Commit Graph

  • 0007febea3 Oh for fuck's sake. Don't rename Velocity because they do a bunch of pointless reflection internally to look default implementations by class name. Sigh. Michael Bayne 2010-10-13 00:30:19 +00:00
  • 3462ee2661 Exception is already Serializable and GWT accepts Serializable in lieu of IsSerializable. If possible, it would be nice to keep GWT references out of narya-base (where com.threerings.web ends up). Michael Bayne 2010-10-12 18:05:30 +00:00
  • f0a9dc8dd7 We must be vigilant about overriding Ant's retarded defaults. In the latest version of Ant, it issues a warning if you fail to do so. Genius! Michael Bayne 2010-10-12 17:44:18 +00:00
  • f4af1915b0 Build narya-tools with jarjar so as to hide its included dependencies in the unfortunate event that someone puts it on a real project classpath. Michael Bayne 2010-10-12 17:35:41 +00:00
  • 480893d0b0 No gwt-dev needed, we're not compiling. Par Winzell 2010-10-07 20:32:42 +00:00
  • bdff124525 Strings have length, Zell. Par Winzell 2010-10-07 20:23:24 +00:00
  • 72d0016963 I guess I have to actually use a Java5 SDK for Narya. How's that whole 'convert to java 6' thing going, yohoho folks? Par Winzell 2010-10-07 20:10:34 +00:00
  • d4ee1a9d4e We don't have GWT until we have gwt-utils, too, I guess. Par Winzell 2010-10-07 19:59:57 +00:00
  • f7bd94543f Delete brief experiment. Par Winzell 2010-10-07 19:29:48 +00:00
  • e0e47a1a1a Build some GWT bits, if GWT is present. Par Winzell 2010-10-07 19:22:20 +00:00
  • d26267c704 GWT frontend and servlet for editing runtime configuration. Par Winzell 2010-10-07 19:21:48 +00:00
  • 3069062034 Some basic GWT infrastructure. Par Winzell 2010-10-07 19:21:24 +00:00
  • 1b30ef96cf Quell enthusiastic logging for an inevitable action. Ray Greenwell 2010-10-07 17:28:49 +00:00
  • ba118de010 Looks like this should just return if it can't get the value, rather than set the (possibly non-nullable) value to null. Andrzej Kapolka 2010-10-05 21:03:09 +00:00
  • 26d497861f Fix a bug where the client & server happen to attempt to change location nearly simultaneously. Mike Thomas 2010-10-05 20:54:41 +00:00
  • b73e76d19b Let's use StringUtil.shortClassName() which handles null and also snips off the package, since that's generally TMI anyway. Michael Bayne 2010-10-05 20:27:31 +00:00
  • adf1f5af37 Null values don't have classes. Andrzej Kapolka 2010-10-05 20:23:03 +00:00
  • e91aeaf3de ConnectionManager needs to be injectable for things not using BindingConnectionManager Charlie Groves 2010-09-22 01:22:49 +00:00
  • 55fb708832 Restructured BindingConnectionManager so that the listen sockets are included with all the other sockets and we do only a single select. Nixed the SocketChannelAcceptor as it relies on having its own Selector which is not desirable here or in Samsara. Michael Bayne 2010-09-20 23:29:18 +00:00
  • 22aefb9aef Some tidying up and comment improvements. Michael Bayne 2010-09-20 19:40:50 +00:00
  • e574ad7215 Source headers. Michael Bayne 2010-09-20 19:16:15 +00:00
  • 95dd3265d8 Wrong "its" Dave Hoover 2010-09-17 20:38:27 +00:00
  • 308a6f9d52 Turns out presents' datagram support doesn't make sense at an individual datagram level, and requires a supporting Connection provided by the ConnectionManager. As such, there's no point in adding an additional selector for datagrams outside the ConnectionManager, and this moves it back into BindingConectionManager. Charlie Groves 2010-09-16 22:47:58 +00:00
  • 08a4a5281d tomato tomato Charlie Groves 2010-09-16 20:02:20 +00:00
  • e75d4917c8 Not used Charlie Groves 2010-09-16 18:23:30 +00:00
  • 68a5c21cda Allow the select wait for ConnectionManager to be injected eg bind(Integer.class).annotatedWith(IncomingEventWait.class).toInstance(1); will make the select wait 1 millisecond. Charlie Groves 2010-09-16 18:23:10 +00:00
  • fc13de8b9a Nix unused import Dave Hoover 2010-09-16 17:45:51 +00:00
  • ffd5b490bf Fix javadoc Dave Hoover 2010-09-16 17:45:50 +00:00
  • 1645c23be5 No @Overrides for interfaces in java 1.5 Dave Hoover 2010-09-16 17:45:50 +00:00
  • 5c411bf6e7 There's a method for that. Ray Greenwell 2010-09-16 17:38:29 +00:00
  • e24ddecabb Fucking Java 5 Charlie Groves 2010-09-16 03:13:42 +00:00
  • a0d4296b4e Split tcp socket opening and datagram listening out of ConnectionManager to allow sockets to remain open when a ConnectionManager using them shuts down. Charlie Groves 2010-09-16 02:49:29 +00:00
  • 0d4837e71d ConnectionManager allowed a server to listen on multiple tcp and datagram ports, but it only closed the last opened. Close all opened listening sockets. Charlie Groves 2010-09-14 22:47:45 +00:00
  • c2ebf2884b Ray points out this can usefully be an AtomicInteger. Par Winzell 2010-09-13 17:35:23 +00:00
  • 6f7101404b Don't go beyond 100. Par Winzell 2010-09-13 17:15:37 +00:00
  • 3fd045a01d Error error. Par Winzell 2010-09-11 17:06:29 +00:00
  • 2180623f37 Don't modify the set we're iterating over. Par Winzell 2010-09-11 17:02:48 +00:00
  • d5dcc4a1eb Nixed some cruft, updated some source headers. Michael Bayne 2010-09-11 04:30:16 +00:00
  • ba0ba7f659 Instead of NPEing when we reset _oin at a bad time, let's throw an exception we can better handle and reduce some log spew. Mike Thomas 2010-09-09 17:35:44 +00:00
  • 83e5576be0 IDEA module files. You'll need to name your SDK 'Flex SDK (player 9)' for these to work. I have eschewed modules exporting the modules they depend on forward; partly because this doesn't work right for Flex compilation anyway, and partly because we do explicit dependencies everywhere else, so why not here? Each module explicitly lists the libraries and modules on which it depends. Par Winzell 2010-09-09 17:23:57 +00:00
  • 908bee05c4 Stop lying Charlie Groves 2010-09-08 19:29:51 +00:00
  • be64ef83f1 Create the directories for generated files if necessary Charlie Groves 2010-09-07 23:45:04 +00:00
  • d35526a3c4 Unused imports Charlie Groves 2010-09-07 21:03:33 +00:00
  • e7bd54fa2d Fix up imports, cuddle braces, use fancy .concat(). Par Winzell 2010-09-07 20:08:25 +00:00
  • 73e44953d7 Argh. Java6->5 fixes. Par Winzell 2010-09-07 19:26:42 +00:00
  • 59588467f7 Forgot the final message sort. Par Winzell 2010-09-07 19:13:02 +00:00
  • 628e03beed As the first application of the new invokeNodeRequest, rip all the chat collection stuff out of CrowdPeerManager and into the chat package where it belongs. Par Winzell 2010-09-07 19:10:00 +00:00
  • 5365b53ee6 Let's take invokdeNodeRequest the final distance; NodeRequest allows each peer to submit a result, so let's pass those results back to the caller instead of throwing them away. With this, we should be able to stop shoving all complex peer operations into PeerManager subclasses, and instead sprinkling them as NodeRequests in the sub-packages in which they belong. Par Winzell 2010-09-07 19:08:26 +00:00
  • a9c4b051e0 Let's not flip out on templated fields. Par Winzell 2010-09-06 01:15:49 +00:00
  • 21d985b468 Speeling Dave Hoover 2010-09-03 00:29:55 +00:00
  • 68bb635152 Extending Object should be implicit. Bruno Garcia 2010-08-30 22:52:38 +00:00
  • 73d2bd1741 Rather than delaying datagrams if we've exceeded the throttle, just drop them. Andrzej Kapolka 2010-08-28 02:30:17 +00:00
  • 90929ab5d7 Fix for reconnection issue in Project X. When clients disconnect, we need to reset the throttle state. When they reconnect, we need to resend the message rate. Andrzej Kapolka 2010-08-28 02:14:09 +00:00
  • 9bdf1f8ef1 Serialize the request early so that developers find out as soon as possible if they're fucking up. Also fixed up a comment. Michael Bayne 2010-08-27 23:02:23 +00:00
  • d99d17ca7a Make NodeRequests a bit fancier for hitting multiple nodes. Mike Thomas 2010-08-27 01:06:06 +00:00
  • 0e575a8b6f Allow peers to specify a namespace. If they do, only hook up to peers with node names starting with the namespace. Charlie Groves 2010-08-24 22:51:15 +00:00
  • 543a0d8a57 Allow subclasses to change how signal handlers are registered Charlie Groves 2010-08-23 22:28:29 +00:00
  • 8818511502 Pass the receiver code through Charlie Groves 2010-08-16 21:39:48 +00:00
  • 2cdc445b1f Have ServiceDescription carry skipAS like it does the rest of the processing information Charlie Groves 2010-08-12 21:00:49 +00:00
  • c12e1e44b3 Multimap is pretty rad Charlie Groves 2010-08-12 20:59:47 +00:00
  • 696f77beb5 Fix possible ClassCastException, because DSet.contains(E entry) will type-erase to DSet.Entry. Ensure objects tested for containment in our Set view implement that. Ray Greenwell 2010-08-11 20:34:27 +00:00
  • 891ca8db44 Since we have a runtime type token here, it's best to avoid the @SuppressWarnings annotation. Ray Greenwell 2010-08-11 20:32:26 +00:00
  • 11c52c06eb Whitespace Dave Hoover 2010-08-11 18:09:12 +00:00
  • 1108d3c0db Factory methods for StreamableArrayList and StreamableHashSet Dave Hoover 2010-08-11 18:09:11 +00:00
  • 0bce227a40 Map/Set streaming for actionscript. Ray Greenwell 2010-08-11 03:53:06 +00:00
  • 4bf5490c35 Removed accidentally committed debuggery. Ray Greenwell 2010-08-11 03:39:30 +00:00
  • f98e5e9abf Gonna need that name before we try to load it Charlie Groves 2010-08-11 03:19:01 +00:00
  • 7badbbd20f This AssertionError is new. Mdb added it. It freaks something out in p-x. Ray Greenwell 2010-08-11 00:42:32 +00:00
  • c1ac294171 Succincter Charlie Groves 2010-08-11 00:27:48 +00:00
  • 0771006ab0 Expand these tests for maps, sets, collections. Ray Greenwell 2010-08-10 23:38:25 +00:00
  • ee1d915e91 List streams differently now. Ray Greenwell 2010-08-10 23:20:39 +00:00
  • 2f039b370d Changed this, as List is streamed differently now that the bug with it is fixed. Ray Greenwell 2010-08-10 23:20:03 +00:00
  • 931d72f4b5 Streaming changes. - Fix the bug with streaming an object that had a field typed as merely 'List'. This will result in an incompatible change. Notably, BootstrapData has a List field, but this shouldn't be accessed until after versions are checked. The bug also meant that such fields could only hold an ArrayList. Now they may hold any List type and will turn into an ArrayList on the other end... - Support Map and Set streaming in the same way. - Support using unmodifiedUTF if the "com.threerings.io.unmodifiedUTFStreaming" system property is set to the String "true". Ray Greenwell 2010-08-10 23:18:02 +00:00
  • 5c7d59e597 This is so succinct now, we can dispense with the helper method altogether. Michael Bayne 2010-08-10 22:09:33 +00:00
  • 83c5fcfb15 Whoa, Daddy-O! This is almost like a real abstraction in Java. Michael Bayne 2010-08-10 22:07:48 +00:00
  • 89dfe715d8 Switch class to streamer mapping to a Map. Whoa! I used a linked hash map to preserve the registration order, in case that's significant. It would be nice if Guava had an ImmutableLinkedHashMap, but I don't see such a thing. Michael Bayne 2010-08-10 22:04:05 +00:00
  • 41ba838b20 More supported types, and a less heinously formatted blob. Michael Bayne 2010-08-10 21:46:59 +00:00
  • 7967925920 More types! Michael Bayne 2010-08-10 21:30:05 +00:00
  • 589b2128e3 Made StreamableTest actually test something. Specifically: Michael Bayne 2010-08-10 21:24:26 +00:00
  • 4ea22628f7 Generalize velocity creation and classloader setting for code generation ant tasks into GenTask Charlie Groves 2010-08-10 00:38:46 +00:00
  • 1236121ff6 Add names to the builtin FieldMarshallers to make them differentiable when printed Charlie Groves 2010-08-09 22:23:32 +00:00
  • decb56970c Use the Maps builder method. Assign the DEFAULT_MAP if we unstream an empty map, too. Comment DEFAULT_MAP. Ray Greenwell 2010-08-09 18:37:54 +00:00
  • e3ae54ddb4 Add a comment so people don't accidentally new StreamableHashMap(), and expect it to work. Nathan Curtis 2010-08-09 18:34:26 +00:00
  • b6174f659f If no map class is passed in (should only happen on deserialization, and the real source map is provided on readObject), pass an empty immutable map to super, which is less weird and probably better than just ignoring and forcibly not calling the super constructor, which is what was happening. Nathan Curtis 2010-08-09 18:30:45 +00:00
  • 4e8c8bce3c Comment fixes. Object[] is put in _marshallers as a wrapped BasicStreamer below, so don't act like we're handling it with the generic marshaller. Charlie Groves 2010-08-08 22:19:17 +00:00
  • 00a5e66f92 Style. Bruno Garcia 2010-08-03 02:48:42 +00:00
  • 03ae523e62 Typo. Bruno Garcia 2010-07-29 20:35:45 +00:00
  • 7b86f2d8db Prefer StringBuilder where possible. Ray Greenwell 2010-07-22 01:20:02 +00:00
  • 1bfc69aa49 Seems byte was only partially added here. Andrzej Kapolka 2010-07-20 01:11:55 +00:00
  • 73900f891d Oh dear, fix already used variable name. Ray Greenwell 2010-07-19 18:08:54 +00:00
  • 353fdcd7cf I'm trying to excise uses of com.samskivert.util.Collections, which is full of some weird old stuff, much of it replaced by using Guava or newer 1.5 features. Ray Greenwell 2010-07-19 17:53:59 +00:00
  • 126ee1f738 Guavafication. A little ugly. Ray Greenwell 2010-07-16 01:19:36 +00:00
  • dc11108af4 Some tweaks to getTimeString. Dave Hoover 2010-07-14 20:59:02 +00:00
  • 6bff220be7 Prefer java standard libraries, then guava, then samskivert. Ray Greenwell 2010-07-14 19:28:14 +00:00
  • cdebcfa71d StringUtil.replace is deprecated Dave Hoover 2010-07-13 21:45:39 +00:00
  • 3340e270ae Whitespace Dave Hoover 2010-07-13 21:45:38 +00:00
  • b8e029fdad In with the new logging Dave Hoover 2010-07-09 23:33:38 +00:00
  • 46835601af Missing newline in report. Par Winzell 2010-07-02 13:32:49 +00:00
  • 4ddc60e369 Integer was getting envious of Short's MIN/MAX_VALUE constants, so let's appease it. Mike Thomas 2010-07-01 20:17:52 +00:00
  • 2f704712b2 I am of two minds whether this is the "right" way or not, but Eclipse's warnings sure think so. Since the generics portion has been erased we can only check against <?> but we at least are noting that it's a parameterized type. Mike Thomas 2010-06-28 17:19:41 +00:00