4 /* FIXME: better place? */
5 int vasprintf (char **ptr, const char *fmt, va_list ap);
6 char *bprintf(const char*fmt, ...) _XBT_GNUC_PRINTF(1,2);
10 - Rename cursor to iterator
12 log.h still contains @name which break doxygen:
13 xbt/log.h:/** \name DEBUG
14 xbt/log.h:/** \name VERB
15 xbt/log.h:/** \name INFO
16 xbt/log.h:/** \name WARN
17 xbt/log.h:/** \name ERROR
18 xbt/log.h:/** \name CRITICAL
30 * Check the gcc version on powerpc. We disabled -floop-optimize on powerpc,
31 but versions above 3.4.0 should be ok.
32 * check whether we have better than jmp_buf to implement exceptions, and
33 use it (may need to generate a public .h, as glib does)
41 (errors, logs ; dynars, dicts, hooks, pools; config, rrdb)
44 * Mallocators and/or memory pool so that we can cleanly kill an actor
47 * Better split casual errors from programing errors.
48 The first ones should be repported to the user, the second should kill
49 the program (or, yet better, only the msg handler)
50 * Allows the use of an error handler depending on the current module (ie,
51 the same philosophy as log4c using GSL's error functions)
54 * Hijack message from a given category to another for a while (to mask
55 initializations, and more)
56 * Allow each actor to have its own setting
57 * a init/exit mecanism for logging appender
58 * Several appenders; fix the setting stuff to change the appender
59 * more logging appenders (take those from Ralf in l2)
62 * speed up the cursors, for example using the contexts when available
65 * better formalisation of what modules are (amok deeply needs it)
66 configuration + init() + exit() + dependencies
67 * allow to load them at runtime
68 check in erlang how they upgrade them without downtime
71 * we may need a round-robin database module, and a statistical one
72 * a hook module *may* help cleaning up some parts. Not sure yet.
73 * Some of the datacontainer modules seem to overlap. Kill some of them?
74 - replace fifo with dynars
75 - replace set with SWAG
82 * implement the P2P protocols that macedon does. They constitute great
86 * Spawn threads handling the communication
87 - Data sending cannot be delegated if we want to be kept informed
88 (*easily*) of errors here.
89 - Actor execution flow shouldn't be interrupted
90 - It should be allowed to access (both in read and write access)
91 any data available (ie, referenced) from the actor without
92 requesting to check for a condition before.
93 (in other word, no mutex or assimilated)
94 - I know that enforcing those rules prevent the implementation of
95 really cleaver stuff. Keeping the stuff simple for the users is more
96 important to me than allowing them to do cleaver tricks. Black magic
97 should be done *within* gras to reach a good performance level.
99 - Data receiving can be delegated (and should)
100 The first step here is a "simple" mailbox mecanism, with a fifo of
101 messages protected by semaphore.
102 The rest is rather straightforward too.
104 * use poll(2) instead of select(2) when available. (first need to check
105 the advantage of doing so ;)
107 Another idea we spoke about was to simulate this feature with a bunch of
108 threads blocked in a read(1) on each incomming socket. The latency is
109 reduced by the cost of a syscall, but the more I think about it, the
110 less I find the idea adapted to our context.
112 * timeout the send/recv too (hard to do in RL)
114 * multiplex on incoming SOAP over HTTP (once datadesc can deal with it)
116 * The module syntax/API is too complex.
117 - Everybody opens a server socket (or almost), and nobody open two of
118 them. This should be done automatically without user intervention.
119 - I'd like to offer the possibility to speak to someone, not to speak on
120 a socket. Users shouldn't care about such technical details.
121 - the idea of host_cookie in NWS seem to match my needs, but we still
122 need a proper name ;)
123 - this would allow to exchange a "socket" between peer :)
124 - the creation needs to identify the peer actor within the process
126 * when a send failed because the socket was closed on the other side,
127 try to reopen it seamlessly. Needs exceptions or another way to
128 differentiate between the several system_error.
129 * cache accepted sockets and close the old ones after a while.
130 Depends on the previous item; difficult to achieve with firewalls
133 * Implement gras_datadesc_cpy to speedup things in the simulator
134 (and allow to have several "actors" within the same unix process).
135 For now, we mimick closely the RL even in SG. It was easier to do
136 since the datadesc layer is unchanged, but it is not needed and
138 gras_datadesc_cpy needs to provide the size of the corresponding messages, so
139 that we can report it into the simulator.
140 * Add a XML wire protocol alongside to the binary one (for SOAP/HTTP)
144 * Inter-arch conversions
146 - Convert in the same buffer when size increase
147 - Exchange (on net) structures in one shoot when possible.
148 - Port to really exotic platforms (Cray is not IEEE ;)
149 * datadesc_set_cste: give the value by default when receiving.
150 - It's not transfered anymore, which is good for functions pointer.
152 - Cleanup the code (bison?)
153 - Factorize code in union/struct field adding
154 - Handle typedefs (needs love from DataDesc/)
155 - Handle unions with annotate
157 - Handle long long and long double
158 - Forbid "char", allow "signed char" and "unsigned char", or user code won't be
159 portable to ARM, at least.
160 - Handle struct/union/enum embeeded within another container
161 (needs modifications in DataDesc, too)
165 - Check struct { struct { int a } b; }
167 * gras_datadesc_import_nws?
170 * Other message types than oneway & RPC are possible:
171 - forwarding request, group communication
174 * Group communication
175 * Message declarations in a tree manner (such as log channels)?
177 [GRASPE] (platform expender)
178 * Tool to visualize/deploy and manage in RL
179 * pull method of source diffusion in graspe-slave
181 [Actors] (parallelism in GRAS)
182 * An actor is a user process.
183 It has a highly sequential control flow from its birth until its death.
184 The timers won't stop the current execution to branch elsewhere, they
185 will be delayed until the actor is ready to listen. Likewise, no signal
186 delivery. The goal is to KISS for users.
187 * You can fork a new actor, even on remote hosts.
188 * They are implemented as threads in RL, but this is still a distributed
189 memory *model*. If you want to share data with another actor, send it
190 using the message interface to explicit who's responsible of this data.
191 * data exchange between actors placed within the same UNIX process is
192 *implemented* by memcopy, but that's an implementation detail.
194 [Other, more general issues]
195 * watchdog in RL (ie, while (1) { fork; exec the child, wait in father })
196 * Allow [homogeneous] dico to be sent
197 * Make GRAS thread safe by mutexing what needs to be
204 * finish this module (still missing the saturate part)
205 * add a version guessing the appropriate datasizes automatically
207 * provide a way to retrieve the host load as in NWS
208 * log control, management, dynamic token ring
209 * a way using SSH to ask a remote host to open a socket back on me