1 SimGrid (3.4-svn) unstable; urgency=high
4 * fix a bug on struct sizeof computation, which prevented the
5 exchange of arrays of structs in some conditions
6 * Allow the exchange of 0-long dynamic vectors.
7 - for that, use -1 as indicator of dynamic size instead of 0
8 - This implied to change any size from unsigned long to long,
9 reducing a bit communication abilities, but I guess that with
10 64bits being quite common, this is more than enough.
11 - This also induce a protocol change, thus bumping network protocol
12 version from 0 to 1 (if we have external users, we have to get
13 clean on that point too ;)
16 * Allow to control the simulation from a trace file.
17 New functions MSG_action_register() and MSG_action_trace_run()
18 The first one allows to associate a function execution to each
19 kind of action while the second one parses a trace file and
20 triggers the corresponding actions within the system.
21 For now, only a toy example is provided in examples/msg/actions
22 * Add an exemple of process migration in examples/msg/migration
23 * Fix a bug in task exchange which broke MSG_task_get_sender()
24 Add a teshsuite regression test for that.
25 [Bug: if MSG_task_get_sender() is called after sender exit,
29 * Add SIMIX_process_set_name() to change the name of the current
30 process in the log messages.
33 * Also include strbuff from xbt.h public header
34 * xbt_ex_display(): do not free the exception after displaying
35 This allows to do more with the given exception afterward.
36 Users should call xbt_ex_free() themselves.
38 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
40 SimGrid (3.3) stable; urgency=high
44 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
47 * Introduce the SIMIX module: factorize code between MSG and GRAS.
50 Until now, GRAS were using MSG as an interface to SURF. It was
51 quite difficult because both interface have several differences
52 (MSG channels vs GRAS sockets were the most notable point).
54 This also opens the gate to SMPI (which should occur soon) and speed
55 up simulations by to 40% (even if it were not the main goal).
57 **************************************
58 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
59 **************************************
60 The point is that events occuring at the exact same timestamp are
61 not scheduled in the same order with the old and new version. This
62 may be enough to completely change the execution of simulations in
63 some cases. Sorry for the inconvenience.
65 * Cleanup and upgrade the XML format to push further scalability
66 issues (check http://hal.inria.fr/inria-00256883/ for more info)
68 * Improve the testing infrastructure with tesh. Now a very large part of
69 the code is tested not only by being run but also by checking that the
70 output match an expected output [Mt].
72 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
73 is really less memory-demanding, which should allow you to use
74 larger files in SimGrid [AL].
76 * Inform valgrind about our contextes, so that it becomes usable
77 with the default (and more effecient) version of SimGrid
78 [contributed by Sékou Diakite, many thanks]
81 * Introduce a listener thread in charge of receiving incomming
82 messages from the network. It allows to overlap communication and
83 computation but most notably, it removes some stupid deadlocks due
84 to the fact that so far, a process could not send and receive at
85 the same time. This made most non trivial communication schema
87 * Convert the PIDs from long int to int to match the MSG ones (and
89 * New function: gras_agent_spawn() to launch a new process on
90 current host. Only working in simulation for now. [Mt]
91 * New function: gras_os_hostport() returning a constant form (ie,
92 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
95 * Make the backtrace of exceptions more human readable [Mt]
96 * New module: xbt/str [Mt]
97 a ton of string utility functions (split, join, printf to a newly
98 allocated buffer, trim, etc)
99 * New module: xbt/hash [Mt]
100 SHA1 hashing algorithm (more to come if needed)
101 * New module: xbt/synchro [Mt]
102 synchronization tools (mutex and conditions) working the same way
103 in simulation and in real life (mainly useful for GRAS, but not
105 * New module: xbt/queue [Mt]
106 classical producer/consumer synchronization scheme
107 * xbt_dynar_new_sync() creates a synchronized dynar. All access
108 (using the classical functions will get serialized) [Mt]
109 * Make dictionary internal table dynamic. No need to specify its size
110 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
112 * Make sure the log channels are organized as a tree under windows
113 (because of ANSI C compatibility issue, any channel were child of
117 * Cleaned many thing in surf and fixed a few bugs [AL].
118 * Add a nice command line configuration mechanism to compose models [AL].
119 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
120 the previous one (KCCFLN05). It relies on something that looks like
121 a max-min sharing mechanism but cannot be written as such. A new solver
122 was thus designed [AL].
123 * Add a new solver to lmm. Based on Lagrange optimization and
124 gradient-based descent, it enables to efficiently maximise systems s.a
126 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
128 This solver enables to propose two new network models for TCP Reno and
129 TCP Vegas based on Low's work. These models still need to be fully
130 tested though [Pedro Velho].
133 * Bug fix in SD_simulate. Now the time bound given as argument is
135 * Use the new parallel task model (ptask_L07) as default.
136 * Use the SURF command line configuration mechanism.
137 * 0-size tasks (for synchronization) should now work.
139 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
141 SimGrid (3.2) stable; urgency=high
145 We still experience issues on this platform, but we believe that at
148 GRAS API BREAKAGE (for simplification purpose, sorry):
149 * the gras_msgtype_by_name is not used anymore. Instead of
150 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
151 you can write (and must)
152 gras_msg_send(toserver, "request", &request);
153 - If you still want to pass a gras_msgtype_t to the function (to cache
154 the type and avoid the lookup time), use the gras_msg_send_() variant.
155 - Impacted functions:
156 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
157 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
158 * The callbacks are now expected to return 0 when everything went well
159 (just like the main() function)
161 GRAS new features and improvements:
162 * New module mecanism where user code can use per process globals [Mt]
163 This is similar to gras_userdata_*() functions, but for libraries. It
164 factorize some code developped over and over in the examples and AMOK.
165 It has still to be documented and used (only amok/peermanagement is
167 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
168 loss when they were small enough to fit into the buffer and sent quickly
169 enough so that they can all get received in one shoot.
170 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
171 if not found. Use the *_or_null() variant for the old semantic.
172 * In gras_msg_handle, do not discard messages without callback.
173 They are probably messages to be explicitly awaited later (ie, proofs of
174 mis-synchronization in userland since they are sent before being awaited)
176 * gras_socket_meas_send/recv: semantic changed!
177 The numerical arguments used to be (1) the total amount of data to send
178 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
179 messages. This was need for the fool willing to send more than MAXINT
180 bytes on quite fat pipes.
183 * Do really rename the hostmanagement module to peermanagement. [Mt]
184 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
185 but this is rather new and this was documented in the module
186 documentation (poor excuses, I admit)
187 * Bandwidth measurement semantic changed! This follows the changes to
188 gras_socket_meas_send/recv explained above.
191 * A sequential mode has been added to the workstations. When a workstation
192 is in sequential mode, it can execute only one task, and the other tasks
193 are waiting in a FIFO. [Christophe Thiery]
196 * The KCCFLN05 workstation model now handles parallel tasks. It is the
197 model for SIMDAG. [Christophe Thiery]
198 * Bug fix in the maxmin solver: Some values were close to 0 instead of
199 equal to 0, which caused some bad behaviors in
200 saturated_constraint_set_update. I now use a threshold mechanism like in
204 * When running manually src/testall, you select specific units [Mt]
205 testall is the result of our cunit mecanism, and should replace all
206 the scripty thingy around since bash don't run easily on billware.
208 * A mallocator system has been added. [Christophe Thiery]
209 Mallocators allow you to recycle your unused objects instead of freeing them
210 and allocating new ones.
212 Documentation update:
213 * FAQ reworking + New FAQs:
214 - "Valgrind spits tons of errors!" [Mt]
215 - "How to repport bugs" [Mt]
216 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
217 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
218 - Communication time measurement within MSG [AL]
219 - I experience weird communication times when I change the latency [AL]
222 - an introduction to the framework and to the used communication model
223 - an initiatic tour introducing the most proheminent features:
225 . Lesson 0: Installing GRAS
226 . Lesson 1: Setting up your own project
227 o Part 2: Message passing
228 . Lesson 2: Exchanging simple messages
229 . Lesson 3: Passing arguments to the processes (in SG)
230 . Lesson 4: Attaching callbacks to messages
231 . Lesson 5: Using globals in processes
232 . Lesson 6: Logging informations properly
233 . Lesson 7: Using internal timers
234 . Lesson 8: Handling errors through exceptions
235 . Lesson 9: Exchanging simple data
236 . Lesson 10: Remote Procedure Calling (RPC)
237 . Lesson 11: Explicitely waiting for messages
238 . Recapping of message passing features in GRAS
239 - A HOWTO section containing:
240 o HOWTO design a GRAS application
241 More are due, of course. They will come latter. In the meanwhile, you can
242 check the examples which are still here.
244 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
246 SimGrid (3.1) stable; urgency=high
250 There was a stack corruption somewhere, visible only when optimizing
251 with these versions. [Vince]
254 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
255 module that enabled to play with Directed Acyclic Graphs. It is built
256 directly on top of SURF and provides an API rather close to the old
257 SG. Some old codes using SG are currently under rewrite to check that
258 all needful functions are provided. [Christophe Thiery]
261 * Complete rewrite of the KCCFLN05 workstation model. It is now an
262 extension of the classical CLM03 model that gracefully handles
263 failures. This is now the default model for MSG and GRAS. It doesn't
264 handle parallel tasks yet though. [AL]
265 * Bug fix: Weights were not correctly set in the network part.
266 WARNING: This may have resulted in incorrect results with simulations
267 where there are more than one flow on a given link. [AL]
270 * After a (long ?) discussion on simgrid-devel, we have decided that the
271 convention we had on units was stupid. That is why it has been decided
272 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
273 WARNING : This means that all previous platform files will not work as
274 such with this version! A warning is issued to ask users to update
276 A conversion script can be found in the contrib module of the CVS, under
277 the name contrib/platform_generation/surfxml_update.pl [MQ]
280 * Bug fix: Processes were started in reverse order, wrt deployment file.
281 WARNING: if your code relies on this bug, please fix it. [AL]
282 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
283 being executed on two different locations. [AL]
284 * Bug fix: Failures are now better supported thanks to Derrick's tests
285 (there was many failure situations I hadn't thought of and that weren't
286 correctly handled). [AL]
287 * New function: MSG_host_is_avail indicates you whether a given m_host_t
291 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
292 Exception occurring on server-side are propagated back to client (!).
294 API CHANGE: the callback changed their prototype. Change:
295 int my_handler(gras_socket_t expeditor, void *payload_data) {
297 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
298 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
300 * New! function: gras_msg_handleall to deal with all messages arriving
301 within a given period.
302 * New! function: gras_socket_server_range to get a server socket in a
303 range of port numbers (ease to avoid port number conflicts) [MQ]
304 * New! gras processes display their backtrace when they get a SIGUSR1
305 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
306 Sweet to debug RL processes [MQ]
310 - Do not force experiment sizes to be expressed in kb, or it becomes
311 impossible to measure the latency this way (needs one byte-long tests)
312 WARNING: this changes the amok_bw_* function semantic. [MQ]
313 - Implements the link saturation stuff. [MQ]
314 * Peer management module:
315 New! module factorizing code that we wrote over and over [MQ].
318 * New module: cunit (my jUnit implementation in ansi C) [MQ]
319 - Test units are placed directly into the library code, they get extracted
320 automatically and placed into the src/testall binary.
321 - Convert most of the XBT tests to this system.
322 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
323 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
324 live in the same C code. This required to change a little bit the API
325 of surfxml parsing but shouldn't be an issue for end-users. [AL]
326 * New module: sparse graph structure with basic algorithms (this is work
327 in progress and the API is not considered to be frozen yet). [AL]
328 * Display more information on backtraces: source line & function names are
329 now displayed just like valgrind does (rely on addr2line tool) [MQ]
330 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
331 * Reworked a little bit some #include statements to load only required
332 headers. Some user code that relied on SimGrid to include stdlib or
333 stdio may need to include it by themselves. [AL]
334 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
336 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
337 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
338 but I guess that almost nobody used those parts. [MQ]
340 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
342 SimGrid (3.0.1) stable; urgency=low
345 * Unfortunately, I had missed 5 misnamed functions:
346 xbt_fifo_item_t xbt_fifo_newitem(void);
347 void xbt_fifo_freeitem(xbt_fifo_item_t);
348 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
349 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
350 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
351 They're now deprecated. Please use their new versions:
352 xbt_fifo_item_t xbt_fifo_new_item(void);
353 void xbt_fifo_free_item(xbt_fifo_item_t);
354 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
355 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
356 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
358 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
359 * Documentation: xbt_log module unmercifully reworked [MQ]
360 * Bugfix: there was a problem with the ending of contexts with
361 the pthread backend. It caused some weird deadlock or behavior
362 depending on the pthread implementation. [AL]
363 * Bugfix: get the exceptions raised in the simulator repport where
364 and why they come from when they are not catched in time [AL, MQ]
367 * Bugfix: Do repport the error when two non-connected hosts try to
368 exchange data (Thanks to Flavien for stumbling into this one) [AL]
371 * Add additionnal checkings on communications. Assert that two
372 communicating hosts are connected by a set of links... [AL]
375 * Add additionnal checkings on channel values in communication [AL]
376 * New: MSG_task_get_source to see on which host a task was generated [HC]
377 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
378 the number of tasks waiting to be received on channel and sent
380 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
381 waits for the first task coming from a given host.. [AL]
383 GRAS new functionnalities: [MQ]
384 * Enhance the parsing macro to allow the size of multidimentional objects
385 to be given thru annotations.
386 * New example (and documentation): Matrix Multiplication a la RPC
387 (as when I was young!) and fix a bunch of bugs found on the way.
389 GRAS performance improvements: [MQ]
391 * Reduce the amount of cbps creation/destruction by making it static to
392 datadesc_send/recv() and using a (newly created) cbps_reset (based on
395 * Change libdata to a set so that we can search for stuff by ID (and thus
396 reduce the insane amount of dict lookups)
399 * Actually implement gras_datadesc_copy() so that we don't have to mimick
400 RL communication on top of SG since it's so uneffective.
401 It may also be used for inter-thread communication in RL, one day.
402 * Use gras_datadesc_copy() to exchange messages on top of SG
404 - improve message exchange performance on top of SG
405 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
406 * Don't exchange on the network the size of the used part of buffer,
407 instead, specify the possible buffer size to read().
409 - reduces the amount of read/write calls (one pair per exchange)
410 - reduces the amount of exchanged data (the size)
411 - allows to retrieve all arrived data on receiver side, if we don't need
412 it right now (subsequent read will peek the buffer)
413 - allows the receiver to proceed with the begining of the stream before
414 everything is arrived
415 - make it possible to build an iov transport (using readv/writev)
417 - take care of the data with non-stable storage (like stacked data),
419 * If possible, TCP send uses vector I/O (when writev() is here)
420 - Don't use it for receive since we send data sizes and data on the
421 same stream, so we wouldn't be able to chain large amount of chunks
422 before having to flush the stuff to read the size.
423 * Rework the transport plugin mecanism to simplify it and reduce the
424 amount of pointer dereferencement when searching for the right function
427 * I guess that now, we do almost as few system calls as possible while
428 doing as few data copy as possible.
430 To improve it further, we could try to send all the sizes first and then
431 all the data (to use iov on receiving size), but it's only a partial
432 solution: when you have 2 dimensional data, the sizes of the second
433 dimension is data of the first dimension, so you need 3 streams.
435 I'm not sure the potential performance gains justify the coding burden.
437 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
439 SimGrid (3.00) stable; urgency=high
442 * New! Give the possibility to hijack the surf parser and thus bypass
443 MSG_create_environment and MSG_launch_application. Have a look at
444 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
446 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
448 SimGrid (2.96) unstable; urgency=low
453 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
454 This deprecates the xbt_error_t mecanisms.
455 It modifies (simplifies) all XBT and GRAS API.
456 MSG API keeps unchanged (exceptions raised by XBT are catched from
457 within MSG and masked with existing error handling facilities)
460 * New! Add a FATPIPE model. [AL]
461 * New! Add a parallel task model. [AL]
462 * New! Add automatically a loopback interface (in the default
463 network model) if none was precised.
466 * Bugfix: MSG_process_resume now works with the current running process.
468 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
469 * Modification of MSG_task_get_compute_duration. Once a task has been
470 processed, the value returned by this function is now equal to 0. [AL]
471 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
472 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
473 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
474 of a m_task. For now, MSG_task_cancel only works with computation
476 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
477 of the processor (in Mflop/s) regardless of the current load on the
479 * API Change: use proper naming convention for MSG_getClock and
480 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
482 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
483 This function changes the priority of a computation task. This priority
484 doesn't affect the transfer rate. A priority of 2 will make a task
485 receive two times more cpu power than the other ones. This function
486 has been added to suit the needs of Nguyen The Loc and hasn't been that
487 much tested yet. So if it fails, please report it and send me your code.
489 * API Change: removed all functions and types that were marked "deprecated"
490 since many months. Renamed MSG_global_init_args to MSG_global_init.
492 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
494 SimGrid (2.95) unstable; urgency=low
497 * Steal some nice code to GNU pth to fix context detection and usage [AL]
498 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
499 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
502 * Bugfix: Allow absolute paths for platform description files [MQ]
503 * Bugfix: do free the variables after use. Leads to drastic performance
505 * Implement max_duration (ie, timeouts) on resources [AL]
508 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
510 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
511 the needed support (timeouts) [AL]
514 * Implement measurement sockets. You can now get the bandwidth between two
515 hosts thanks to AMOK (see below). [MQ]
516 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
517 dynar over the network (yeah) [MQ]
518 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
519 * simple/static token-ring example. [Alexandre Colucci and MQ]
520 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
524 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
525 type parsing. "short" and "long" modifiers were also ignored; other
526 modifier (such as reference level) are still ignored. [MQ]
527 * Embeed the buffer size within the buffer itself on SG. [MQ]
528 That way, send() are atomic and cannot get intermixed anymore (at least
529 the ones which are less than 100k; bigger messages still have the issue)
530 * Array size pushed by the field, not by the field type (or each
531 and every long int will push stuff to the cbps) [MQ]
532 * use select() to sleep since it allows to portably sleep less than one
535 GRAS (minor cleanups)
536 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
539 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
540 * type category 'ignored' killed as it was never used and were difficult
542 * whether a type can cycle or not is now a flag, leaving room for more
543 flags (as "ignored", if we feel the need one day ;)
544 * Rename raw sockets to measurement sockets since "raw" has another
545 meaning in networking community.
548 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
549 and offers features not belonging to GRAS but that most applications
550 need. One day, it may be a set of plugins loadable at runtime.
551 * New module: bandwidth
552 bandwidth measurement between arbitrary nodes running this module. [MQ]
554 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
556 SimGrid (2.94) unstable; urgency=low
558 The first beta release of SimGrid 3 !
562 * Update the main page and the FAQ. Adding references to gforge.
565 * Add a gras_os_getpid function.
568 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
569 * Extend the logs so that they also print PID, hostname, date, ... if
571 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
572 and kill the old version which were in testsuite/
573 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
577 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
578 for users. Instead, each programming environment (GRAS, MSG,...) use it
579 when needed to provide such a feature to users.
580 Don't shortcut the mecanism or you will also shortcut the virtualization
581 you need on the simulator.
585 * Cleanups in configury with regard to compile optimization/warning flags.
586 Also add -fno-loop-optimize to any powerpc since it's the optimization
587 killing gcc (< 3.4.0).
588 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
589 needed by MSG examples complications
590 * Borrow configury beautifications from PHP
593 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
594 warning (thanks loris for stumbling into this one).
595 * Bugfix: stop loading private headers (gras_config.h) from the public
599 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
600 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
601 RL would imply a DNS resolver, which is *hard* to do in a portable way
602 (and therefore delayed).
603 * Implement a real timer mecanism and use it in timing macros. This allows
604 to avoid rounding errors and get a 0.000005 sec precision in timing
605 macros. While I was at it, various cleanups:
606 - allow to declare more than one timed section per file (fix a stupid bug)
607 - move some private declaration to the right place
608 - merge conditional execution and timing macros into emulation module
609 - document the module
610 - make sure the module cleanups its mess on gras_exit
611 * Documentation improvements:
612 - (new) how to compile applications using GRAS
613 - (new) emulation support (timing macros)
615 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
617 SimGrid (2.93) unstable; urgency=low
619 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
622 - Use Paje properly where used. Still to be sanitized properly.
623 - Portability fix: Add an implementation of the contexts using pthread
627 - Add xbt_procname(): returns the name of the current process.
628 Use it to show the current process's name in all logging.
630 - fix detection of older flex version and the reaction, since we do
631 depend on modern ones (we use lex_destroy)
632 - Better separation of SG and RL in the libs: remove all simulation code
633 from libgras. As a result, this lib is now only 200k when stripped.
634 Some of the xbt modules may also be duplicated (two sets and such) and
635 should be cleaned/killed before SG3.
636 - Insist on using xlC on AIX because of weird problems involving gcc there.
637 - Cleanup the make remote stuff. This is now done by scripts
638 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
639 mainly for our private use, but we're working on changing them to user
642 - Bugfix: flush the socket on close only if there is some *output*.
643 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
645 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
648 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
650 SimGrid (2.92) unstable; urgency=low
652 Alpha 3 on the path to SimGrid 3
656 - New! First try of benchmarking macros.
657 - New! First try so that gras_stub_generator generate deployment and
658 remote compilation helpers.
660 - Bugfix: Initialization fix in msg_test.
664 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
666 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
668 - API Change: gras_os_sleep to take the amount of seconds as a double.
669 Accepting an int was error prone since it was the only location where
670 seconds were coded as such. It leaded to damn rounding errors.
671 - Bugfix: Hard to belive that timers ever worked before this.
673 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
675 SimGrid (2.91) unstable; urgency=low
677 Alpha 2 on the path to SimGrid 3
681 - Bug fix in the lmm_solver.
683 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
684 through the function MSG_paje_output.
685 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
686 - It is possible to bound the rate of a communication in MSG with
687 MSG_task_put_bounded() (was already in the previous version but I had forgotten
688 to write it in the changelog).
689 - Bug fix to let GRAS run on top of MSG until we move it directly on top
694 - Various cleanups to the autotools stuff
695 - Begin to move Gras examples to examples/gras/
696 - Let make distcheck work again (yeah!)
698 - documentation overhauled using doxygen.
699 gtk-doc-tools is dead in SimGrid now.
700 - Automatically extract all existing logging categories, and add the list
701 to the documentation (long standing one, to say the less)
703 - Cleanup the known architecture table. Reorder the entries to group what
704 should be, and use a more consistent naming scheme.
705 (some of the test dataset are still to be regenerated)
706 - New! Allow library to register globals on each process just as userdata
708 This is implemented using a xbt_dict and not a xbt_set, so we loose the
709 lookup time (for now).
710 Use it in msg and trp.
711 This cleans a lot the internals and helps enforcing privacy of the
712 headers between the gras components.
713 - New! Add a timer mechanism, not unlike cron(8) and at(1).
714 - Bugfix: gras_os_time was delirious in RL.
715 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
717 - Reenable GRAS now that it works.
719 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
721 SimGrid (2.90) unstable; urgency=low
723 Alpha 1 on the path to SimGrid 3
725 * It is a long time since the last release of SimGrid. I'm sorry about
726 that but as I had told you, I was rewriting a lot of things. I apologize
727 to those who had been reporting bugs to me and that I had not answered.
728 If your bug is still in the new version, please tell me. Here is a
729 summary of the main changes.
731 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
732 lead by Martin Quinson. As a consequence SimGrid gains a lot in
733 portability, speed, and a lot more but you'll figure it out later.
734 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
735 I wanted to release the new MSG as soon as possible and I have
736 broken GRAS, which is the reason why, for now, only MSG is fully
737 functional. A laconic description of these projects is available
738 in the documentation.
740 * REVOLUTION 2: I have removed SG and I am now using a new simulation
741 kernel optimized for our needs (called SURF but only the developers
742 should use it). Hence, MSG is now roughly 30 times faster and I think
743 that by rewriting a little bit MSG, I could event speed it up a little
744 bit more. Beside the gain in speed, it is also much easier to encode a
745 new platform model with SURF than it was with SG. More to come...
747 * REVOLUTION 3: I have tried to change a little as possible the API of
748 MSG but a few things really had to disappear. The main differences
749 with the previous version are :
750 1) no more m_links_t and the corresponding functions. Platforms are
751 directly read from a XML description and cannot be hard-coded
752 anymore. The same format is used for application deployment
753 description. The new format is described in the documentation.
754 Have a look in tools/platform_generation. There is a tiny script
755 that converts from the old platform format to the new one. Concerning
756 the application deployment format, parsing the old one is tricky.
757 I think most of you should however be able to convert your files. If
758 it is really an issue, I can write a C code that does the conversion.
760 2) the toolbox tbx does not exist anymore. We now have a library
761 with much more data-structures but without the hash-tables (we have
762 dictionaries that are much faster).
764 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
766 *****************************************************************************
767 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
768 * don't feel like converting the oldies. (Mt) *
769 *****************************************************************************
772 Version 2.90: "the long awaited one"
773 - Finished rewriting and debugging MSG. Rewrote the documentation.
774 - disable GRAS for now since it needs to be ported to the newest SG
777 - Finish the port to windows (using mingw32 for cross-compile)
780 - Main loop and datastructures of SURF. A cpu resource object is
781 functional. Surf can thus be used to create cpu's with variable
782 performance on which you can execute some actions.
784 2004-11-15 Martin Quinson
785 - Port to ARM. Simply added the alignment and size descriptions. Should
786 work, but the ARM machines are so slow that I didn't had the opportunity
787 to 'make check' over there yet.
789 2004-11-15 Arnaud Legrand
790 - Trace manager now written. It uses a heap structure and is therefore
791 expected to be efficient. It may however be speeded up (particularly
792 when many events occur at the same date) by using red and black
793 trees. One day maybe...
794 - Max-min linear system solver written. It uses a sparse matrix
795 structure taking advantage of its expected use. Most operations are
796 O(1) and free/calloc are called as few as possible. The computation of
797 the minimum could however be improved by using a red and black tree
800 2004-11-03 Arnaud Legrand
801 - Rename every gras_* function that was in xbt/ to its xbt_
803 - Add a heap and a doubly-linked list to xbt
804 - Added a dichotomy to the dictionaries. make check works as well before
805 so I assume that the patch is correct. I do not know however if things
806 run effectively faster than before now. :)
808 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
809 SimGrid, and the version number is bumped to 2.x
811 2004-10-29 Martin Quinson
812 - Introduction of the remote errors.
813 They are the result of a RMI/RPC on the remote machine.
814 ErrCodes being scalar values, you can't get the host on which those
815 errors did happen. Extending the error mechanism as in Gnome is possible.
816 No idea yet whether it is a good idea.
818 2004-10-28 Martin Quinson
819 - Interface revolution: the Starred Structure Eradication.
820 I used to do typedef struct {} toto_t; and then handle *toto_t.
821 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
822 - ???_t is a valid type (builded with typedef)
823 - s_toto_t is a structure (access to fields with .)
824 - s_toto is a structure needing 'struct' keyword to be used
825 - e_toto_t is an enum
826 - toto_t is an 'object' (struct*)
828 typedef struct s_toto {} s_toto_t, *toto_t;
829 typedef enum {} e_toto_t;
830 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
831 s_toto_t) is private.
833 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
834 so that it can set the variable to NULL. It was so for dicts and sets,
835 it changed for dynars.
837 - Fix a bunch of memleaks in dict_remove
838 - Fix a bug in sg/server_socket opening: it failed all the time.
840 2004-10-07 Martin Quinson
841 - Speed up dynar lookup operation a bit.
843 gras_dynar_get is dead.
845 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
846 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
847 (which gives you the address of the stored data).
849 gras_dynar_get_as is an helpful macro which allows you to retrieve a
850 copy of the data using an affectation to do the job and not a memcpy.
852 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
853 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
855 It does not really speedup the dynar test because they are
856 setting elements all the time (and look them seldom). But the dict does
857 far more lookup than setting.
859 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
861 2004-10-05 Martin Quinson
862 - Allow to (en/dis)able the cycle detection at run time.
864 Whether we should check for cycle or not is now a property of each
865 datatype. When you think there may be some cycle, use datadesc_cycle_set.
866 datadesc_cycle_unset allow to remove this property when previously set.
868 Note that the cycle detection is off by default since it impacts the
869 performance. Watch the data you feed GRAS with ;)
871 This property is hereditary. Any element embedded in a structure having it
872 set have it set for the time of this data exchange.
874 You should set it both on sender and receiver side. If you don't set it on
875 sender side, it will enter an endless loop. If you forget on receiver
876 side, the cycles won't be recreated after communication.
878 - Header reorganization.
879 Kill gras_private.h, each submodule must load the headers it needs.
881 2004-10-04 Martin Quinson
882 - Interface revolution: do not try to survive to malloc failure.
884 Now, gras_malloc and friends call gras_abort() on failure.
885 As a conclusion, malloc_error is not a valid error anymore, and all
886 functions for which it was the only gras_error_t return value are
887 changed. They now return void, or there result directly.
888 This simplify the API a lot.
890 2004-09-29 Martin Quinson
891 - Re-enable raw sockets.
892 Created by gras_socket_{client,server}_ext;
893 Used with gras_raw_{send,recv}
896 It should allow to kill the last bits of gras first version soon.
898 This is not completely satisfactory yet (duplicate code with
899 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
902 - Simplify transport plugin (internal) interface by not passing any
903 argument to _server and _client, but embedding them in the socket
906 2004-09-28 Martin Quinson
907 - Finish the port to AIX.
908 autoconf was my problem (segfault within the malloc replacement
909 function. No idea why)
911 2004-09-16 Martin Quinson
912 - Fix some size_t madness on 64bit architectures.
914 2004-09-08 Martin Quinson
915 - Reduce the number of system headers loaded, overload some more system
916 calls (such as malloc to cast the result of the system one, and work
918 - Fix and reintroduce the config support
920 2004-09-07 Martin Quinson
921 - Source code reorganization to allow Arnaud to surf all over there.
922 - Allow to document the logging categories.
923 - Remove all uppercase from logging categories and useless cleanup in names.
925 2004-08-18 Martin Quinson
926 Version 0.6.2 (protocol not changed; API changed)
927 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
928 gras_error_t), and NULL when not found. Functions expecting a msgtype
929 as argument (msg_wait; msg_send) deal with NULL argument by providing a
930 hopefully usefull message.
931 - Portability to prehistoric sparcs again
933 2004-08-17 Martin Quinson
934 Version 0.6.1 (protocol not changed; ABI not changed)
935 - prealloc some buffers to speed things up
937 2004-08-11 Martin Quinson
938 Version 0.6 (protocol not changed; ABI expended)
939 - The parsing macro can deal with the references, provided that you add
940 the relevant annotations (using GRAS_ANNOTE(size,field_name))
942 2004-08-09 Martin Quinson
943 Version 0.5 (protocol not changed; ABI changed)
944 - Allow to off turn the cycle detection code in data exchange at
945 compilation time. It should be at run time, but I'm short of time (and
946 the config stuff is still broken). That way, we keep dict out of the
947 critical path, which is good because the performance is poor:
948 - search not dichotomial yet
949 - dynar give no way to access their content and memcpy everytime
950 - In composed data description (struct, ref and so on), stop foolness of
951 keeping the subtype's ID, but store the type itself. This keeps sets out
952 of the critical path, which is good since they rely on dynar and
953 dictionnaries. The only loose of that is that we cannot detect the
954 redeclaration of a structure/union with another content (but I'm not sure
955 the code detected well this error before anyway). We still can detect
956 the redefinition discrepancy for the other types.
957 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
958 it breaks the code because of type-punning used all over the place).
959 This breaks on all non-gcc architectures (for now).
961 All those changes (plus the buffer of last time) allow me to gain 2 order
962 of magnitude on cruel tests consisting of 800000 array of integers on two
963 level of a hierarchical structure (200 secondes -> 4 secondes)
966 - the selector of reference must now return the type it points to, not
969 2004-08-06 Martin Quinson
970 Version 0.4 (protocol changed; ABI not changed)
971 - Allow to pass --gras-log argument to processes in simulation mode. Really.
972 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
973 - Implement a buffer transport, and use it by default (it relies on tcp in
974 real life and on sg in simulation).
975 That's a bit hackish since I had a new field to the structure to store
976 its data without interfering with the subtype ones. Inheritance
977 is tricky in C. And that's a kind of reverse inheritance with one class
978 derivating two classes. Or maybe a game with java interfaces. Anyway,
979 that's damn hard in C (at least).
980 Moreover, I got tired while trying to ensure plugin separation and
981 genericity in SG mode. MSG wants me to do weird things, so let's go for
982 cruel hacks (temporarily of course ;).
983 See comment in transport_private.h:71
984 - do not include all the _interface headers in private but in the files
985 which really need them (to cut the compilation time when they are
988 2004-07-26 Martin Quinson
989 Version 0.3 (protocol not changed; ABI changed)
990 - Major overhault of the datadesc interface to simplify it:
991 - shorted the function names:
992 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
993 - add a trivial way to push/pop integers into the cbps without malloc.
994 This allows to make really generic sub_type description, which simply
995 pop their size of the stack.
996 - add a function gras_datadesc_ref_pop_arr() which does what users want
997 most of the time: Declare a dynamic array (which pops its size of the
998 stack) and declare a reference to it. Poor name, but anyway.
999 - kill the post-send callback, add a post-receive one
1001 2004-07-23 Martin Quinson
1002 Version 0.2 (protocol changed; ABI changed)
1003 - add some testing for cpbs in the test cases, and fix some more bugs.
1004 This invalidate again the little64 data file, since I cannot regenerate
1006 - remove an awfull optimization in the logging stuff, allowing me to:
1007 - understand it again
1008 - learn gcc how to check that the argument match the provided format
1009 - fix all errors revealed by gcc after that
1010 - internal keys of dict are not \0 terminated. Deal with it properly in
1011 loggings instead of segfaulting when the user want to see the logs :-/
1013 2004-07-22 Martin Quinson
1014 - Fix some stupid bug preventing cbps (callback postit) from working
1016 2004-07-21 Martin Quinson
1017 - Some documentation cleanups
1018 - remove the useless last argument of msgtype_declare
1019 - rename the Virtu functions to fit into the 'os' namespace
1020 - move headers src/include -> src/include/gras/ and stop fooling with
1021 gras -> . symbolic link
1022 - make distcheck is now successful
1024 2004-07-19 Martin Quinson
1026 - Build shared library also
1027 - Install html doc to the right location
1028 - stop removing maintainer files in make clean
1029 - build tests only on make check
1031 2004-07-13 Martin Quinson
1033 - No major issue in previous version => change versionning schema
1034 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1035 corresponding dataset.
1037 2004-07-11 Martin Quinson
1039 - Get it working with any kind of structure (we can compute the padding
1040 bytes remotely for all the architectures I have access to)
1041 - Implement the structure parsing macro (still not quite robust/complete)
1042 - Improvement to the remote testing toysuite
1044 2004-07-10 Martin Quinson
1045 [autoconf mechanism]
1046 - get ride of a bunch of deprecated macros
1047 - actually run the test for two-compliment, not only compile it :-/
1048 - test whether the structures get packed (and bail out if yes. Damn.
1049 Alignment is a serious matter)
1050 - test whether the structures get compacted (but respecting the alignment
1051 constraints of each types)
1052 - test whether the array fields of structures can straddle alignment boundaries
1054 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1055 (took me 2 hours to find that bug, looking at the wrong place)
1056 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1057 before sending/receiving and are used to compute the offsets of fields
1058 - Given that padding size depend even on compiler options, keep track of
1059 alignment and aligned_size only for the current architecture. Not a big
1060 deal since we send structure fields one after the other (seems
1062 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1063 program, let it work on linux/gcc/little32. portability todo.
1065 2004-07-08 Martin Quinson
1066 - import and improve remote compilation support from FAST
1067 - make sure make check works on half a dozen of machines out there
1069 2004-07-07 Martin Quinson
1070 Let's say it's version 0.0.3 ;)
1071 - Implement conversions (yuhu!)
1072 - Let it work on solaris (beside conversion, of course)
1073 - Stupid me, using rand() to generate the conversion datatests in not wise.
1075 2004-07-06 Martin Quinson
1076 - Let make dist work, since I'm gonna need it to compile on remote hosts
1077 - Let Tests/datadesc_usage write the architecture on which the file was
1078 generated as first byte.
1079 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1080 architecture descriptions.
1081 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1082 on those architectures.
1083 - Optimization: send/recv array of scalar in one shoot
1085 2004-07-05 Martin Quinson
1086 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1088 - Plug a whole bunch of memleaks
1089 - each process now have to call gras_{init,exit}. One day, their log
1090 settings will be separated
1091 - Continue the code factorisation between SG, RL and common in Transport.
1093 2004-07-04 Martin Quinson
1095 - Redistribution between SG and RL.
1096 We wanna have to accept in SG, so move accepted related parts of RL in
1097 the common part. (more precisely, the dynar of all known sockets is no
1098 more a static in transport.c, but part of the process_data)
1100 [gras_stub_generator]
1101 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1103 2004-07-03 Martin Quinson
1104 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1105 them all up in one shot)
1107 - Fix the ugly case of reference to dynamic array.
1108 - New (semi-public) function gras_datadesc_size to allow the messaging
1109 layer to malloc the needed space for the buffer.
1111 - gras_socket_close now expect the socket to close (and not its address to
1112 put NULL in it after it). This is because the socket passed to handlers
1113 is one of their argument (=> not writable).
1115 - propagate the interface cleanup from last week in datadesc, ie remove a
1116 superfluous level of indirection. User pass adress of variable
1117 containing data (both when sending and receiving), and not of a variable
1118 being a pointer to the data. Let's say that I like it better ;)
1119 The price for that is constructs like "int msg=*(int*)payload" in
1120 handlers, but it's a fine price, IMHO.
1122 - Let it work in RL (yuhu)
1124 2004-06-21 Martin Quinson
1126 - porting SG plugin and SG select to new standards (works almost).
1127 - plug memleaks and fix bugs around.
1130 - cleanup the prototype of data recv and force users to specify when they
1131 want to handle references to objects. Test case working even for cycles.
1132 - plug memleaks. Valgrind is perfectly ok with this.
1134 2004-06-12 Martin Quinson
1136 - cleanup the separation between plugin and main code in plugin creation
1138 2004-06-11 Martin Quinson
1140 - Reput hook for raw sockets, needed for BW experiments
1141 - kill a few lines of dead code
1142 [Data description] Interface cleanup
1143 - gras_datadesc_by_name returns the searched type or NULL.
1144 That way, no variable is needed to use a type desc once, which makes
1146 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1147 two parameters were strings (field name, type name), leading to
1149 [Dicos] Interface cleanup
1150 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1151 This is consistant with the dynar API.
1153 2004-04-21 Martin Quinson
1155 - Porting to new standards.
1157 - interface cleanup.
1158 There is no bag anymore, no need to take extra provision to mask the
1159 pointers behind "ID".
1160 Better splitup of functions between files create/exchange/convert.
1161 This is still a bit artificial since convert and receive are so
1162 interleaved, but anyway.
1164 - add a queued message list to procdata (the ones not matching criteria
1166 - factorize some more code between SG and RL wrt procdata
1168 - use gras_exit in example to track memleaks
1169 - get rid of gs_example now that GS is properly integrated into gras
1170 - update run_test to integrate the lastest tests (datadesc)
1172 - rename WARNINGn macros to WARNn since it prooved error-prone
1174 2004-04-19 Martin Quinson
1176 - register init/exit functions within gras module mechanism
1177 - send/receive function.
1178 Convertion is not implemented, but short-cutted if not needed.
1179 struct/array elements are sent one by one (instead of block-wise), but
1180 nobody really cares (yet). Get a prototype before optimizing.
1181 - tests (using a file socket) for DD send/receive on:
1182 - base types: int, float
1183 - array: fixed size, string (ie ref to dynamic string)
1184 - structure: homogeneous, heterogeneous
1185 - chained list, graph with cycle
1186 Believe it or not, valgrind is not too unhappy with the results. The
1187 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1191 - Bugfix when using a filename explicitely (instead of '-')
1193 2004-04-09 Martin Quinson
1195 - factorize more code between RL and SG in socket creation
1196 - Complete the implementation and tests of:
1198 o file (only in RL, and mainly for debugging)
1200 I lost 3 days to design a portable address resolver, and then decided
1201 that the prototype mainly have to run on my box.
1202 Addressing portability too early may be like optimizing too early :-/
1204 - use gras_init in the Tests instead of the crappy parse_log_opt
1205 (the latter function is removed)
1206 [Conditional execution]
1207 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1208 [Code reorganisation]
1209 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1210 Build examples against the RL library, since there is no way to disable
1211 its creation for now.
1213 For information, the beginning of coding on GRAS was back in june
1214 2003. I guess that every line has been rewritten at least twice since