1 SimGrid (3.4-svn) unstable; urgency=high
4 * Also include strbuff from xbt.h public header
7 * fix a bug on struct sizeof computation, which prevented the
8 exchange of arrays of structs in some conditions
11 * Allow to control the simulation from a trace file.
12 New functions MSG_action_register() and MSG_action_trace_run()
13 The first one allows to associate a function execution to each
14 kind of action while the second one parses a trace file and
15 triggers the corresponding actions within the system.
16 For now, only a toy example is provided in examples/msg/actions
17 * Add an exemple of process migration in examples/msg/migration
18 * Fix a bug in task exchange which broke MSG_task_get_sender()
19 Add a teshsuite regression test for that.
20 [Bug: if MSG_task_get_sender() is called after sender exit,
24 * Add SIMIX_process_set_name() to change the name of the current
25 process in the log messages.
28 * xbt_ex_display(): do not free the exception after displaying
29 This allows to do more with the given exception afterward.
30 Users should call xbt_ex_free() themselves.
32 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
34 SimGrid (3.3) stable; urgency=high
38 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
41 * Introduce the SIMIX module: factorize code between MSG and GRAS.
44 Until now, GRAS were using MSG as an interface to SURF. It was
45 quite difficult because both interface have several differences
46 (MSG channels vs GRAS sockets were the most notable point).
48 This also opens the gate to SMPI (which should occur soon) and speed
49 up simulations by to 40% (even if it were not the main goal).
51 **************************************
52 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
53 **************************************
54 The point is that events occuring at the exact same timestamp are
55 not scheduled in the same order with the old and new version. This
56 may be enough to completely change the execution of simulations in
57 some cases. Sorry for the inconvenience.
59 * Cleanup and upgrade the XML format to push further scalability
60 issues (check http://hal.inria.fr/inria-00256883/ for more info)
62 * Improve the testing infrastructure with tesh. Now a very large part of
63 the code is tested not only by being run but also by checking that the
64 output match an expected output [Mt].
66 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
67 is really less memory-demanding, which should allow you to use
68 larger files in SimGrid [AL].
70 * Inform valgrind about our contextes, so that it becomes usable
71 with the default (and more effecient) version of SimGrid
72 [contributed by Sékou Diakite, many thanks]
75 * Introduce a listener thread in charge of receiving incomming
76 messages from the network. It allows to overlap communication and
77 computation but most notably, it removes some stupid deadlocks due
78 to the fact that so far, a process could not send and receive at
79 the same time. This made most non trivial communication schema
81 * Convert the PIDs from long int to int to match the MSG ones (and
83 * New function: gras_agent_spawn() to launch a new process on
84 current host. Only working in simulation for now. [Mt]
85 * New function: gras_os_hostport() returning a constant form (ie,
86 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
89 * Make the backtrace of exceptions more human readable [Mt]
90 * New module: xbt/str [Mt]
91 a ton of string utility functions (split, join, printf to a newly
92 allocated buffer, trim, etc)
93 * New module: xbt/hash [Mt]
94 SHA1 hashing algorithm (more to come if needed)
95 * New module: xbt/synchro [Mt]
96 synchronization tools (mutex and conditions) working the same way
97 in simulation and in real life (mainly useful for GRAS, but not
99 * New module: xbt/queue [Mt]
100 classical producer/consumer synchronization scheme
101 * xbt_dynar_new_sync() creates a synchronized dynar. All access
102 (using the classical functions will get serialized) [Mt]
103 * Make dictionary internal table dynamic. No need to specify its size
104 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
106 * Make sure the log channels are organized as a tree under windows
107 (because of ANSI C compatibility issue, any channel were child of
111 * Cleaned many thing in surf and fixed a few bugs [AL].
112 * Add a nice command line configuration mechanism to compose models [AL].
113 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
114 the previous one (KCCFLN05). It relies on something that looks like
115 a max-min sharing mechanism but cannot be written as such. A new solver
116 was thus designed [AL].
117 * Add a new solver to lmm. Based on Lagrange optimization and
118 gradient-based descent, it enables to efficiently maximise systems s.a
120 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
122 This solver enables to propose two new network models for TCP Reno and
123 TCP Vegas based on Low's work. These models still need to be fully
124 tested though [Pedro Velho].
127 * Bug fix in SD_simulate. Now the time bound given as argument is
129 * Use the new parallel task model (ptask_L07) as default.
130 * Use the SURF command line configuration mechanism.
131 * 0-size tasks (for synchronization) should now work.
133 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
135 SimGrid (3.2) stable; urgency=high
139 We still experience issues on this platform, but we believe that at
142 GRAS API BREAKAGE (for simplification purpose, sorry):
143 * the gras_msgtype_by_name is not used anymore. Instead of
144 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
145 you can write (and must)
146 gras_msg_send(toserver, "request", &request);
147 - If you still want to pass a gras_msgtype_t to the function (to cache
148 the type and avoid the lookup time), use the gras_msg_send_() variant.
149 - Impacted functions:
150 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
151 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
152 * The callbacks are now expected to return 0 when everything went well
153 (just like the main() function)
155 GRAS new features and improvements:
156 * New module mecanism where user code can use per process globals [Mt]
157 This is similar to gras_userdata_*() functions, but for libraries. It
158 factorize some code developped over and over in the examples and AMOK.
159 It has still to be documented and used (only amok/peermanagement is
161 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
162 loss when they were small enough to fit into the buffer and sent quickly
163 enough so that they can all get received in one shoot.
164 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
165 if not found. Use the *_or_null() variant for the old semantic.
166 * In gras_msg_handle, do not discard messages without callback.
167 They are probably messages to be explicitly awaited later (ie, proofs of
168 mis-synchronization in userland since they are sent before being awaited)
170 * gras_socket_meas_send/recv: semantic changed!
171 The numerical arguments used to be (1) the total amount of data to send
172 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
173 messages. This was need for the fool willing to send more than MAXINT
174 bytes on quite fat pipes.
177 * Do really rename the hostmanagement module to peermanagement. [Mt]
178 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
179 but this is rather new and this was documented in the module
180 documentation (poor excuses, I admit)
181 * Bandwidth measurement semantic changed! This follows the changes to
182 gras_socket_meas_send/recv explained above.
185 * A sequential mode has been added to the workstations. When a workstation
186 is in sequential mode, it can execute only one task, and the other tasks
187 are waiting in a FIFO. [Christophe Thiery]
190 * The KCCFLN05 workstation model now handles parallel tasks. It is the
191 model for SIMDAG. [Christophe Thiery]
192 * Bug fix in the maxmin solver: Some values were close to 0 instead of
193 equal to 0, which caused some bad behaviors in
194 saturated_constraint_set_update. I now use a threshold mechanism like in
198 * When running manually src/testall, you select specific units [Mt]
199 testall is the result of our cunit mecanism, and should replace all
200 the scripty thingy around since bash don't run easily on billware.
202 * A mallocator system has been added. [Christophe Thiery]
203 Mallocators allow you to recycle your unused objects instead of freeing them
204 and allocating new ones.
206 Documentation update:
207 * FAQ reworking + New FAQs:
208 - "Valgrind spits tons of errors!" [Mt]
209 - "How to repport bugs" [Mt]
210 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
211 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
212 - Communication time measurement within MSG [AL]
213 - I experience weird communication times when I change the latency [AL]
216 - an introduction to the framework and to the used communication model
217 - an initiatic tour introducing the most proheminent features:
219 . Lesson 0: Installing GRAS
220 . Lesson 1: Setting up your own project
221 o Part 2: Message passing
222 . Lesson 2: Exchanging simple messages
223 . Lesson 3: Passing arguments to the processes (in SG)
224 . Lesson 4: Attaching callbacks to messages
225 . Lesson 5: Using globals in processes
226 . Lesson 6: Logging informations properly
227 . Lesson 7: Using internal timers
228 . Lesson 8: Handling errors through exceptions
229 . Lesson 9: Exchanging simple data
230 . Lesson 10: Remote Procedure Calling (RPC)
231 . Lesson 11: Explicitely waiting for messages
232 . Recapping of message passing features in GRAS
233 - A HOWTO section containing:
234 o HOWTO design a GRAS application
235 More are due, of course. They will come latter. In the meanwhile, you can
236 check the examples which are still here.
238 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
240 SimGrid (3.1) stable; urgency=high
244 There was a stack corruption somewhere, visible only when optimizing
245 with these versions. [Vince]
248 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
249 module that enabled to play with Directed Acyclic Graphs. It is built
250 directly on top of SURF and provides an API rather close to the old
251 SG. Some old codes using SG are currently under rewrite to check that
252 all needful functions are provided. [Christophe Thiery]
255 * Complete rewrite of the KCCFLN05 workstation model. It is now an
256 extension of the classical CLM03 model that gracefully handles
257 failures. This is now the default model for MSG and GRAS. It doesn't
258 handle parallel tasks yet though. [AL]
259 * Bug fix: Weights were not correctly set in the network part.
260 WARNING: This may have resulted in incorrect results with simulations
261 where there are more than one flow on a given link. [AL]
264 * After a (long ?) discussion on simgrid-devel, we have decided that the
265 convention we had on units was stupid. That is why it has been decided
266 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
267 WARNING : This means that all previous platform files will not work as
268 such with this version! A warning is issued to ask users to update
270 A conversion script can be found in the contrib module of the CVS, under
271 the name contrib/platform_generation/surfxml_update.pl [MQ]
274 * Bug fix: Processes were started in reverse order, wrt deployment file.
275 WARNING: if your code relies on this bug, please fix it. [AL]
276 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
277 being executed on two different locations. [AL]
278 * Bug fix: Failures are now better supported thanks to Derrick's tests
279 (there was many failure situations I hadn't thought of and that weren't
280 correctly handled). [AL]
281 * New function: MSG_host_is_avail indicates you whether a given m_host_t
285 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
286 Exception occurring on server-side are propagated back to client (!).
288 API CHANGE: the callback changed their prototype. Change:
289 int my_handler(gras_socket_t expeditor, void *payload_data) {
291 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
292 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
294 * New! function: gras_msg_handleall to deal with all messages arriving
295 within a given period.
296 * New! function: gras_socket_server_range to get a server socket in a
297 range of port numbers (ease to avoid port number conflicts) [MQ]
298 * New! gras processes display their backtrace when they get a SIGUSR1
299 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
300 Sweet to debug RL processes [MQ]
304 - Do not force experiment sizes to be expressed in kb, or it becomes
305 impossible to measure the latency this way (needs one byte-long tests)
306 WARNING: this changes the amok_bw_* function semantic. [MQ]
307 - Implements the link saturation stuff. [MQ]
308 * Peer management module:
309 New! module factorizing code that we wrote over and over [MQ].
312 * New module: cunit (my jUnit implementation in ansi C) [MQ]
313 - Test units are placed directly into the library code, they get extracted
314 automatically and placed into the src/testall binary.
315 - Convert most of the XBT tests to this system.
316 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
317 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
318 live in the same C code. This required to change a little bit the API
319 of surfxml parsing but shouldn't be an issue for end-users. [AL]
320 * New module: sparse graph structure with basic algorithms (this is work
321 in progress and the API is not considered to be frozen yet). [AL]
322 * Display more information on backtraces: source line & function names are
323 now displayed just like valgrind does (rely on addr2line tool) [MQ]
324 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
325 * Reworked a little bit some #include statements to load only required
326 headers. Some user code that relied on SimGrid to include stdlib or
327 stdio may need to include it by themselves. [AL]
328 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
330 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
331 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
332 but I guess that almost nobody used those parts. [MQ]
334 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
336 SimGrid (3.0.1) stable; urgency=low
339 * Unfortunately, I had missed 5 misnamed functions:
340 xbt_fifo_item_t xbt_fifo_newitem(void);
341 void xbt_fifo_freeitem(xbt_fifo_item_t);
342 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
343 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
344 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
345 They're now deprecated. Please use their new versions:
346 xbt_fifo_item_t xbt_fifo_new_item(void);
347 void xbt_fifo_free_item(xbt_fifo_item_t);
348 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
349 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
350 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
352 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
353 * Documentation: xbt_log module unmercifully reworked [MQ]
354 * Bugfix: there was a problem with the ending of contexts with
355 the pthread backend. It caused some weird deadlock or behavior
356 depending on the pthread implementation. [AL]
357 * Bugfix: get the exceptions raised in the simulator repport where
358 and why they come from when they are not catched in time [AL, MQ]
361 * Bugfix: Do repport the error when two non-connected hosts try to
362 exchange data (Thanks to Flavien for stumbling into this one) [AL]
365 * Add additionnal checkings on communications. Assert that two
366 communicating hosts are connected by a set of links... [AL]
369 * Add additionnal checkings on channel values in communication [AL]
370 * New: MSG_task_get_source to see on which host a task was generated [HC]
371 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
372 the number of tasks waiting to be received on channel and sent
374 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
375 waits for the first task coming from a given host.. [AL]
377 GRAS new functionnalities: [MQ]
378 * Enhance the parsing macro to allow the size of multidimentional objects
379 to be given thru annotations.
380 * New example (and documentation): Matrix Multiplication a la RPC
381 (as when I was young!) and fix a bunch of bugs found on the way.
383 GRAS performance improvements: [MQ]
385 * Reduce the amount of cbps creation/destruction by making it static to
386 datadesc_send/recv() and using a (newly created) cbps_reset (based on
389 * Change libdata to a set so that we can search for stuff by ID (and thus
390 reduce the insane amount of dict lookups)
393 * Actually implement gras_datadesc_copy() so that we don't have to mimick
394 RL communication on top of SG since it's so uneffective.
395 It may also be used for inter-thread communication in RL, one day.
396 * Use gras_datadesc_copy() to exchange messages on top of SG
398 - improve message exchange performance on top of SG
399 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
400 * Don't exchange on the network the size of the used part of buffer,
401 instead, specify the possible buffer size to read().
403 - reduces the amount of read/write calls (one pair per exchange)
404 - reduces the amount of exchanged data (the size)
405 - allows to retrieve all arrived data on receiver side, if we don't need
406 it right now (subsequent read will peek the buffer)
407 - allows the receiver to proceed with the begining of the stream before
408 everything is arrived
409 - make it possible to build an iov transport (using readv/writev)
411 - take care of the data with non-stable storage (like stacked data),
413 * If possible, TCP send uses vector I/O (when writev() is here)
414 - Don't use it for receive since we send data sizes and data on the
415 same stream, so we wouldn't be able to chain large amount of chunks
416 before having to flush the stuff to read the size.
417 * Rework the transport plugin mecanism to simplify it and reduce the
418 amount of pointer dereferencement when searching for the right function
421 * I guess that now, we do almost as few system calls as possible while
422 doing as few data copy as possible.
424 To improve it further, we could try to send all the sizes first and then
425 all the data (to use iov on receiving size), but it's only a partial
426 solution: when you have 2 dimensional data, the sizes of the second
427 dimension is data of the first dimension, so you need 3 streams.
429 I'm not sure the potential performance gains justify the coding burden.
431 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
433 SimGrid (3.00) stable; urgency=high
436 * New! Give the possibility to hijack the surf parser and thus bypass
437 MSG_create_environment and MSG_launch_application. Have a look at
438 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
440 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
442 SimGrid (2.96) unstable; urgency=low
447 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
448 This deprecates the xbt_error_t mecanisms.
449 It modifies (simplifies) all XBT and GRAS API.
450 MSG API keeps unchanged (exceptions raised by XBT are catched from
451 within MSG and masked with existing error handling facilities)
454 * New! Add a FATPIPE model. [AL]
455 * New! Add a parallel task model. [AL]
456 * New! Add automatically a loopback interface (in the default
457 network model) if none was precised.
460 * Bugfix: MSG_process_resume now works with the current running process.
462 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
463 * Modification of MSG_task_get_compute_duration. Once a task has been
464 processed, the value returned by this function is now equal to 0. [AL]
465 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
466 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
467 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
468 of a m_task. For now, MSG_task_cancel only works with computation
470 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
471 of the processor (in Mflop/s) regardless of the current load on the
473 * API Change: use proper naming convention for MSG_getClock and
474 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
476 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
477 This function changes the priority of a computation task. This priority
478 doesn't affect the transfer rate. A priority of 2 will make a task
479 receive two times more cpu power than the other ones. This function
480 has been added to suit the needs of Nguyen The Loc and hasn't been that
481 much tested yet. So if it fails, please report it and send me your code.
483 * API Change: removed all functions and types that were marked "deprecated"
484 since many months. Renamed MSG_global_init_args to MSG_global_init.
486 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
488 SimGrid (2.95) unstable; urgency=low
491 * Steal some nice code to GNU pth to fix context detection and usage [AL]
492 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
493 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
496 * Bugfix: Allow absolute paths for platform description files [MQ]
497 * Bugfix: do free the variables after use. Leads to drastic performance
499 * Implement max_duration (ie, timeouts) on resources [AL]
502 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
504 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
505 the needed support (timeouts) [AL]
508 * Implement measurement sockets. You can now get the bandwidth between two
509 hosts thanks to AMOK (see below). [MQ]
510 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
511 dynar over the network (yeah) [MQ]
512 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
513 * simple/static token-ring example. [Alexandre Colucci and MQ]
514 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
518 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
519 type parsing. "short" and "long" modifiers were also ignored; other
520 modifier (such as reference level) are still ignored. [MQ]
521 * Embeed the buffer size within the buffer itself on SG. [MQ]
522 That way, send() are atomic and cannot get intermixed anymore (at least
523 the ones which are less than 100k; bigger messages still have the issue)
524 * Array size pushed by the field, not by the field type (or each
525 and every long int will push stuff to the cbps) [MQ]
526 * use select() to sleep since it allows to portably sleep less than one
529 GRAS (minor cleanups)
530 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
533 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
534 * type category 'ignored' killed as it was never used and were difficult
536 * whether a type can cycle or not is now a flag, leaving room for more
537 flags (as "ignored", if we feel the need one day ;)
538 * Rename raw sockets to measurement sockets since "raw" has another
539 meaning in networking community.
542 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
543 and offers features not belonging to GRAS but that most applications
544 need. One day, it may be a set of plugins loadable at runtime.
545 * New module: bandwidth
546 bandwidth measurement between arbitrary nodes running this module. [MQ]
548 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
550 SimGrid (2.94) unstable; urgency=low
552 The first beta release of SimGrid 3 !
556 * Update the main page and the FAQ. Adding references to gforge.
559 * Add a gras_os_getpid function.
562 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
563 * Extend the logs so that they also print PID, hostname, date, ... if
565 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
566 and kill the old version which were in testsuite/
567 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
571 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
572 for users. Instead, each programming environment (GRAS, MSG,...) use it
573 when needed to provide such a feature to users.
574 Don't shortcut the mecanism or you will also shortcut the virtualization
575 you need on the simulator.
579 * Cleanups in configury with regard to compile optimization/warning flags.
580 Also add -fno-loop-optimize to any powerpc since it's the optimization
581 killing gcc (< 3.4.0).
582 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
583 needed by MSG examples complications
584 * Borrow configury beautifications from PHP
587 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
588 warning (thanks loris for stumbling into this one).
589 * Bugfix: stop loading private headers (gras_config.h) from the public
593 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
594 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
595 RL would imply a DNS resolver, which is *hard* to do in a portable way
596 (and therefore delayed).
597 * Implement a real timer mecanism and use it in timing macros. This allows
598 to avoid rounding errors and get a 0.000005 sec precision in timing
599 macros. While I was at it, various cleanups:
600 - allow to declare more than one timed section per file (fix a stupid bug)
601 - move some private declaration to the right place
602 - merge conditional execution and timing macros into emulation module
603 - document the module
604 - make sure the module cleanups its mess on gras_exit
605 * Documentation improvements:
606 - (new) how to compile applications using GRAS
607 - (new) emulation support (timing macros)
609 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
611 SimGrid (2.93) unstable; urgency=low
613 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
616 - Use Paje properly where used. Still to be sanitized properly.
617 - Portability fix: Add an implementation of the contexts using pthread
621 - Add xbt_procname(): returns the name of the current process.
622 Use it to show the current process's name in all logging.
624 - fix detection of older flex version and the reaction, since we do
625 depend on modern ones (we use lex_destroy)
626 - Better separation of SG and RL in the libs: remove all simulation code
627 from libgras. As a result, this lib is now only 200k when stripped.
628 Some of the xbt modules may also be duplicated (two sets and such) and
629 should be cleaned/killed before SG3.
630 - Insist on using xlC on AIX because of weird problems involving gcc there.
631 - Cleanup the make remote stuff. This is now done by scripts
632 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
633 mainly for our private use, but we're working on changing them to user
636 - Bugfix: flush the socket on close only if there is some *output*.
637 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
639 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
642 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
644 SimGrid (2.92) unstable; urgency=low
646 Alpha 3 on the path to SimGrid 3
650 - New! First try of benchmarking macros.
651 - New! First try so that gras_stub_generator generate deployment and
652 remote compilation helpers.
654 - Bugfix: Initialization fix in msg_test.
658 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
660 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
662 - API Change: gras_os_sleep to take the amount of seconds as a double.
663 Accepting an int was error prone since it was the only location where
664 seconds were coded as such. It leaded to damn rounding errors.
665 - Bugfix: Hard to belive that timers ever worked before this.
667 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
669 SimGrid (2.91) unstable; urgency=low
671 Alpha 2 on the path to SimGrid 3
675 - Bug fix in the lmm_solver.
677 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
678 through the function MSG_paje_output.
679 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
680 - It is possible to bound the rate of a communication in MSG with
681 MSG_task_put_bounded() (was already in the previous version but I had forgotten
682 to write it in the changelog).
683 - Bug fix to let GRAS run on top of MSG until we move it directly on top
688 - Various cleanups to the autotools stuff
689 - Begin to move Gras examples to examples/gras/
690 - Let make distcheck work again (yeah!)
692 - documentation overhauled using doxygen.
693 gtk-doc-tools is dead in SimGrid now.
694 - Automatically extract all existing logging categories, and add the list
695 to the documentation (long standing one, to say the less)
697 - Cleanup the known architecture table. Reorder the entries to group what
698 should be, and use a more consistent naming scheme.
699 (some of the test dataset are still to be regenerated)
700 - New! Allow library to register globals on each process just as userdata
702 This is implemented using a xbt_dict and not a xbt_set, so we loose the
703 lookup time (for now).
704 Use it in msg and trp.
705 This cleans a lot the internals and helps enforcing privacy of the
706 headers between the gras components.
707 - New! Add a timer mechanism, not unlike cron(8) and at(1).
708 - Bugfix: gras_os_time was delirious in RL.
709 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
711 - Reenable GRAS now that it works.
713 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
715 SimGrid (2.90) unstable; urgency=low
717 Alpha 1 on the path to SimGrid 3
719 * It is a long time since the last release of SimGrid. I'm sorry about
720 that but as I had told you, I was rewriting a lot of things. I apologize
721 to those who had been reporting bugs to me and that I had not answered.
722 If your bug is still in the new version, please tell me. Here is a
723 summary of the main changes.
725 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
726 lead by Martin Quinson. As a consequence SimGrid gains a lot in
727 portability, speed, and a lot more but you'll figure it out later.
728 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
729 I wanted to release the new MSG as soon as possible and I have
730 broken GRAS, which is the reason why, for now, only MSG is fully
731 functional. A laconic description of these projects is available
732 in the documentation.
734 * REVOLUTION 2: I have removed SG and I am now using a new simulation
735 kernel optimized for our needs (called SURF but only the developers
736 should use it). Hence, MSG is now roughly 30 times faster and I think
737 that by rewriting a little bit MSG, I could event speed it up a little
738 bit more. Beside the gain in speed, it is also much easier to encode a
739 new platform model with SURF than it was with SG. More to come...
741 * REVOLUTION 3: I have tried to change a little as possible the API of
742 MSG but a few things really had to disappear. The main differences
743 with the previous version are :
744 1) no more m_links_t and the corresponding functions. Platforms are
745 directly read from a XML description and cannot be hard-coded
746 anymore. The same format is used for application deployment
747 description. The new format is described in the documentation.
748 Have a look in tools/platform_generation. There is a tiny script
749 that converts from the old platform format to the new one. Concerning
750 the application deployment format, parsing the old one is tricky.
751 I think most of you should however be able to convert your files. If
752 it is really an issue, I can write a C code that does the conversion.
754 2) the toolbox tbx does not exist anymore. We now have a library
755 with much more data-structures but without the hash-tables (we have
756 dictionaries that are much faster).
758 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
760 *****************************************************************************
761 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
762 * don't feel like converting the oldies. (Mt) *
763 *****************************************************************************
766 Version 2.90: "the long awaited one"
767 - Finished rewriting and debugging MSG. Rewrote the documentation.
768 - disable GRAS for now since it needs to be ported to the newest SG
771 - Finish the port to windows (using mingw32 for cross-compile)
774 - Main loop and datastructures of SURF. A cpu resource object is
775 functional. Surf can thus be used to create cpu's with variable
776 performance on which you can execute some actions.
778 2004-11-15 Martin Quinson
779 - Port to ARM. Simply added the alignment and size descriptions. Should
780 work, but the ARM machines are so slow that I didn't had the opportunity
781 to 'make check' over there yet.
783 2004-11-15 Arnaud Legrand
784 - Trace manager now written. It uses a heap structure and is therefore
785 expected to be efficient. It may however be speeded up (particularly
786 when many events occur at the same date) by using red and black
787 trees. One day maybe...
788 - Max-min linear system solver written. It uses a sparse matrix
789 structure taking advantage of its expected use. Most operations are
790 O(1) and free/calloc are called as few as possible. The computation of
791 the minimum could however be improved by using a red and black tree
794 2004-11-03 Arnaud Legrand
795 - Rename every gras_* function that was in xbt/ to its xbt_
797 - Add a heap and a doubly-linked list to xbt
798 - Added a dichotomy to the dictionaries. make check works as well before
799 so I assume that the patch is correct. I do not know however if things
800 run effectively faster than before now. :)
802 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
803 SimGrid, and the version number is bumped to 2.x
805 2004-10-29 Martin Quinson
806 - Introduction of the remote errors.
807 They are the result of a RMI/RPC on the remote machine.
808 ErrCodes being scalar values, you can't get the host on which those
809 errors did happen. Extending the error mechanism as in Gnome is possible.
810 No idea yet whether it is a good idea.
812 2004-10-28 Martin Quinson
813 - Interface revolution: the Starred Structure Eradication.
814 I used to do typedef struct {} toto_t; and then handle *toto_t.
815 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
816 - ???_t is a valid type (builded with typedef)
817 - s_toto_t is a structure (access to fields with .)
818 - s_toto is a structure needing 'struct' keyword to be used
819 - e_toto_t is an enum
820 - toto_t is an 'object' (struct*)
822 typedef struct s_toto {} s_toto_t, *toto_t;
823 typedef enum {} e_toto_t;
824 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
825 s_toto_t) is private.
827 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
828 so that it can set the variable to NULL. It was so for dicts and sets,
829 it changed for dynars.
831 - Fix a bunch of memleaks in dict_remove
832 - Fix a bug in sg/server_socket opening: it failed all the time.
834 2004-10-07 Martin Quinson
835 - Speed up dynar lookup operation a bit.
837 gras_dynar_get is dead.
839 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
840 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
841 (which gives you the address of the stored data).
843 gras_dynar_get_as is an helpful macro which allows you to retrieve a
844 copy of the data using an affectation to do the job and not a memcpy.
846 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
847 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
849 It does not really speedup the dynar test because they are
850 setting elements all the time (and look them seldom). But the dict does
851 far more lookup than setting.
853 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
855 2004-10-05 Martin Quinson
856 - Allow to (en/dis)able the cycle detection at run time.
858 Whether we should check for cycle or not is now a property of each
859 datatype. When you think there may be some cycle, use datadesc_cycle_set.
860 datadesc_cycle_unset allow to remove this property when previously set.
862 Note that the cycle detection is off by default since it impacts the
863 performance. Watch the data you feed GRAS with ;)
865 This property is hereditary. Any element embedded in a structure having it
866 set have it set for the time of this data exchange.
868 You should set it both on sender and receiver side. If you don't set it on
869 sender side, it will enter an endless loop. If you forget on receiver
870 side, the cycles won't be recreated after communication.
872 - Header reorganization.
873 Kill gras_private.h, each submodule must load the headers it needs.
875 2004-10-04 Martin Quinson
876 - Interface revolution: do not try to survive to malloc failure.
878 Now, gras_malloc and friends call gras_abort() on failure.
879 As a conclusion, malloc_error is not a valid error anymore, and all
880 functions for which it was the only gras_error_t return value are
881 changed. They now return void, or there result directly.
882 This simplify the API a lot.
884 2004-09-29 Martin Quinson
885 - Re-enable raw sockets.
886 Created by gras_socket_{client,server}_ext;
887 Used with gras_raw_{send,recv}
890 It should allow to kill the last bits of gras first version soon.
892 This is not completely satisfactory yet (duplicate code with
893 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
896 - Simplify transport plugin (internal) interface by not passing any
897 argument to _server and _client, but embedding them in the socket
900 2004-09-28 Martin Quinson
901 - Finish the port to AIX.
902 autoconf was my problem (segfault within the malloc replacement
903 function. No idea why)
905 2004-09-16 Martin Quinson
906 - Fix some size_t madness on 64bit architectures.
908 2004-09-08 Martin Quinson
909 - Reduce the number of system headers loaded, overload some more system
910 calls (such as malloc to cast the result of the system one, and work
912 - Fix and reintroduce the config support
914 2004-09-07 Martin Quinson
915 - Source code reorganization to allow Arnaud to surf all over there.
916 - Allow to document the logging categories.
917 - Remove all uppercase from logging categories and useless cleanup in names.
919 2004-08-18 Martin Quinson
920 Version 0.6.2 (protocol not changed; API changed)
921 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
922 gras_error_t), and NULL when not found. Functions expecting a msgtype
923 as argument (msg_wait; msg_send) deal with NULL argument by providing a
924 hopefully usefull message.
925 - Portability to prehistoric sparcs again
927 2004-08-17 Martin Quinson
928 Version 0.6.1 (protocol not changed; ABI not changed)
929 - prealloc some buffers to speed things up
931 2004-08-11 Martin Quinson
932 Version 0.6 (protocol not changed; ABI expended)
933 - The parsing macro can deal with the references, provided that you add
934 the relevant annotations (using GRAS_ANNOTE(size,field_name))
936 2004-08-09 Martin Quinson
937 Version 0.5 (protocol not changed; ABI changed)
938 - Allow to off turn the cycle detection code in data exchange at
939 compilation time. It should be at run time, but I'm short of time (and
940 the config stuff is still broken). That way, we keep dict out of the
941 critical path, which is good because the performance is poor:
942 - search not dichotomial yet
943 - dynar give no way to access their content and memcpy everytime
944 - In composed data description (struct, ref and so on), stop foolness of
945 keeping the subtype's ID, but store the type itself. This keeps sets out
946 of the critical path, which is good since they rely on dynar and
947 dictionnaries. The only loose of that is that we cannot detect the
948 redeclaration of a structure/union with another content (but I'm not sure
949 the code detected well this error before anyway). We still can detect
950 the redefinition discrepancy for the other types.
951 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
952 it breaks the code because of type-punning used all over the place).
953 This breaks on all non-gcc architectures (for now).
955 All those changes (plus the buffer of last time) allow me to gain 2 order
956 of magnitude on cruel tests consisting of 800000 array of integers on two
957 level of a hierarchical structure (200 secondes -> 4 secondes)
960 - the selector of reference must now return the type it points to, not
963 2004-08-06 Martin Quinson
964 Version 0.4 (protocol changed; ABI not changed)
965 - Allow to pass --gras-log argument to processes in simulation mode. Really.
966 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
967 - Implement a buffer transport, and use it by default (it relies on tcp in
968 real life and on sg in simulation).
969 That's a bit hackish since I had a new field to the structure to store
970 its data without interfering with the subtype ones. Inheritance
971 is tricky in C. And that's a kind of reverse inheritance with one class
972 derivating two classes. Or maybe a game with java interfaces. Anyway,
973 that's damn hard in C (at least).
974 Moreover, I got tired while trying to ensure plugin separation and
975 genericity in SG mode. MSG wants me to do weird things, so let's go for
976 cruel hacks (temporarily of course ;).
977 See comment in transport_private.h:71
978 - do not include all the _interface headers in private but in the files
979 which really need them (to cut the compilation time when they are
982 2004-07-26 Martin Quinson
983 Version 0.3 (protocol not changed; ABI changed)
984 - Major overhault of the datadesc interface to simplify it:
985 - shorted the function names:
986 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
987 - add a trivial way to push/pop integers into the cbps without malloc.
988 This allows to make really generic sub_type description, which simply
989 pop their size of the stack.
990 - add a function gras_datadesc_ref_pop_arr() which does what users want
991 most of the time: Declare a dynamic array (which pops its size of the
992 stack) and declare a reference to it. Poor name, but anyway.
993 - kill the post-send callback, add a post-receive one
995 2004-07-23 Martin Quinson
996 Version 0.2 (protocol changed; ABI changed)
997 - add some testing for cpbs in the test cases, and fix some more bugs.
998 This invalidate again the little64 data file, since I cannot regenerate
1000 - remove an awfull optimization in the logging stuff, allowing me to:
1001 - understand it again
1002 - learn gcc how to check that the argument match the provided format
1003 - fix all errors revealed by gcc after that
1004 - internal keys of dict are not \0 terminated. Deal with it properly in
1005 loggings instead of segfaulting when the user want to see the logs :-/
1007 2004-07-22 Martin Quinson
1008 - Fix some stupid bug preventing cbps (callback postit) from working
1010 2004-07-21 Martin Quinson
1011 - Some documentation cleanups
1012 - remove the useless last argument of msgtype_declare
1013 - rename the Virtu functions to fit into the 'os' namespace
1014 - move headers src/include -> src/include/gras/ and stop fooling with
1015 gras -> . symbolic link
1016 - make distcheck is now successful
1018 2004-07-19 Martin Quinson
1020 - Build shared library also
1021 - Install html doc to the right location
1022 - stop removing maintainer files in make clean
1023 - build tests only on make check
1025 2004-07-13 Martin Quinson
1027 - No major issue in previous version => change versionning schema
1028 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1029 corresponding dataset.
1031 2004-07-11 Martin Quinson
1033 - Get it working with any kind of structure (we can compute the padding
1034 bytes remotely for all the architectures I have access to)
1035 - Implement the structure parsing macro (still not quite robust/complete)
1036 - Improvement to the remote testing toysuite
1038 2004-07-10 Martin Quinson
1039 [autoconf mechanism]
1040 - get ride of a bunch of deprecated macros
1041 - actually run the test for two-compliment, not only compile it :-/
1042 - test whether the structures get packed (and bail out if yes. Damn.
1043 Alignment is a serious matter)
1044 - test whether the structures get compacted (but respecting the alignment
1045 constraints of each types)
1046 - test whether the array fields of structures can straddle alignment boundaries
1048 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1049 (took me 2 hours to find that bug, looking at the wrong place)
1050 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1051 before sending/receiving and are used to compute the offsets of fields
1052 - Given that padding size depend even on compiler options, keep track of
1053 alignment and aligned_size only for the current architecture. Not a big
1054 deal since we send structure fields one after the other (seems
1056 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1057 program, let it work on linux/gcc/little32. portability todo.
1059 2004-07-08 Martin Quinson
1060 - import and improve remote compilation support from FAST
1061 - make sure make check works on half a dozen of machines out there
1063 2004-07-07 Martin Quinson
1064 Let's say it's version 0.0.3 ;)
1065 - Implement conversions (yuhu!)
1066 - Let it work on solaris (beside conversion, of course)
1067 - Stupid me, using rand() to generate the conversion datatests in not wise.
1069 2004-07-06 Martin Quinson
1070 - Let make dist work, since I'm gonna need it to compile on remote hosts
1071 - Let Tests/datadesc_usage write the architecture on which the file was
1072 generated as first byte.
1073 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1074 architecture descriptions.
1075 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1076 on those architectures.
1077 - Optimization: send/recv array of scalar in one shoot
1079 2004-07-05 Martin Quinson
1080 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1082 - Plug a whole bunch of memleaks
1083 - each process now have to call gras_{init,exit}. One day, their log
1084 settings will be separated
1085 - Continue the code factorisation between SG, RL and common in Transport.
1087 2004-07-04 Martin Quinson
1089 - Redistribution between SG and RL.
1090 We wanna have to accept in SG, so move accepted related parts of RL in
1091 the common part. (more precisely, the dynar of all known sockets is no
1092 more a static in transport.c, but part of the process_data)
1094 [gras_stub_generator]
1095 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1097 2004-07-03 Martin Quinson
1098 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1099 them all up in one shot)
1101 - Fix the ugly case of reference to dynamic array.
1102 - New (semi-public) function gras_datadesc_size to allow the messaging
1103 layer to malloc the needed space for the buffer.
1105 - gras_socket_close now expect the socket to close (and not its address to
1106 put NULL in it after it). This is because the socket passed to handlers
1107 is one of their argument (=> not writable).
1109 - propagate the interface cleanup from last week in datadesc, ie remove a
1110 superfluous level of indirection. User pass adress of variable
1111 containing data (both when sending and receiving), and not of a variable
1112 being a pointer to the data. Let's say that I like it better ;)
1113 The price for that is constructs like "int msg=*(int*)payload" in
1114 handlers, but it's a fine price, IMHO.
1116 - Let it work in RL (yuhu)
1118 2004-06-21 Martin Quinson
1120 - porting SG plugin and SG select to new standards (works almost).
1121 - plug memleaks and fix bugs around.
1124 - cleanup the prototype of data recv and force users to specify when they
1125 want to handle references to objects. Test case working even for cycles.
1126 - plug memleaks. Valgrind is perfectly ok with this.
1128 2004-06-12 Martin Quinson
1130 - cleanup the separation between plugin and main code in plugin creation
1132 2004-06-11 Martin Quinson
1134 - Reput hook for raw sockets, needed for BW experiments
1135 - kill a few lines of dead code
1136 [Data description] Interface cleanup
1137 - gras_datadesc_by_name returns the searched type or NULL.
1138 That way, no variable is needed to use a type desc once, which makes
1140 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1141 two parameters were strings (field name, type name), leading to
1143 [Dicos] Interface cleanup
1144 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1145 This is consistant with the dynar API.
1147 2004-04-21 Martin Quinson
1149 - Porting to new standards.
1151 - interface cleanup.
1152 There is no bag anymore, no need to take extra provision to mask the
1153 pointers behind "ID".
1154 Better splitup of functions between files create/exchange/convert.
1155 This is still a bit artificial since convert and receive are so
1156 interleaved, but anyway.
1158 - add a queued message list to procdata (the ones not matching criteria
1160 - factorize some more code between SG and RL wrt procdata
1162 - use gras_exit in example to track memleaks
1163 - get rid of gs_example now that GS is properly integrated into gras
1164 - update run_test to integrate the lastest tests (datadesc)
1166 - rename WARNINGn macros to WARNn since it prooved error-prone
1168 2004-04-19 Martin Quinson
1170 - register init/exit functions within gras module mechanism
1171 - send/receive function.
1172 Convertion is not implemented, but short-cutted if not needed.
1173 struct/array elements are sent one by one (instead of block-wise), but
1174 nobody really cares (yet). Get a prototype before optimizing.
1175 - tests (using a file socket) for DD send/receive on:
1176 - base types: int, float
1177 - array: fixed size, string (ie ref to dynamic string)
1178 - structure: homogeneous, heterogeneous
1179 - chained list, graph with cycle
1180 Believe it or not, valgrind is not too unhappy with the results. The
1181 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1185 - Bugfix when using a filename explicitely (instead of '-')
1187 2004-04-09 Martin Quinson
1189 - factorize more code between RL and SG in socket creation
1190 - Complete the implementation and tests of:
1192 o file (only in RL, and mainly for debugging)
1194 I lost 3 days to design a portable address resolver, and then decided
1195 that the prototype mainly have to run on my box.
1196 Addressing portability too early may be like optimizing too early :-/
1198 - use gras_init in the Tests instead of the crappy parse_log_opt
1199 (the latter function is removed)
1200 [Conditional execution]
1201 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1202 [Code reorganisation]
1203 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1204 Build examples against the RL library, since there is no way to disable
1205 its creation for now.
1207 For information, the beginning of coding on GRAS was back in june
1208 2003. I guess that every line has been rewritten at least twice since