+8
|
* Support of stats sizes subcommand via the binary protocol.Toru Maesaka2009-01-031-1/+1
|
* Support of stats reset subcommand via the binary protocol.Toru Maesaka2009-01-031-4/+14
|
* Moved inclusion of malloc.h to slabs.h and cleaned whitespaces.Toru Maesaka2009-01-031-7/+0
|
* Support of malloc stats via the binary protocol.Toru Maesaka2009-01-031-16/+3
|
* Support of slabber stats via the binary protocol.Toru Maesaka2009-01-031-1/+1
|
* Support of item stats via the binary protocol.Toru Maesaka2009-01-031-14/+22
|
* Code cleaned by running devtools/clean-whitespace.pl script.Toru Maesaka2009-01-031-13/+13
|
* Applied Trond's false assertion removal patch and stamped to 1.3.1Trond Norbye2009-01-031-4/+0
|
* Minor formatting changes.Toru Maesaka2009-01-031-3/+3
|
* Storage of values with negative lengths was undefined, potential crash.dormando2009-01-021-1/+2
| | | | Reported on the mailing list by Marcus Hunger <hunger@sipgate.de>
* Don't leave stale data on failed set attempt.dormando2009-01-021-0/+11
| | | | This is mildly less confusing for folks with OOM errors or who store items for long periods of time, but occasionally attempt setting a very large item. While I don't believe this actually fixes a class of bugs we've seen due to refcount leaks, I make this change for consistency.
* Don't leak references in process_get_command.dormando2009-01-021-2/+11
|
* Don't leak reference on OOM error.dormando2009-01-021-0/+3
|
* Add DTrace probes for Solaris/etc.Trond Norbye2009-01-021-17/+59
|