- Nov 09, 2011
-
-
Konstantin Osipov authored
-
Konstantin Osipov authored
-
Konstantin Osipov authored
-
Konstantin Osipov authored
-
- Nov 07, 2011
-
-
Konstantin Osipov authored
-
Konstantin Osipov authored
-
Konstantin Osipov authored
Fix a bug in next slot calculation (creeped in with code review fixes).
-
Dmitry Simonenko authored
-
Dmitry Simonenko authored
-
- Nov 04, 2011
-
-
Konstantin Osipov authored
Conflicts: mod/box/box.m mod/box/index.m
-
Konstantin Osipov authored
Add comments. Rewrite put_slot loop to make it *both* clear and efficient. There is no need to check for equality in put_slot loop once we meet first non-dirty element: a non-dirty link indicates the end of the collision chain. Do not move elements up the chain in put_slot loop: we may use index in the hash/tree as tuple id in secondary keys.
-
- Nov 03, 2011
-
-
Konstantin Osipov authored
-
Konstantin Osipov authored
-
- Nov 01, 2011
-
-
Konstantin Osipov authored
Fix a compile failiure when BFD is not installed. Make test results execution-order independent (fiber id depends on how many fibers were created so far).
-
Konstantin Osipov authored
-
- Oct 28, 2011
-
-
Konstantin Osipov authored
Conflicts: test/box/lua.result test/box/lua.test
-
Konstantin Osipov authored
Add box.fiber.create(), box.fiber.resume(), fiber.yield(), box.fiber.detach(). Add tests. Makce fiber cancellation implementation more robust and quick, so that it is actually usable from Lua. Debug.
-
Dmitry Simonenko authored
-
- Oct 26, 2011
-
-
Dmitry Simonenko authored
-
Dmitry Simonenko authored
-
- Oct 20, 2011
-
-
Konstantin Osipov authored
Implement box.tuple.unpack(), which converts the tuple to a list of lua atoms.
-
Konstantin Osipov authored
-
Konstantin Osipov authored
-
Konstantin Osipov authored
-
Konstantin Shulgin authored
-
Konstantin Shulgin authored
-
Dmitry Simonenko authored
-
- Oct 19, 2011
-
-
Konstantin Osipov authored
Add basic support for splice in box.update() operation.
-
Dmitry Simonenko authored
-
Dmitry Simonenko authored
-
Dmitry Simonenko authored
-
- Oct 18, 2011
-
-
Dmitry Simonenko authored
-
Dmitry Simonenko authored
-
Konstantin Osipov authored
-
Konstantin Osipov authored
Creating a new userdata object whenever a fiber is accessed is inefficient and complicates fiber garbage collection: if a fiber is referenced to by multiple userdata instances, it's hard to know when to free it up. Implement 'memoize' pattern for fiber userdata to deal with these two issues. After this patch, no matter how and where from one gets a reference to a fiber, it's the same userdata object. As a side effect, getting fiber identity in Lua is also easy -- one can just directly compare two fibers to tell whether or not they are the same.
-
Konstantin Osipov authored
-
- Oct 17, 2011
-
-
Konstantin Osipov authored
Add fiber.cancel(), fiber.sleep(), fiber.testcancel(), fiber.id(). Draft the fiber library interface description.
-
- Oct 13, 2011
-
-
Konstantin Osipov authored
Update ER_NO_SUCH_PROC, if the procedure doesn't exist, quote the procedure name, since it can be empty.
-
Konstantin Osipov authored
Check that --background actually works, fix the coding style.
-
Konstantin Osipov authored
A fix and a test case for https://bugs.launchpad.net/tarantool/+bug/750658 --background neither closes nor redirects stdin/stdout/stderr --background option didn't work properly, since stdin/stdout/ stderr streams were left open and pointing to a terminal. This lead to a hang when tarantool was started from a shell script or over ssh. If --background option is given, we need to fork, try to create a pid file, close stdin/stdout/stderr and then initialize the logging subsystem.
-