1 2004-10-29 Martin Quinson
2 - Introduction of the remote errors.
3 They are the result of a RMI/RPC on the remote machine.
4 ErrCodes being scalar values, you can't get the host on which those
5 errors did happen. Extending the error mecanism as in Gnome is possible.
6 No idea yet whether it is a good idea.
8 2004-10-28 Martin Quinson
9 - Interface revolution: the Starred Structure Eradication.
10 I used to do typedef struct {} toto_t; and then handle *toto_t.
11 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
12 - ???_t is a valid type (builded with typedef)
13 - s_toto_t is a structure (access to fields with .)
14 - s_toto is a structure needing 'struct' keyword to be used
16 - toto_t is an 'object' (struct*)
18 typedef struct s_toto {} s_toto_t, *toto_t;
19 typedef enum {} e_toto_t;
20 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
23 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
24 so that it can set the variable to NULL. It was so for dicts and sets,
25 it changed for dynars.
27 - Fix a bunch of memleaks in dict_remove
28 - Fix a bug in sg/server_socket opening: it failed all the time.
30 2004-10-07 Martin Quinson
31 - Speed up dynar lookup operation a bit.
33 gras_dynar_get is dead.
35 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
36 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
37 (which gives you the address of the stored data).
39 gras_dynar_get_as is an helpful macro which allows you to retrieve a
40 copy of the data using an affectation to do the job and not a memcpy.
42 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
43 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
45 It does not really speedup the dynar test because they are
46 setting elements all the time (and look them seldom). But the dict does
47 far more lookup than setting.
49 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
51 2004-10-05 Martin Quinson
52 - Allow to (en/dis)able the cycle detection at run time.
54 Whether we should check for cycle or not is now a property of each
55 datatype. When you think there may be some cycle, use datadesc_cycle_set.
56 datadesc_cycle_unset allow to remove this property when previously set.
58 Note that the cycle detection is off by default since it impacts the
59 performance. Watch the data you feed GRAS with ;)
61 This property is hereditary. Any element embeeded in a structure having it
62 set have it set for the time of this data exchange.
64 You should set it both on sender and receiver side. If you don't set it on
65 sender side, it will enter an endless loop. If you forget on receiver
66 side, the cycles won't be recreated after communication.
68 - Header reorganization.
69 Kill gras_private.h, each submodule must load the headers it needs.
71 2004-10-04 Martin Quinson
72 - Interface revolution: do not try to survive to malloc failure.
74 Now, gras_malloc and friends call gras_abort() on failure.
75 As a conclusion, malloc_error is not a valid error anymore, and all
76 functions for which it was the only gras_error_t return value are
77 changed. They now return void, or there result directly.
78 This simplify the API a lot.
80 2004-09-29 Martin Quinson
81 - Re-enable raw sockets.
82 Created by gras_socket_{client,server}_ext;
83 Used with gras_raw_{send,recv}
86 It should allow to kill the last bits of gras first version soon.
88 This is not completely satisfactory yet (dupplicate code with
89 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
92 - Simplify transport plugin (internal) interface by not passing any
93 argument to _server and _client, but embeeding them in the socket
96 2004-09-28 Martin Quinson
97 - Finish the port to AIX.
98 autoconf was my problem (segfault within the malloc replacement
99 function. No idea why)
101 2004-09-16 Martin Quinson
102 - Fix some size_t madness on 64bit architectures.
104 2004-09-08 Martin Quinson
105 - Reduce the number of system headers loaded, overload some more system
106 calls (such as malloc to cast the result of the system one, and work
108 - Fix and reintroduce the config support
110 2004-09-07 Martin Quinson
111 - Source code reorganization to allow Arnaud to surf all over there.
112 - Allow to document the logging categories.
113 - Remove all uppercase from logging categories and useless cleanup in names.
115 2004-08-18 Martin Quinson
116 Version 0.6.2 (protocol not changed; API changed)
117 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
118 gras_error_t), and NULL when not found. Functions expecting a msgtype
119 as argument (msg_wait; msg_send) deal with NULL argument by providing a
120 hopefully usefull message.
121 - Portability to prehistoric sparcs again
123 2004-08-17 Martin Quinson
124 Version 0.6.1 (protocol not changed; ABI not changed)
125 - prealloc some buffers to speed things up
127 2004-08-11 Martin Quinson
128 Version 0.6 (protocol not changed; ABI expended)
129 - The parsing macro can deal with the references, provided that you add
130 the relevant annotations (using GRAS_ANNOTE(size,field_name))
132 2004-08-09 Martin Quinson
133 Version 0.5 (protocol not changed; ABI changed)
134 - Allow to off turn the cycle detection code in data exchange at
135 compilation time. It should be at run time, but I'm short of time (and
136 the config stuff is still broken). That way, we keep dict out of the
137 critical path, which is good because the performance is poor:
138 - search not dichotomial yet
139 - dynar give no way to access their content and memcpy everytime
140 - In composed data description (struct, ref and so on), stop foolness of
141 keeping the subtype's ID, but store the type itself. This keeps sets out
142 of the critical path, which is good since they rely on dynar and
143 dictionnaries. The only loose of that is that we cannot detect the
144 redeclaration of a structure/union with another content (but I'm not sure
145 the code detected well this error before anyway). We still can detect
146 the redefinition discrepancy for the other types.
147 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
148 it breaks the code because of type-punning used all over the place).
149 This breaks on all non-gcc architectures (for now).
151 All those changes (plus the buffer of last time) allow me to gain 2 order
152 of magnitude on cruel tests consisting of 800000 array of integers on two
153 level of a hierarchical structure (200 secondes -> 4 secondes)
156 - the selector of reference must now return the type it points to, not
159 2004-08-06 Martin Quinson
160 Version 0.4 (protocol changed; ABI not changed)
161 - Allow to pass --gras-log argument to processes in simulation mode. Really.
162 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
163 - Implement a buffer transport, and use it by default (it relies on tcp in
164 real life and on sg in simulation).
165 That's a bit hackish since I had a new field to the structure to store
166 its data without interfering with the subtype ones. Inheritance
167 is tricky in C. And that's a kind of reverse inheritance with one class
168 derivating two classes. Or maybe a game with java interfaces. Anyway,
169 that's damn hard in C (at least).
170 Moreover, I got tired while trying to ensure plugin separation and
171 genericity in SG mode. MSG wants me to do weird things, so let's go for
172 cruel hacks (temporarily of course ;).
173 See comment in transport_private.h:71
174 - do not include all the _interface headers in private but in the files
175 which really need them (to cut the compilation time when they are
178 2004-07-26 Martin Quinson
179 Version 0.3 (protocol not changed; ABI changed)
180 - Major overhault of the datadesc interface to simplify it:
181 - shorted the function names:
182 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
183 - add a trivial way to push/pop integers into the cbps without malloc.
184 This allows to make really generic sub_type description, which simply
185 pop their size of the stack.
186 - add a function gras_datadesc_ref_pop_arr() which does what users want
187 most of the time: Declare a dynamic array (which pops its size of the
188 stack) and declare a reference to it. Poor name, but anyway.
189 - kill the post-send callback, add a post-receive one
191 2004-07-23 Martin Quinson
192 Version 0.2 (protocol changed; ABI changed)
193 - add some testing for cpbs in the test cases, and fix some more bugs.
194 This invalidate again the little64 data file, since I cannot regenerate
196 - remove an awfull optimization in the logging stuff, allowing me to:
197 - understand it again
198 - learn gcc how to check that the argument match the provided format
199 - fix all errors revealed by gcc after that
200 - internal keys of dict are not \0 terminated. Deal with it properly in
201 loggings instead of segfaulting when the user want to see the logs :-/
203 2004-07-22 Martin Quinson
204 - Fix some stupid bug preventing cbps (callback postit) from working
206 2004-07-21 Martin Quinson
207 - Some documentation cleanups
208 - remove the useless last argument of msgtype_declare
209 - rename the Virtu functions to fit into the 'os' namespace
210 - move headers src/include -> src/include/gras/ and stop fooling with
211 gras -> . symbolic link
212 - make distcheck is now successful
214 2004-07-19 Martin Quinson
216 - Build shared library also
217 - Install html doc to the right location
218 - stop removing maintainer files in make clean
219 - build tests only on make check
221 2004-07-13 Martin Quinson
223 - No major issue in previous version => change versionning schema
224 - Re-enable little64 convertion test now that Abdou kindly regenerated the
225 corresponding dataset.
227 2004-07-11 Martin Quinson
229 - Get it working with any kind of structure (we can compute the padding
230 bytes remotely for all the architectures I have access to)
231 - Implement the structure parsing macro (still not quite robust/complete)
232 - Improvement to the remote testing toysuite
234 2004-07-10 Martin Quinson
236 - get ride of a bunch of deprecated macros
237 - actually run the test for two-compliment, not only compile it :-/
238 - test whether the structures get packed (and bail out if yes. Damn.
239 Alignment is a serious matter)
240 - test whether the structures get compacted (but respecting the alignment
241 constraints of each types)
242 - test whether the array fields of structures can straddle alignment boundaries
244 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
245 (took me 2 hours to find that bug, looking at the wrong place)
246 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
247 before sending/receiving and are used to compute the offsets of fields
248 - Given that padding size depend even on compiler options, keep track of
249 alignment and aligned_size only for the current architecture. Not a big
250 deal since we send structure fields one after the other (seems
252 - Add the datastructure used for IEEE paper by the PBIO guys to the test
253 program, let it work on linux/gcc/little32. portability todo.
255 2004-07-08 Martin Quinson
256 - import and improve remote compilation support from FAST
257 - make sure make check works on half a dozen of machines out there
259 2004-07-07 Martin Quinson
260 Let's say it's version 0.0.3 ;)
261 - Implement conversions (yuhu!)
262 - Let it work on solaris (beside conversion, of course)
263 - Stupid me, using rand() to generate the conversion datatests in not wise.
265 2004-07-06 Martin Quinson
266 - Let make dist work, since I'm gonna need it to compile on remote hosts
267 - Let Tests/datadesc_usage write the architecture on which the file was
268 generated as first byte.
269 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
270 architecture descriptions.
271 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
272 on those architectures.
273 - Optimization: send/recv array of scalar in one shoot
275 2004-07-05 Martin Quinson
276 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
278 - Plug a whole bunch of memleaks
279 - each process now have to call gras_{init,exit}. One day, their log
280 settings will be separated
281 - Continue the code factorisation between SG, RL and common in Transport.
283 2004-07-04 Martin Quinson
285 - Redistribution between SG and RL.
286 We wanna have to accept in SG, so move accepted related parts of RL in
287 the common part. (more precisely, the dynar of all known sockets is no
288 more a static in transport.c, but part of the process_data)
290 [gras_stub_generator]
291 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
293 2004-07-03 Martin Quinson
294 - Create a new log channel tbx containing dict, set, log, dynar (to shut
295 them all up in one shot)
297 - Fix the ugly case of reference to dynamic array.
298 - New (semi-public) function gras_datadesc_size to allow the messaging
299 layer to malloc the needed space for the buffer.
301 - gras_socket_close now expect the socket to close (and not its address to
302 put NULL in it after it). This is because the socket passed to handlers
303 is one of their argument (=> not writable).
305 - propagate the interface cleanup from last week in datadesc, ie remove a
306 superfluous level of indirection. User pass adress of variable
307 containing data (both when sending and receiving), and not of a variable
308 being a pointer to the data. Let's say that I like it better ;)
309 The price for that is constructs like "int msg=*(int*)payload" in
310 handlers, but it's a fine price, IMHO.
312 - Let it work in RL (yuhu)
314 2004-06-21 Martin Quinson
316 - porting SG plugin and SG select to new standards (works almost).
317 - plug memleaks and fix bugs around.
320 - cleanup the prototype of data recv and force users to specify when they
321 want to handle references to objects. Test case working even for cycles.
322 - plug memleaks. Valgrind is perfectly ok with this.
324 2004-06-12 Martin Quinson
326 - cleanup the separation between plugin and main code in plugin creation
328 2004-06-11 Martin Quinson
330 - Reput hook for raw sockets, needed for BW experiments
331 - kill a few lines of dead code
332 [Data description] Interface cleanup
333 - gras_datadesc_by_name returns the searched type or NULL.
334 That way, no variable is needed to use a type desc once, which makes
336 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
337 two parameters were strings (field name, type name), leading to
339 [Dicos] Interface cleanup
340 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
341 This is consistant with the dynar API.
343 2004-04-21 Martin Quinson
345 - Porting to new standards.
348 There is no bag anymore, no need to take extra provision to mask the
349 pointers behind "ID".
350 Better splitup of functions between files create/exchange/convert.
351 This is still a bit artificial since convert and receive are so
352 interleaved, but anyway.
354 - add a queued message list to procdata (the ones not matching criteria
356 - factorize some more code between SG and RL wrt procdata
358 - use gras_exit in example to track memleaks
359 - get rid of gs_example now that GS is properly integrated into gras
360 - update run_test to integrate the lastest tests (datadesc)
362 - rename WARNINGn macros to WARNn since it prooved error-prone
364 2004-04-19 Martin Quinson
366 - register init/exit functions within gras module mecanism
367 - send/receive function.
368 Convertion is not implemented, but short-cutted if not needed.
369 struct/array elements are sent one by one (instead of block-wise), but
370 nobody really cares (yet). Get a prototype before optimizing.
371 - tests (using a file socket) for DD send/receive on:
372 - base types: int, float
373 - array: fixed size, string (ie ref to dynamic string)
374 - structure: homogeneous, heterogeneous
375 - chained list, graph with cycle
376 Believe it or not, valgrind is not too unhappy with the results. The
377 cycle happily segfaults, but the others are ok. And I'm sick of pointers
381 - Bugfix when using a filename explicitely (instead of '-')
383 2004-04-09 Martin Quinson
385 - factorize more code between RL and SG in socket creation
386 - Complete the implementation and tests of:
388 o file (only in RL, and mainly for debugging)
390 I lost 3 days to design a portable address resolver, and then decided
391 that the prototype mainly have to run on my box.
392 Addressing portability too early may be like optimizing too early :-/
394 - use gras_init in the Tests instead of the crappy parse_log_opt
395 (the latter function is removed)
396 [Conditional execution]
397 - New functions: gras_if_RL/gras_if_SG (basic support for this)
398 [Code reorganisation]
399 - Get rid of libgrasutils.a since it makes more trouble than it solves.
400 Build examples against the RL library, since there is no way to disable
401 its creation for now.