1 2004-11-03 Arnaud Legrand
2 - Rename every gras_* function that was in xbt/ to its xbt_
4 - Add a heap and a doubly-linked list to xbt
6 2004-10-29 Martin Quinson
7 - Introduction of the remote errors.
8 They are the result of a RMI/RPC on the remote machine.
9 ErrCodes being scalar values, you can't get the host on which those
10 errors did happen. Extending the error mecanism as in Gnome is possible.
11 No idea yet whether it is a good idea.
13 2004-10-28 Martin Quinson
14 - Interface revolution: the Starred Structure Eradication.
15 I used to do typedef struct {} toto_t; and then handle *toto_t.
16 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
17 - ???_t is a valid type (builded with typedef)
18 - s_toto_t is a structure (access to fields with .)
19 - s_toto is a structure needing 'struct' keyword to be used
21 - toto_t is an 'object' (struct*)
23 typedef struct s_toto {} s_toto_t, *toto_t;
24 typedef enum {} e_toto_t;
25 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
28 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
29 so that it can set the variable to NULL. It was so for dicts and sets,
30 it changed for dynars.
32 - Fix a bunch of memleaks in dict_remove
33 - Fix a bug in sg/server_socket opening: it failed all the time.
35 2004-10-07 Martin Quinson
36 - Speed up dynar lookup operation a bit.
38 gras_dynar_get is dead.
40 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
41 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
42 (which gives you the address of the stored data).
44 gras_dynar_get_as is an helpful macro which allows you to retrieve a
45 copy of the data using an affectation to do the job and not a memcpy.
47 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
48 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
50 It does not really speedup the dynar test because they are
51 setting elements all the time (and look them seldom). But the dict does
52 far more lookup than setting.
54 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
56 2004-10-05 Martin Quinson
57 - Allow to (en/dis)able the cycle detection at run time.
59 Whether we should check for cycle or not is now a property of each
60 datatype. When you think there may be some cycle, use datadesc_cycle_set.
61 datadesc_cycle_unset allow to remove this property when previously set.
63 Note that the cycle detection is off by default since it impacts the
64 performance. Watch the data you feed GRAS with ;)
66 This property is hereditary. Any element embeeded in a structure having it
67 set have it set for the time of this data exchange.
69 You should set it both on sender and receiver side. If you don't set it on
70 sender side, it will enter an endless loop. If you forget on receiver
71 side, the cycles won't be recreated after communication.
73 - Header reorganization.
74 Kill gras_private.h, each submodule must load the headers it needs.
76 2004-10-04 Martin Quinson
77 - Interface revolution: do not try to survive to malloc failure.
79 Now, gras_malloc and friends call gras_abort() on failure.
80 As a conclusion, malloc_error is not a valid error anymore, and all
81 functions for which it was the only gras_error_t return value are
82 changed. They now return void, or there result directly.
83 This simplify the API a lot.
85 2004-09-29 Martin Quinson
86 - Re-enable raw sockets.
87 Created by gras_socket_{client,server}_ext;
88 Used with gras_raw_{send,recv}
91 It should allow to kill the last bits of gras first version soon.
93 This is not completely satisfactory yet (dupplicate code with
94 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
97 - Simplify transport plugin (internal) interface by not passing any
98 argument to _server and _client, but embeeding them in the socket
101 2004-09-28 Martin Quinson
102 - Finish the port to AIX.
103 autoconf was my problem (segfault within the malloc replacement
104 function. No idea why)
106 2004-09-16 Martin Quinson
107 - Fix some size_t madness on 64bit architectures.
109 2004-09-08 Martin Quinson
110 - Reduce the number of system headers loaded, overload some more system
111 calls (such as malloc to cast the result of the system one, and work
113 - Fix and reintroduce the config support
115 2004-09-07 Martin Quinson
116 - Source code reorganization to allow Arnaud to surf all over there.
117 - Allow to document the logging categories.
118 - Remove all uppercase from logging categories and useless cleanup in names.
120 2004-08-18 Martin Quinson
121 Version 0.6.2 (protocol not changed; API changed)
122 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
123 gras_error_t), and NULL when not found. Functions expecting a msgtype
124 as argument (msg_wait; msg_send) deal with NULL argument by providing a
125 hopefully usefull message.
126 - Portability to prehistoric sparcs again
128 2004-08-17 Martin Quinson
129 Version 0.6.1 (protocol not changed; ABI not changed)
130 - prealloc some buffers to speed things up
132 2004-08-11 Martin Quinson
133 Version 0.6 (protocol not changed; ABI expended)
134 - The parsing macro can deal with the references, provided that you add
135 the relevant annotations (using GRAS_ANNOTE(size,field_name))
137 2004-08-09 Martin Quinson
138 Version 0.5 (protocol not changed; ABI changed)
139 - Allow to off turn the cycle detection code in data exchange at
140 compilation time. It should be at run time, but I'm short of time (and
141 the config stuff is still broken). That way, we keep dict out of the
142 critical path, which is good because the performance is poor:
143 - search not dichotomial yet
144 - dynar give no way to access their content and memcpy everytime
145 - In composed data description (struct, ref and so on), stop foolness of
146 keeping the subtype's ID, but store the type itself. This keeps sets out
147 of the critical path, which is good since they rely on dynar and
148 dictionnaries. The only loose of that is that we cannot detect the
149 redeclaration of a structure/union with another content (but I'm not sure
150 the code detected well this error before anyway). We still can detect
151 the redefinition discrepancy for the other types.
152 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
153 it breaks the code because of type-punning used all over the place).
154 This breaks on all non-gcc architectures (for now).
156 All those changes (plus the buffer of last time) allow me to gain 2 order
157 of magnitude on cruel tests consisting of 800000 array of integers on two
158 level of a hierarchical structure (200 secondes -> 4 secondes)
161 - the selector of reference must now return the type it points to, not
164 2004-08-06 Martin Quinson
165 Version 0.4 (protocol changed; ABI not changed)
166 - Allow to pass --gras-log argument to processes in simulation mode. Really.
167 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
168 - Implement a buffer transport, and use it by default (it relies on tcp in
169 real life and on sg in simulation).
170 That's a bit hackish since I had a new field to the structure to store
171 its data without interfering with the subtype ones. Inheritance
172 is tricky in C. And that's a kind of reverse inheritance with one class
173 derivating two classes. Or maybe a game with java interfaces. Anyway,
174 that's damn hard in C (at least).
175 Moreover, I got tired while trying to ensure plugin separation and
176 genericity in SG mode. MSG wants me to do weird things, so let's go for
177 cruel hacks (temporarily of course ;).
178 See comment in transport_private.h:71
179 - do not include all the _interface headers in private but in the files
180 which really need them (to cut the compilation time when they are
183 2004-07-26 Martin Quinson
184 Version 0.3 (protocol not changed; ABI changed)
185 - Major overhault of the datadesc interface to simplify it:
186 - shorted the function names:
187 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
188 - add a trivial way to push/pop integers into the cbps without malloc.
189 This allows to make really generic sub_type description, which simply
190 pop their size of the stack.
191 - add a function gras_datadesc_ref_pop_arr() which does what users want
192 most of the time: Declare a dynamic array (which pops its size of the
193 stack) and declare a reference to it. Poor name, but anyway.
194 - kill the post-send callback, add a post-receive one
196 2004-07-23 Martin Quinson
197 Version 0.2 (protocol changed; ABI changed)
198 - add some testing for cpbs in the test cases, and fix some more bugs.
199 This invalidate again the little64 data file, since I cannot regenerate
201 - remove an awfull optimization in the logging stuff, allowing me to:
202 - understand it again
203 - learn gcc how to check that the argument match the provided format
204 - fix all errors revealed by gcc after that
205 - internal keys of dict are not \0 terminated. Deal with it properly in
206 loggings instead of segfaulting when the user want to see the logs :-/
208 2004-07-22 Martin Quinson
209 - Fix some stupid bug preventing cbps (callback postit) from working
211 2004-07-21 Martin Quinson
212 - Some documentation cleanups
213 - remove the useless last argument of msgtype_declare
214 - rename the Virtu functions to fit into the 'os' namespace
215 - move headers src/include -> src/include/gras/ and stop fooling with
216 gras -> . symbolic link
217 - make distcheck is now successful
219 2004-07-19 Martin Quinson
221 - Build shared library also
222 - Install html doc to the right location
223 - stop removing maintainer files in make clean
224 - build tests only on make check
226 2004-07-13 Martin Quinson
228 - No major issue in previous version => change versionning schema
229 - Re-enable little64 convertion test now that Abdou kindly regenerated the
230 corresponding dataset.
232 2004-07-11 Martin Quinson
234 - Get it working with any kind of structure (we can compute the padding
235 bytes remotely for all the architectures I have access to)
236 - Implement the structure parsing macro (still not quite robust/complete)
237 - Improvement to the remote testing toysuite
239 2004-07-10 Martin Quinson
241 - get ride of a bunch of deprecated macros
242 - actually run the test for two-compliment, not only compile it :-/
243 - test whether the structures get packed (and bail out if yes. Damn.
244 Alignment is a serious matter)
245 - test whether the structures get compacted (but respecting the alignment
246 constraints of each types)
247 - test whether the array fields of structures can straddle alignment boundaries
249 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
250 (took me 2 hours to find that bug, looking at the wrong place)
251 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
252 before sending/receiving and are used to compute the offsets of fields
253 - Given that padding size depend even on compiler options, keep track of
254 alignment and aligned_size only for the current architecture. Not a big
255 deal since we send structure fields one after the other (seems
257 - Add the datastructure used for IEEE paper by the PBIO guys to the test
258 program, let it work on linux/gcc/little32. portability todo.
260 2004-07-08 Martin Quinson
261 - import and improve remote compilation support from FAST
262 - make sure make check works on half a dozen of machines out there
264 2004-07-07 Martin Quinson
265 Let's say it's version 0.0.3 ;)
266 - Implement conversions (yuhu!)
267 - Let it work on solaris (beside conversion, of course)
268 - Stupid me, using rand() to generate the conversion datatests in not wise.
270 2004-07-06 Martin Quinson
271 - Let make dist work, since I'm gonna need it to compile on remote hosts
272 - Let Tests/datadesc_usage write the architecture on which the file was
273 generated as first byte.
274 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
275 architecture descriptions.
276 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
277 on those architectures.
278 - Optimization: send/recv array of scalar in one shoot
280 2004-07-05 Martin Quinson
281 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
283 - Plug a whole bunch of memleaks
284 - each process now have to call gras_{init,exit}. One day, their log
285 settings will be separated
286 - Continue the code factorisation between SG, RL and common in Transport.
288 2004-07-04 Martin Quinson
290 - Redistribution between SG and RL.
291 We wanna have to accept in SG, so move accepted related parts of RL in
292 the common part. (more precisely, the dynar of all known sockets is no
293 more a static in transport.c, but part of the process_data)
295 [gras_stub_generator]
296 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
298 2004-07-03 Martin Quinson
299 - Create a new log channel tbx containing dict, set, log, dynar (to shut
300 them all up in one shot)
302 - Fix the ugly case of reference to dynamic array.
303 - New (semi-public) function gras_datadesc_size to allow the messaging
304 layer to malloc the needed space for the buffer.
306 - gras_socket_close now expect the socket to close (and not its address to
307 put NULL in it after it). This is because the socket passed to handlers
308 is one of their argument (=> not writable).
310 - propagate the interface cleanup from last week in datadesc, ie remove a
311 superfluous level of indirection. User pass adress of variable
312 containing data (both when sending and receiving), and not of a variable
313 being a pointer to the data. Let's say that I like it better ;)
314 The price for that is constructs like "int msg=*(int*)payload" in
315 handlers, but it's a fine price, IMHO.
317 - Let it work in RL (yuhu)
319 2004-06-21 Martin Quinson
321 - porting SG plugin and SG select to new standards (works almost).
322 - plug memleaks and fix bugs around.
325 - cleanup the prototype of data recv and force users to specify when they
326 want to handle references to objects. Test case working even for cycles.
327 - plug memleaks. Valgrind is perfectly ok with this.
329 2004-06-12 Martin Quinson
331 - cleanup the separation between plugin and main code in plugin creation
333 2004-06-11 Martin Quinson
335 - Reput hook for raw sockets, needed for BW experiments
336 - kill a few lines of dead code
337 [Data description] Interface cleanup
338 - gras_datadesc_by_name returns the searched type or NULL.
339 That way, no variable is needed to use a type desc once, which makes
341 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
342 two parameters were strings (field name, type name), leading to
344 [Dicos] Interface cleanup
345 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
346 This is consistant with the dynar API.
348 2004-04-21 Martin Quinson
350 - Porting to new standards.
353 There is no bag anymore, no need to take extra provision to mask the
354 pointers behind "ID".
355 Better splitup of functions between files create/exchange/convert.
356 This is still a bit artificial since convert and receive are so
357 interleaved, but anyway.
359 - add a queued message list to procdata (the ones not matching criteria
361 - factorize some more code between SG and RL wrt procdata
363 - use gras_exit in example to track memleaks
364 - get rid of gs_example now that GS is properly integrated into gras
365 - update run_test to integrate the lastest tests (datadesc)
367 - rename WARNINGn macros to WARNn since it prooved error-prone
369 2004-04-19 Martin Quinson
371 - register init/exit functions within gras module mecanism
372 - send/receive function.
373 Convertion is not implemented, but short-cutted if not needed.
374 struct/array elements are sent one by one (instead of block-wise), but
375 nobody really cares (yet). Get a prototype before optimizing.
376 - tests (using a file socket) for DD send/receive on:
377 - base types: int, float
378 - array: fixed size, string (ie ref to dynamic string)
379 - structure: homogeneous, heterogeneous
380 - chained list, graph with cycle
381 Believe it or not, valgrind is not too unhappy with the results. The
382 cycle happily segfaults, but the others are ok. And I'm sick of pointers
386 - Bugfix when using a filename explicitely (instead of '-')
388 2004-04-09 Martin Quinson
390 - factorize more code between RL and SG in socket creation
391 - Complete the implementation and tests of:
393 o file (only in RL, and mainly for debugging)
395 I lost 3 days to design a portable address resolver, and then decided
396 that the prototype mainly have to run on my box.
397 Addressing portability too early may be like optimizing too early :-/
399 - use gras_init in the Tests instead of the crappy parse_log_opt
400 (the latter function is removed)
401 [Conditional execution]
402 - New functions: gras_if_RL/gras_if_SG (basic support for this)
403 [Code reorganisation]
404 - Get rid of libgrasutils.a since it makes more trouble than it solves.
405 Build examples against the RL library, since there is no way to disable
406 its creation for now.