1 SimGrid (3.4-svn) unstable; urgency=high
4 * Tested successfully on AIX
7 * fix a bug on struct sizeof computation, which prevented the
8 exchange of arrays of structs in some conditions
9 - added a regression test about this in datadesc_usage
10 * Allow the exchange of 0-long dynamic vectors.
11 - for that, use -1 as indicator of dynamic size instead of 0
12 - This implied to change any size from unsigned long to long,
13 reducing a bit communication abilities, but I guess that with
14 64bits being quite common, this is more than enough.
15 - This also induce a protocol change, thus bumping network protocol
16 version from 0 to 1 (if we have external users, we have to get
17 clean on that point too ;)
18 - added two regression tests about this in datadesc_usage
19 * Be more verbose when propagating local exceptions
21 * Display the status of simulated processes when receiving SIGINT in
25 * Allow to control the simulation from a trace file.
26 New functions MSG_action_register() and MSG_action_trace_run()
27 The first one allows to associate a function execution to each
28 kind of action while the second one parses a trace file and
29 triggers the corresponding actions within the system.
30 For now, only a toy example is provided in examples/msg/actions
31 * Add an exemple of process migration in examples/msg/migration
32 * Fix a bug in task exchange which broke MSG_task_get_sender()
33 Add a teshsuite regression test for that.
34 [Bug: if MSG_task_get_sender() is called after sender exit,
36 * Fix a bug which prevented suspend/resume to work properly
37 * Display the status of simulated processes when receiving SIGINT
38 This fixes a regression of v3.3. due to the introduction of SIMIX
39 * Bug fixing in failure management:
40 - trace could not start by a failure at time 0
41 - failure during communications were not working
44 * Add SIMIX_process_set_name() to change the name of the current
45 process in the log messages.
46 * Store smx_hosts in a dict since we only retrieve them by name
49 * Also include strbuff from xbt.h public header
50 * xbt_ex_display(): do not free the exception after displaying
51 This allows to do more with the given exception afterward.
52 Users should call xbt_ex_free() themselves.
55 * Massive internal cleanups:
56 - Cleanup the initialization/finalization process
57 - Kill a whole bunch of unneeded synchronization:
58 processes run in exclusive manner within the simulator
59 - Move queues from global tables to process data fields
61 - now accept -platform and -hostfile arguments
62 - Pass the right rank value to processes according on hostfile
65 * Do not require doxygen in maintainer mode
67 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
69 SimGrid (3.3) stable; urgency=high
73 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
76 * Introduce the SIMIX module: factorize code between MSG and GRAS.
79 Until now, GRAS were using MSG as an interface to SURF. It was
80 quite difficult because both interface have several differences
81 (MSG channels vs GRAS sockets were the most notable point).
83 This also opens the gate to SMPI (which should occur soon) and speed
84 up simulations by to 40% (even if it were not the main goal).
86 **************************************
87 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
88 **************************************
89 The point is that events occuring at the exact same timestamp are
90 not scheduled in the same order with the old and new version. This
91 may be enough to completely change the execution of simulations in
92 some cases. Sorry for the inconvenience.
94 * Cleanup and upgrade the XML format to push further scalability
95 issues (check http://hal.inria.fr/inria-00256883/ for more info)
97 * Improve the testing infrastructure with tesh. Now a very large part of
98 the code is tested not only by being run but also by checking that the
99 output match an expected output [Mt].
101 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
102 is really less memory-demanding, which should allow you to use
103 larger files in SimGrid [AL].
105 * Inform valgrind about our contextes, so that it becomes usable
106 with the default (and more effecient) version of SimGrid
107 [contributed by Sékou Diakite, many thanks]
110 * Introduce a listener thread in charge of receiving incomming
111 messages from the network. It allows to overlap communication and
112 computation but most notably, it removes some stupid deadlocks due
113 to the fact that so far, a process could not send and receive at
114 the same time. This made most non trivial communication schema
116 * Convert the PIDs from long int to int to match the MSG ones (and
118 * New function: gras_agent_spawn() to launch a new process on
119 current host. Only working in simulation for now. [Mt]
120 * New function: gras_os_hostport() returning a constant form (ie,
121 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
124 * Make the backtrace of exceptions more human readable [Mt]
125 * New module: xbt/str [Mt]
126 a ton of string utility functions (split, join, printf to a newly
127 allocated buffer, trim, etc)
128 * New module: xbt/hash [Mt]
129 SHA1 hashing algorithm (more to come if needed)
130 * New module: xbt/synchro [Mt]
131 synchronization tools (mutex and conditions) working the same way
132 in simulation and in real life (mainly useful for GRAS, but not
134 * New module: xbt/queue [Mt]
135 classical producer/consumer synchronization scheme
136 * xbt_dynar_new_sync() creates a synchronized dynar. All access
137 (using the classical functions will get serialized) [Mt]
138 * Make dictionary internal table dynamic. No need to specify its size
139 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
141 * Make sure the log channels are organized as a tree under windows
142 (because of ANSI C compatibility issue, any channel were child of
146 * Cleaned many thing in surf and fixed a few bugs [AL].
147 * Add a nice command line configuration mechanism to compose models [AL].
148 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
149 the previous one (KCCFLN05). It relies on something that looks like
150 a max-min sharing mechanism but cannot be written as such. A new solver
151 was thus designed [AL].
152 * Add a new solver to lmm. Based on Lagrange optimization and
153 gradient-based descent, it enables to efficiently maximise systems s.a
155 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
157 This solver enables to propose two new network models for TCP Reno and
158 TCP Vegas based on Low's work. These models still need to be fully
159 tested though [Pedro Velho].
162 * Bug fix in SD_simulate. Now the time bound given as argument is
164 * Use the new parallel task model (ptask_L07) as default.
165 * Use the SURF command line configuration mechanism.
166 * 0-size tasks (for synchronization) should now work.
168 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
170 SimGrid (3.2) stable; urgency=high
174 We still experience issues on this platform, but we believe that at
177 GRAS API BREAKAGE (for simplification purpose, sorry):
178 * the gras_msgtype_by_name is not used anymore. Instead of
179 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
180 you can write (and must)
181 gras_msg_send(toserver, "request", &request);
182 - If you still want to pass a gras_msgtype_t to the function (to cache
183 the type and avoid the lookup time), use the gras_msg_send_() variant.
184 - Impacted functions:
185 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
186 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
187 * The callbacks are now expected to return 0 when everything went well
188 (just like the main() function)
190 GRAS new features and improvements:
191 * New module mecanism where user code can use per process globals [Mt]
192 This is similar to gras_userdata_*() functions, but for libraries. It
193 factorize some code developped over and over in the examples and AMOK.
194 It has still to be documented and used (only amok/peermanagement is
196 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
197 loss when they were small enough to fit into the buffer and sent quickly
198 enough so that they can all get received in one shoot.
199 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
200 if not found. Use the *_or_null() variant for the old semantic.
201 * In gras_msg_handle, do not discard messages without callback.
202 They are probably messages to be explicitly awaited later (ie, proofs of
203 mis-synchronization in userland since they are sent before being awaited)
205 * gras_socket_meas_send/recv: semantic changed!
206 The numerical arguments used to be (1) the total amount of data to send
207 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
208 messages. This was need for the fool willing to send more than MAXINT
209 bytes on quite fat pipes.
212 * Do really rename the hostmanagement module to peermanagement. [Mt]
213 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
214 but this is rather new and this was documented in the module
215 documentation (poor excuses, I admit)
216 * Bandwidth measurement semantic changed! This follows the changes to
217 gras_socket_meas_send/recv explained above.
220 * A sequential mode has been added to the workstations. When a workstation
221 is in sequential mode, it can execute only one task, and the other tasks
222 are waiting in a FIFO. [Christophe Thiery]
225 * The KCCFLN05 workstation model now handles parallel tasks. It is the
226 model for SIMDAG. [Christophe Thiery]
227 * Bug fix in the maxmin solver: Some values were close to 0 instead of
228 equal to 0, which caused some bad behaviors in
229 saturated_constraint_set_update. I now use a threshold mechanism like in
233 * When running manually src/testall, you select specific units [Mt]
234 testall is the result of our cunit mecanism, and should replace all
235 the scripty thingy around since bash don't run easily on billware.
237 * A mallocator system has been added. [Christophe Thiery]
238 Mallocators allow you to recycle your unused objects instead of freeing them
239 and allocating new ones.
241 Documentation update:
242 * FAQ reworking + New FAQs:
243 - "Valgrind spits tons of errors!" [Mt]
244 - "How to repport bugs" [Mt]
245 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
246 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
247 - Communication time measurement within MSG [AL]
248 - I experience weird communication times when I change the latency [AL]
251 - an introduction to the framework and to the used communication model
252 - an initiatic tour introducing the most proheminent features:
254 . Lesson 0: Installing GRAS
255 . Lesson 1: Setting up your own project
256 o Part 2: Message passing
257 . Lesson 2: Exchanging simple messages
258 . Lesson 3: Passing arguments to the processes (in SG)
259 . Lesson 4: Attaching callbacks to messages
260 . Lesson 5: Using globals in processes
261 . Lesson 6: Logging informations properly
262 . Lesson 7: Using internal timers
263 . Lesson 8: Handling errors through exceptions
264 . Lesson 9: Exchanging simple data
265 . Lesson 10: Remote Procedure Calling (RPC)
266 . Lesson 11: Explicitely waiting for messages
267 . Recapping of message passing features in GRAS
268 - A HOWTO section containing:
269 o HOWTO design a GRAS application
270 More are due, of course. They will come latter. In the meanwhile, you can
271 check the examples which are still here.
273 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
275 SimGrid (3.1) stable; urgency=high
279 There was a stack corruption somewhere, visible only when optimizing
280 with these versions. [Vince]
283 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
284 module that enabled to play with Directed Acyclic Graphs. It is built
285 directly on top of SURF and provides an API rather close to the old
286 SG. Some old codes using SG are currently under rewrite to check that
287 all needful functions are provided. [Christophe Thiery]
290 * Complete rewrite of the KCCFLN05 workstation model. It is now an
291 extension of the classical CLM03 model that gracefully handles
292 failures. This is now the default model for MSG and GRAS. It doesn't
293 handle parallel tasks yet though. [AL]
294 * Bug fix: Weights were not correctly set in the network part.
295 WARNING: This may have resulted in incorrect results with simulations
296 where there are more than one flow on a given link. [AL]
299 * After a (long ?) discussion on simgrid-devel, we have decided that the
300 convention we had on units was stupid. That is why it has been decided
301 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
302 WARNING : This means that all previous platform files will not work as
303 such with this version! A warning is issued to ask users to update
305 A conversion script can be found in the contrib module of the CVS, under
306 the name contrib/platform_generation/surfxml_update.pl [MQ]
309 * Bug fix: Processes were started in reverse order, wrt deployment file.
310 WARNING: if your code relies on this bug, please fix it. [AL]
311 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
312 being executed on two different locations. [AL]
313 * Bug fix: Failures are now better supported thanks to Derrick's tests
314 (there was many failure situations I hadn't thought of and that weren't
315 correctly handled). [AL]
316 * New function: MSG_host_is_avail indicates you whether a given m_host_t
320 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
321 Exception occurring on server-side are propagated back to client (!).
323 API CHANGE: the callback changed their prototype. Change:
324 int my_handler(gras_socket_t expeditor, void *payload_data) {
326 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
327 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
329 * New! function: gras_msg_handleall to deal with all messages arriving
330 within a given period.
331 * New! function: gras_socket_server_range to get a server socket in a
332 range of port numbers (ease to avoid port number conflicts) [MQ]
333 * New! gras processes display their backtrace when they get a SIGUSR1
334 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
335 Sweet to debug RL processes [MQ]
339 - Do not force experiment sizes to be expressed in kb, or it becomes
340 impossible to measure the latency this way (needs one byte-long tests)
341 WARNING: this changes the amok_bw_* function semantic. [MQ]
342 - Implements the link saturation stuff. [MQ]
343 * Peer management module:
344 New! module factorizing code that we wrote over and over [MQ].
347 * New module: cunit (my jUnit implementation in ansi C) [MQ]
348 - Test units are placed directly into the library code, they get extracted
349 automatically and placed into the src/testall binary.
350 - Convert most of the XBT tests to this system.
351 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
352 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
353 live in the same C code. This required to change a little bit the API
354 of surfxml parsing but shouldn't be an issue for end-users. [AL]
355 * New module: sparse graph structure with basic algorithms (this is work
356 in progress and the API is not considered to be frozen yet). [AL]
357 * Display more information on backtraces: source line & function names are
358 now displayed just like valgrind does (rely on addr2line tool) [MQ]
359 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
360 * Reworked a little bit some #include statements to load only required
361 headers. Some user code that relied on SimGrid to include stdlib or
362 stdio may need to include it by themselves. [AL]
363 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
365 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
366 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
367 but I guess that almost nobody used those parts. [MQ]
369 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
371 SimGrid (3.0.1) stable; urgency=low
374 * Unfortunately, I had missed 5 misnamed functions:
375 xbt_fifo_item_t xbt_fifo_newitem(void);
376 void xbt_fifo_freeitem(xbt_fifo_item_t);
377 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
378 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
379 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
380 They're now deprecated. Please use their new versions:
381 xbt_fifo_item_t xbt_fifo_new_item(void);
382 void xbt_fifo_free_item(xbt_fifo_item_t);
383 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
384 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
385 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
387 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
388 * Documentation: xbt_log module unmercifully reworked [MQ]
389 * Bugfix: there was a problem with the ending of contexts with
390 the pthread backend. It caused some weird deadlock or behavior
391 depending on the pthread implementation. [AL]
392 * Bugfix: get the exceptions raised in the simulator repport where
393 and why they come from when they are not catched in time [AL, MQ]
396 * Bugfix: Do repport the error when two non-connected hosts try to
397 exchange data (Thanks to Flavien for stumbling into this one) [AL]
400 * Add additionnal checkings on communications. Assert that two
401 communicating hosts are connected by a set of links... [AL]
404 * Add additionnal checkings on channel values in communication [AL]
405 * New: MSG_task_get_source to see on which host a task was generated [HC]
406 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
407 the number of tasks waiting to be received on channel and sent
409 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
410 waits for the first task coming from a given host.. [AL]
412 GRAS new functionnalities: [MQ]
413 * Enhance the parsing macro to allow the size of multidimentional objects
414 to be given thru annotations.
415 * New example (and documentation): Matrix Multiplication a la RPC
416 (as when I was young!) and fix a bunch of bugs found on the way.
418 GRAS performance improvements: [MQ]
420 * Reduce the amount of cbps creation/destruction by making it static to
421 datadesc_send/recv() and using a (newly created) cbps_reset (based on
424 * Change libdata to a set so that we can search for stuff by ID (and thus
425 reduce the insane amount of dict lookups)
428 * Actually implement gras_datadesc_copy() so that we don't have to mimick
429 RL communication on top of SG since it's so uneffective.
430 It may also be used for inter-thread communication in RL, one day.
431 * Use gras_datadesc_copy() to exchange messages on top of SG
433 - improve message exchange performance on top of SG
434 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
435 * Don't exchange on the network the size of the used part of buffer,
436 instead, specify the possible buffer size to read().
438 - reduces the amount of read/write calls (one pair per exchange)
439 - reduces the amount of exchanged data (the size)
440 - allows to retrieve all arrived data on receiver side, if we don't need
441 it right now (subsequent read will peek the buffer)
442 - allows the receiver to proceed with the begining of the stream before
443 everything is arrived
444 - make it possible to build an iov transport (using readv/writev)
446 - take care of the data with non-stable storage (like stacked data),
448 * If possible, TCP send uses vector I/O (when writev() is here)
449 - Don't use it for receive since we send data sizes and data on the
450 same stream, so we wouldn't be able to chain large amount of chunks
451 before having to flush the stuff to read the size.
452 * Rework the transport plugin mecanism to simplify it and reduce the
453 amount of pointer dereferencement when searching for the right function
456 * I guess that now, we do almost as few system calls as possible while
457 doing as few data copy as possible.
459 To improve it further, we could try to send all the sizes first and then
460 all the data (to use iov on receiving size), but it's only a partial
461 solution: when you have 2 dimensional data, the sizes of the second
462 dimension is data of the first dimension, so you need 3 streams.
464 I'm not sure the potential performance gains justify the coding burden.
466 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
468 SimGrid (3.00) stable; urgency=high
471 * New! Give the possibility to hijack the surf parser and thus bypass
472 MSG_create_environment and MSG_launch_application. Have a look at
473 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
475 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
477 SimGrid (2.96) unstable; urgency=low
482 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
483 This deprecates the xbt_error_t mecanisms.
484 It modifies (simplifies) all XBT and GRAS API.
485 MSG API keeps unchanged (exceptions raised by XBT are catched from
486 within MSG and masked with existing error handling facilities)
489 * New! Add a FATPIPE model. [AL]
490 * New! Add a parallel task model. [AL]
491 * New! Add automatically a loopback interface (in the default
492 network model) if none was precised.
495 * Bugfix: MSG_process_resume now works with the current running process.
497 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
498 * Modification of MSG_task_get_compute_duration. Once a task has been
499 processed, the value returned by this function is now equal to 0. [AL]
500 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
501 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
502 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
503 of a m_task. For now, MSG_task_cancel only works with computation
505 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
506 of the processor (in Mflop/s) regardless of the current load on the
508 * API Change: use proper naming convention for MSG_getClock and
509 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
511 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
512 This function changes the priority of a computation task. This priority
513 doesn't affect the transfer rate. A priority of 2 will make a task
514 receive two times more cpu power than the other ones. This function
515 has been added to suit the needs of Nguyen The Loc and hasn't been that
516 much tested yet. So if it fails, please report it and send me your code.
518 * API Change: removed all functions and types that were marked "deprecated"
519 since many months. Renamed MSG_global_init_args to MSG_global_init.
521 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
523 SimGrid (2.95) unstable; urgency=low
526 * Steal some nice code to GNU pth to fix context detection and usage [AL]
527 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
528 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
531 * Bugfix: Allow absolute paths for platform description files [MQ]
532 * Bugfix: do free the variables after use. Leads to drastic performance
534 * Implement max_duration (ie, timeouts) on resources [AL]
537 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
539 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
540 the needed support (timeouts) [AL]
543 * Implement measurement sockets. You can now get the bandwidth between two
544 hosts thanks to AMOK (see below). [MQ]
545 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
546 dynar over the network (yeah) [MQ]
547 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
548 * simple/static token-ring example. [Alexandre Colucci and MQ]
549 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
553 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
554 type parsing. "short" and "long" modifiers were also ignored; other
555 modifier (such as reference level) are still ignored. [MQ]
556 * Embeed the buffer size within the buffer itself on SG. [MQ]
557 That way, send() are atomic and cannot get intermixed anymore (at least
558 the ones which are less than 100k; bigger messages still have the issue)
559 * Array size pushed by the field, not by the field type (or each
560 and every long int will push stuff to the cbps) [MQ]
561 * use select() to sleep since it allows to portably sleep less than one
564 GRAS (minor cleanups)
565 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
568 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
569 * type category 'ignored' killed as it was never used and were difficult
571 * whether a type can cycle or not is now a flag, leaving room for more
572 flags (as "ignored", if we feel the need one day ;)
573 * Rename raw sockets to measurement sockets since "raw" has another
574 meaning in networking community.
577 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
578 and offers features not belonging to GRAS but that most applications
579 need. One day, it may be a set of plugins loadable at runtime.
580 * New module: bandwidth
581 bandwidth measurement between arbitrary nodes running this module. [MQ]
583 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
585 SimGrid (2.94) unstable; urgency=low
587 The first beta release of SimGrid 3 !
591 * Update the main page and the FAQ. Adding references to gforge.
594 * Add a gras_os_getpid function.
597 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
598 * Extend the logs so that they also print PID, hostname, date, ... if
600 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
601 and kill the old version which were in testsuite/
602 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
606 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
607 for users. Instead, each programming environment (GRAS, MSG,...) use it
608 when needed to provide such a feature to users.
609 Don't shortcut the mecanism or you will also shortcut the virtualization
610 you need on the simulator.
614 * Cleanups in configury with regard to compile optimization/warning flags.
615 Also add -fno-loop-optimize to any powerpc since it's the optimization
616 killing gcc (< 3.4.0).
617 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
618 needed by MSG examples complications
619 * Borrow configury beautifications from PHP
622 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
623 warning (thanks loris for stumbling into this one).
624 * Bugfix: stop loading private headers (gras_config.h) from the public
628 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
629 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
630 RL would imply a DNS resolver, which is *hard* to do in a portable way
631 (and therefore delayed).
632 * Implement a real timer mecanism and use it in timing macros. This allows
633 to avoid rounding errors and get a 0.000005 sec precision in timing
634 macros. While I was at it, various cleanups:
635 - allow to declare more than one timed section per file (fix a stupid bug)
636 - move some private declaration to the right place
637 - merge conditional execution and timing macros into emulation module
638 - document the module
639 - make sure the module cleanups its mess on gras_exit
640 * Documentation improvements:
641 - (new) how to compile applications using GRAS
642 - (new) emulation support (timing macros)
644 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
646 SimGrid (2.93) unstable; urgency=low
648 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
651 - Use Paje properly where used. Still to be sanitized properly.
652 - Portability fix: Add an implementation of the contexts using pthread
656 - Add xbt_procname(): returns the name of the current process.
657 Use it to show the current process's name in all logging.
659 - fix detection of older flex version and the reaction, since we do
660 depend on modern ones (we use lex_destroy)
661 - Better separation of SG and RL in the libs: remove all simulation code
662 from libgras. As a result, this lib is now only 200k when stripped.
663 Some of the xbt modules may also be duplicated (two sets and such) and
664 should be cleaned/killed before SG3.
665 - Insist on using xlC on AIX because of weird problems involving gcc there.
666 - Cleanup the make remote stuff. This is now done by scripts
667 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
668 mainly for our private use, but we're working on changing them to user
671 - Bugfix: flush the socket on close only if there is some *output*.
672 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
674 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
677 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
679 SimGrid (2.92) unstable; urgency=low
681 Alpha 3 on the path to SimGrid 3
685 - New! First try of benchmarking macros.
686 - New! First try so that gras_stub_generator generate deployment and
687 remote compilation helpers.
689 - Bugfix: Initialization fix in msg_test.
693 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
695 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
697 - API Change: gras_os_sleep to take the amount of seconds as a double.
698 Accepting an int was error prone since it was the only location where
699 seconds were coded as such. It leaded to damn rounding errors.
700 - Bugfix: Hard to belive that timers ever worked before this.
702 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
704 SimGrid (2.91) unstable; urgency=low
706 Alpha 2 on the path to SimGrid 3
710 - Bug fix in the lmm_solver.
712 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
713 through the function MSG_paje_output.
714 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
715 - It is possible to bound the rate of a communication in MSG with
716 MSG_task_put_bounded() (was already in the previous version but I had forgotten
717 to write it in the changelog).
718 - Bug fix to let GRAS run on top of MSG until we move it directly on top
723 - Various cleanups to the autotools stuff
724 - Begin to move Gras examples to examples/gras/
725 - Let make distcheck work again (yeah!)
727 - documentation overhauled using doxygen.
728 gtk-doc-tools is dead in SimGrid now.
729 - Automatically extract all existing logging categories, and add the list
730 to the documentation (long standing one, to say the less)
732 - Cleanup the known architecture table. Reorder the entries to group what
733 should be, and use a more consistent naming scheme.
734 (some of the test dataset are still to be regenerated)
735 - New! Allow library to register globals on each process just as userdata
737 This is implemented using a xbt_dict and not a xbt_set, so we loose the
738 lookup time (for now).
739 Use it in msg and trp.
740 This cleans a lot the internals and helps enforcing privacy of the
741 headers between the gras components.
742 - New! Add a timer mechanism, not unlike cron(8) and at(1).
743 - Bugfix: gras_os_time was delirious in RL.
744 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
746 - Reenable GRAS now that it works.
748 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
750 SimGrid (2.90) unstable; urgency=low
752 Alpha 1 on the path to SimGrid 3
754 * It is a long time since the last release of SimGrid. I'm sorry about
755 that but as I had told you, I was rewriting a lot of things. I apologize
756 to those who had been reporting bugs to me and that I had not answered.
757 If your bug is still in the new version, please tell me. Here is a
758 summary of the main changes.
760 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
761 lead by Martin Quinson. As a consequence SimGrid gains a lot in
762 portability, speed, and a lot more but you'll figure it out later.
763 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
764 I wanted to release the new MSG as soon as possible and I have
765 broken GRAS, which is the reason why, for now, only MSG is fully
766 functional. A laconic description of these projects is available
767 in the documentation.
769 * REVOLUTION 2: I have removed SG and I am now using a new simulation
770 kernel optimized for our needs (called SURF but only the developers
771 should use it). Hence, MSG is now roughly 30 times faster and I think
772 that by rewriting a little bit MSG, I could event speed it up a little
773 bit more. Beside the gain in speed, it is also much easier to encode a
774 new platform model with SURF than it was with SG. More to come...
776 * REVOLUTION 3: I have tried to change a little as possible the API of
777 MSG but a few things really had to disappear. The main differences
778 with the previous version are :
779 1) no more m_links_t and the corresponding functions. Platforms are
780 directly read from a XML description and cannot be hard-coded
781 anymore. The same format is used for application deployment
782 description. The new format is described in the documentation.
783 Have a look in tools/platform_generation. There is a tiny script
784 that converts from the old platform format to the new one. Concerning
785 the application deployment format, parsing the old one is tricky.
786 I think most of you should however be able to convert your files. If
787 it is really an issue, I can write a C code that does the conversion.
789 2) the toolbox tbx does not exist anymore. We now have a library
790 with much more data-structures but without the hash-tables (we have
791 dictionaries that are much faster).
793 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
795 *****************************************************************************
796 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
797 * don't feel like converting the oldies. (Mt) *
798 *****************************************************************************
801 Version 2.90: "the long awaited one"
802 - Finished rewriting and debugging MSG. Rewrote the documentation.
803 - disable GRAS for now since it needs to be ported to the newest SG
806 - Finish the port to windows (using mingw32 for cross-compile)
809 - Main loop and datastructures of SURF. A cpu resource object is
810 functional. Surf can thus be used to create cpu's with variable
811 performance on which you can execute some actions.
813 2004-11-15 Martin Quinson
814 - Port to ARM. Simply added the alignment and size descriptions. Should
815 work, but the ARM machines are so slow that I didn't had the opportunity
816 to 'make check' over there yet.
818 2004-11-15 Arnaud Legrand
819 - Trace manager now written. It uses a heap structure and is therefore
820 expected to be efficient. It may however be speeded up (particularly
821 when many events occur at the same date) by using red and black
822 trees. One day maybe...
823 - Max-min linear system solver written. It uses a sparse matrix
824 structure taking advantage of its expected use. Most operations are
825 O(1) and free/calloc are called as few as possible. The computation of
826 the minimum could however be improved by using a red and black tree
829 2004-11-03 Arnaud Legrand
830 - Rename every gras_* function that was in xbt/ to its xbt_
832 - Add a heap and a doubly-linked list to xbt
833 - Added a dichotomy to the dictionaries. make check works as well before
834 so I assume that the patch is correct. I do not know however if things
835 run effectively faster than before now. :)
837 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
838 SimGrid, and the version number is bumped to 2.x
840 2004-10-29 Martin Quinson
841 - Introduction of the remote errors.
842 They are the result of a RMI/RPC on the remote machine.
843 ErrCodes being scalar values, you can't get the host on which those
844 errors did happen. Extending the error mechanism as in Gnome is possible.
845 No idea yet whether it is a good idea.
847 2004-10-28 Martin Quinson
848 - Interface revolution: the Starred Structure Eradication.
849 I used to do typedef struct {} toto_t; and then handle *toto_t.
850 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
851 - ???_t is a valid type (builded with typedef)
852 - s_toto_t is a structure (access to fields with .)
853 - s_toto is a structure needing 'struct' keyword to be used
854 - e_toto_t is an enum
855 - toto_t is an 'object' (struct*)
857 typedef struct s_toto {} s_toto_t, *toto_t;
858 typedef enum {} e_toto_t;
859 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
860 s_toto_t) is private.
862 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
863 so that it can set the variable to NULL. It was so for dicts and sets,
864 it changed for dynars.
866 - Fix a bunch of memleaks in dict_remove
867 - Fix a bug in sg/server_socket opening: it failed all the time.
869 2004-10-07 Martin Quinson
870 - Speed up dynar lookup operation a bit.
872 gras_dynar_get is dead.
874 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
875 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
876 (which gives you the address of the stored data).
878 gras_dynar_get_as is an helpful macro which allows you to retrieve a
879 copy of the data using an affectation to do the job and not a memcpy.
881 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
882 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
884 It does not really speedup the dynar test because they are
885 setting elements all the time (and look them seldom). But the dict does
886 far more lookup than setting.
888 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
890 2004-10-05 Martin Quinson
891 - Allow to (en/dis)able the cycle detection at run time.
893 Whether we should check for cycle or not is now a property of each
894 datatype. When you think there may be some cycle, use datadesc_cycle_set.
895 datadesc_cycle_unset allow to remove this property when previously set.
897 Note that the cycle detection is off by default since it impacts the
898 performance. Watch the data you feed GRAS with ;)
900 This property is hereditary. Any element embedded in a structure having it
901 set have it set for the time of this data exchange.
903 You should set it both on sender and receiver side. If you don't set it on
904 sender side, it will enter an endless loop. If you forget on receiver
905 side, the cycles won't be recreated after communication.
907 - Header reorganization.
908 Kill gras_private.h, each submodule must load the headers it needs.
910 2004-10-04 Martin Quinson
911 - Interface revolution: do not try to survive to malloc failure.
913 Now, gras_malloc and friends call gras_abort() on failure.
914 As a conclusion, malloc_error is not a valid error anymore, and all
915 functions for which it was the only gras_error_t return value are
916 changed. They now return void, or there result directly.
917 This simplify the API a lot.
919 2004-09-29 Martin Quinson
920 - Re-enable raw sockets.
921 Created by gras_socket_{client,server}_ext;
922 Used with gras_raw_{send,recv}
925 It should allow to kill the last bits of gras first version soon.
927 This is not completely satisfactory yet (duplicate code with
928 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
931 - Simplify transport plugin (internal) interface by not passing any
932 argument to _server and _client, but embedding them in the socket
935 2004-09-28 Martin Quinson
936 - Finish the port to AIX.
937 autoconf was my problem (segfault within the malloc replacement
938 function. No idea why)
940 2004-09-16 Martin Quinson
941 - Fix some size_t madness on 64bit architectures.
943 2004-09-08 Martin Quinson
944 - Reduce the number of system headers loaded, overload some more system
945 calls (such as malloc to cast the result of the system one, and work
947 - Fix and reintroduce the config support
949 2004-09-07 Martin Quinson
950 - Source code reorganization to allow Arnaud to surf all over there.
951 - Allow to document the logging categories.
952 - Remove all uppercase from logging categories and useless cleanup in names.
954 2004-08-18 Martin Quinson
955 Version 0.6.2 (protocol not changed; API changed)
956 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
957 gras_error_t), and NULL when not found. Functions expecting a msgtype
958 as argument (msg_wait; msg_send) deal with NULL argument by providing a
959 hopefully usefull message.
960 - Portability to prehistoric sparcs again
962 2004-08-17 Martin Quinson
963 Version 0.6.1 (protocol not changed; ABI not changed)
964 - prealloc some buffers to speed things up
966 2004-08-11 Martin Quinson
967 Version 0.6 (protocol not changed; ABI expended)
968 - The parsing macro can deal with the references, provided that you add
969 the relevant annotations (using GRAS_ANNOTE(size,field_name))
971 2004-08-09 Martin Quinson
972 Version 0.5 (protocol not changed; ABI changed)
973 - Allow to off turn the cycle detection code in data exchange at
974 compilation time. It should be at run time, but I'm short of time (and
975 the config stuff is still broken). That way, we keep dict out of the
976 critical path, which is good because the performance is poor:
977 - search not dichotomial yet
978 - dynar give no way to access their content and memcpy everytime
979 - In composed data description (struct, ref and so on), stop foolness of
980 keeping the subtype's ID, but store the type itself. This keeps sets out
981 of the critical path, which is good since they rely on dynar and
982 dictionnaries. The only loose of that is that we cannot detect the
983 redeclaration of a structure/union with another content (but I'm not sure
984 the code detected well this error before anyway). We still can detect
985 the redefinition discrepancy for the other types.
986 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
987 it breaks the code because of type-punning used all over the place).
988 This breaks on all non-gcc architectures (for now).
990 All those changes (plus the buffer of last time) allow me to gain 2 order
991 of magnitude on cruel tests consisting of 800000 array of integers on two
992 level of a hierarchical structure (200 secondes -> 4 secondes)
995 - the selector of reference must now return the type it points to, not
998 2004-08-06 Martin Quinson
999 Version 0.4 (protocol changed; ABI not changed)
1000 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1001 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1002 - Implement a buffer transport, and use it by default (it relies on tcp in
1003 real life and on sg in simulation).
1004 That's a bit hackish since I had a new field to the structure to store
1005 its data without interfering with the subtype ones. Inheritance
1006 is tricky in C. And that's a kind of reverse inheritance with one class
1007 derivating two classes. Or maybe a game with java interfaces. Anyway,
1008 that's damn hard in C (at least).
1009 Moreover, I got tired while trying to ensure plugin separation and
1010 genericity in SG mode. MSG wants me to do weird things, so let's go for
1011 cruel hacks (temporarily of course ;).
1012 See comment in transport_private.h:71
1013 - do not include all the _interface headers in private but in the files
1014 which really need them (to cut the compilation time when they are
1017 2004-07-26 Martin Quinson
1018 Version 0.3 (protocol not changed; ABI changed)
1019 - Major overhault of the datadesc interface to simplify it:
1020 - shorted the function names:
1021 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1022 - add a trivial way to push/pop integers into the cbps without malloc.
1023 This allows to make really generic sub_type description, which simply
1024 pop their size of the stack.
1025 - add a function gras_datadesc_ref_pop_arr() which does what users want
1026 most of the time: Declare a dynamic array (which pops its size of the
1027 stack) and declare a reference to it. Poor name, but anyway.
1028 - kill the post-send callback, add a post-receive one
1030 2004-07-23 Martin Quinson
1031 Version 0.2 (protocol changed; ABI changed)
1032 - add some testing for cpbs in the test cases, and fix some more bugs.
1033 This invalidate again the little64 data file, since I cannot regenerate
1035 - remove an awfull optimization in the logging stuff, allowing me to:
1036 - understand it again
1037 - learn gcc how to check that the argument match the provided format
1038 - fix all errors revealed by gcc after that
1039 - internal keys of dict are not \0 terminated. Deal with it properly in
1040 loggings instead of segfaulting when the user want to see the logs :-/
1042 2004-07-22 Martin Quinson
1043 - Fix some stupid bug preventing cbps (callback postit) from working
1045 2004-07-21 Martin Quinson
1046 - Some documentation cleanups
1047 - remove the useless last argument of msgtype_declare
1048 - rename the Virtu functions to fit into the 'os' namespace
1049 - move headers src/include -> src/include/gras/ and stop fooling with
1050 gras -> . symbolic link
1051 - make distcheck is now successful
1053 2004-07-19 Martin Quinson
1055 - Build shared library also
1056 - Install html doc to the right location
1057 - stop removing maintainer files in make clean
1058 - build tests only on make check
1060 2004-07-13 Martin Quinson
1062 - No major issue in previous version => change versionning schema
1063 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1064 corresponding dataset.
1066 2004-07-11 Martin Quinson
1068 - Get it working with any kind of structure (we can compute the padding
1069 bytes remotely for all the architectures I have access to)
1070 - Implement the structure parsing macro (still not quite robust/complete)
1071 - Improvement to the remote testing toysuite
1073 2004-07-10 Martin Quinson
1074 [autoconf mechanism]
1075 - get ride of a bunch of deprecated macros
1076 - actually run the test for two-compliment, not only compile it :-/
1077 - test whether the structures get packed (and bail out if yes. Damn.
1078 Alignment is a serious matter)
1079 - test whether the structures get compacted (but respecting the alignment
1080 constraints of each types)
1081 - test whether the array fields of structures can straddle alignment boundaries
1083 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1084 (took me 2 hours to find that bug, looking at the wrong place)
1085 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1086 before sending/receiving and are used to compute the offsets of fields
1087 - Given that padding size depend even on compiler options, keep track of
1088 alignment and aligned_size only for the current architecture. Not a big
1089 deal since we send structure fields one after the other (seems
1091 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1092 program, let it work on linux/gcc/little32. portability todo.
1094 2004-07-08 Martin Quinson
1095 - import and improve remote compilation support from FAST
1096 - make sure make check works on half a dozen of machines out there
1098 2004-07-07 Martin Quinson
1099 Let's say it's version 0.0.3 ;)
1100 - Implement conversions (yuhu!)
1101 - Let it work on solaris (beside conversion, of course)
1102 - Stupid me, using rand() to generate the conversion datatests in not wise.
1104 2004-07-06 Martin Quinson
1105 - Let make dist work, since I'm gonna need it to compile on remote hosts
1106 - Let Tests/datadesc_usage write the architecture on which the file was
1107 generated as first byte.
1108 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1109 architecture descriptions.
1110 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1111 on those architectures.
1112 - Optimization: send/recv array of scalar in one shoot
1114 2004-07-05 Martin Quinson
1115 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1117 - Plug a whole bunch of memleaks
1118 - each process now have to call gras_{init,exit}. One day, their log
1119 settings will be separated
1120 - Continue the code factorisation between SG, RL and common in Transport.
1122 2004-07-04 Martin Quinson
1124 - Redistribution between SG and RL.
1125 We wanna have to accept in SG, so move accepted related parts of RL in
1126 the common part. (more precisely, the dynar of all known sockets is no
1127 more a static in transport.c, but part of the process_data)
1129 [gras_stub_generator]
1130 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1132 2004-07-03 Martin Quinson
1133 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1134 them all up in one shot)
1136 - Fix the ugly case of reference to dynamic array.
1137 - New (semi-public) function gras_datadesc_size to allow the messaging
1138 layer to malloc the needed space for the buffer.
1140 - gras_socket_close now expect the socket to close (and not its address to
1141 put NULL in it after it). This is because the socket passed to handlers
1142 is one of their argument (=> not writable).
1144 - propagate the interface cleanup from last week in datadesc, ie remove a
1145 superfluous level of indirection. User pass adress of variable
1146 containing data (both when sending and receiving), and not of a variable
1147 being a pointer to the data. Let's say that I like it better ;)
1148 The price for that is constructs like "int msg=*(int*)payload" in
1149 handlers, but it's a fine price, IMHO.
1151 - Let it work in RL (yuhu)
1153 2004-06-21 Martin Quinson
1155 - porting SG plugin and SG select to new standards (works almost).
1156 - plug memleaks and fix bugs around.
1159 - cleanup the prototype of data recv and force users to specify when they
1160 want to handle references to objects. Test case working even for cycles.
1161 - plug memleaks. Valgrind is perfectly ok with this.
1163 2004-06-12 Martin Quinson
1165 - cleanup the separation between plugin and main code in plugin creation
1167 2004-06-11 Martin Quinson
1169 - Reput hook for raw sockets, needed for BW experiments
1170 - kill a few lines of dead code
1171 [Data description] Interface cleanup
1172 - gras_datadesc_by_name returns the searched type or NULL.
1173 That way, no variable is needed to use a type desc once, which makes
1175 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1176 two parameters were strings (field name, type name), leading to
1178 [Dicos] Interface cleanup
1179 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1180 This is consistant with the dynar API.
1182 2004-04-21 Martin Quinson
1184 - Porting to new standards.
1186 - interface cleanup.
1187 There is no bag anymore, no need to take extra provision to mask the
1188 pointers behind "ID".
1189 Better splitup of functions between files create/exchange/convert.
1190 This is still a bit artificial since convert and receive are so
1191 interleaved, but anyway.
1193 - add a queued message list to procdata (the ones not matching criteria
1195 - factorize some more code between SG and RL wrt procdata
1197 - use gras_exit in example to track memleaks
1198 - get rid of gs_example now that GS is properly integrated into gras
1199 - update run_test to integrate the lastest tests (datadesc)
1201 - rename WARNINGn macros to WARNn since it prooved error-prone
1203 2004-04-19 Martin Quinson
1205 - register init/exit functions within gras module mechanism
1206 - send/receive function.
1207 Convertion is not implemented, but short-cutted if not needed.
1208 struct/array elements are sent one by one (instead of block-wise), but
1209 nobody really cares (yet). Get a prototype before optimizing.
1210 - tests (using a file socket) for DD send/receive on:
1211 - base types: int, float
1212 - array: fixed size, string (ie ref to dynamic string)
1213 - structure: homogeneous, heterogeneous
1214 - chained list, graph with cycle
1215 Believe it or not, valgrind is not too unhappy with the results. The
1216 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1220 - Bugfix when using a filename explicitely (instead of '-')
1222 2004-04-09 Martin Quinson
1224 - factorize more code between RL and SG in socket creation
1225 - Complete the implementation and tests of:
1227 o file (only in RL, and mainly for debugging)
1229 I lost 3 days to design a portable address resolver, and then decided
1230 that the prototype mainly have to run on my box.
1231 Addressing portability too early may be like optimizing too early :-/
1233 - use gras_init in the Tests instead of the crappy parse_log_opt
1234 (the latter function is removed)
1235 [Conditional execution]
1236 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1237 [Code reorganisation]
1238 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1239 Build examples against the RL library, since there is no way to disable
1240 its creation for now.
1242 For information, the beginning of coding on GRAS was back in june
1243 2003. I guess that every line has been rewritten at least twice since