1 2004-11-04 Martin Quinson
2 - Allow to (en/dis)able the cycle detection at run time.
4 Whether we should check for cycle or not is now a property of each
5 datatype. When you think there may be some cycle, use datadesc_cycle_set.
6 datadesc_cycle_unset allow to remove this property when previously set.
8 Note that the cycle detection is off by default since it impacts the
9 performance. Watch the data you feed GRAS with ;)
11 This property is hereditary. Any element embeeded in a structure having it
12 set have it set for the time of this data exchange.
14 You should set it both on sender and receiver side. If you don't set it on
15 sender side, it will enter an endless loop. If you forget on receiver
16 side, the cycles won't be recreated after communication.
18 - Header reorganization.
19 Kill gras_private.h, each submodule must load the headers it needs.
21 2004-10-04 Martin Quinson
22 - Interface revolution: do not try to survive to malloc failure.
24 Now, gras_malloc and friends call gras_abort() on failure.
25 As a conclusion, malloc_error is not a valid error anymore, and all
26 functions for which it was the only gras_error_t return value are
27 changed. They now return void, or there result directly.
28 This simplify the API a lot.
30 2004-09-29 Martin Quinson
31 - Re-enable raw sockets.
32 Created by gras_socket_{client,server}_ext;
33 Used with gras_raw_{send,recv}
36 It should allow to kill the last bits of gras first version soon.
38 This is not completely satisfactory yet (dupplicate code with
39 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
42 - Simplify transport plugin (internal) interface by not passing any
43 argument to _server and _client, but embeeding them in the socket
46 2004-09-28 Martin Quinson
47 - Finish the port to AIX.
48 autoconf was my problem (segfault within the malloc replacement
49 function. No idea why)
51 2004-09-16 Martin Quinson
52 - Fix some size_t madness on 64bit architectures.
54 2004-09-08 Martin Quinson
55 - Reduce the number of system headers loaded, overload some more system
56 calls (such as malloc to cast the result of the system one, and work
58 - Fix and reintroduce the config support
60 2004-09-07 Martin Quinson
61 - Source code reorganization to allow Arnaud to surf all over there.
62 - Allow to document the logging categories.
63 - Remove all uppercase from logging categories and useless cleanup in names.
65 2004-08-18 Martin Quinson
66 Version 0.6.2 (protocol not changed; API changed)
67 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
68 gras_error_t), and NULL when not found. Functions expecting a msgtype
69 as argument (msg_wait; msg_send) deal with NULL argument by providing a
70 hopefully usefull message.
71 - Portability to prehistoric sparcs again
73 2004-08-17 Martin Quinson
74 Version 0.6.1 (protocol not changed; ABI not changed)
75 - prealloc some buffers to speed things up
77 2004-08-11 Martin Quinson
78 Version 0.6 (protocol not changed; ABI expended)
79 - The parsing macro can deal with the references, provided that you add
80 the relevant annotations (using GRAS_ANNOTE(size,field_name))
82 2004-08-09 Martin Quinson
83 Version 0.5 (protocol not changed; ABI changed)
84 - Allow to off turn the cycle detection code in data exchange at
85 compilation time. It should be at run time, but I'm short of time (and
86 the config stuff is still broken). That way, we keep dict out of the
87 critical path, which is good because the performance is poor:
88 - search not dichotomial yet
89 - dynar give no way to access their content and memcpy everytime
90 - In composed data description (struct, ref and so on), stop foolness of
91 keeping the subtype's ID, but store the type itself. This keeps sets out
92 of the critical path, which is good since they rely on dynar and
93 dictionnaries. The only loose of that is that we cannot detect the
94 redeclaration of a structure/union with another content (but I'm not sure
95 the code detected well this error before anyway). We still can detect
96 the redefinition discrepancy for the other types.
97 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
98 it breaks the code because of type-punning used all over the place).
99 This breaks on all non-gcc architectures (for now).
101 All those changes (plus the buffer of last time) allow me to gain 2 order
102 of magnitude on cruel tests consisting of 800000 array of integers on two
103 level of a hierarchical structure (200 secondes -> 4 secondes)
106 - the selector of reference must now return the type it points to, not
109 2004-08-06 Martin Quinson
110 Version 0.4 (protocol changed; ABI not changed)
111 - Allow to pass --gras-log argument to processes in simulation mode. Really.
112 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
113 - Implement a buffer transport, and use it by default (it relies on tcp in
114 real life and on sg in simulation).
115 That's a bit hackish since I had a new field to the structure to store
116 its data without interfering with the subtype ones. Inheritance
117 is tricky in C. And that's a kind of reverse inheritance with one class
118 derivating two classes. Or maybe a game with java interfaces. Anyway,
119 that's damn hard in C (at least).
120 Moreover, I got tired while trying to ensure plugin separation and
121 genericity in SG mode. MSG wants me to do weird things, so let's go for
122 cruel hacks (temporarily of course ;).
123 See comment in transport_private.h:71
124 - do not include all the _interface headers in private but in the files
125 which really need them (to cut the compilation time when they are
128 2004-07-26 Martin Quinson
129 Version 0.3 (protocol not changed; ABI changed)
130 - Major overhault of the datadesc interface to simplify it:
131 - shorted the function names:
132 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
133 - add a trivial way to push/pop integers into the cbps without malloc.
134 This allows to make really generic sub_type description, which simply
135 pop their size of the stack.
136 - add a function gras_datadesc_ref_pop_arr() which does what users want
137 most of the time: Declare a dynamic array (which pops its size of the
138 stack) and declare a reference to it. Poor name, but anyway.
139 - kill the post-send callback, add a post-receive one
141 2004-07-23 Martin Quinson
142 Version 0.2 (protocol changed; ABI changed)
143 - add some testing for cpbs in the test cases, and fix some more bugs.
144 This invalidate again the little64 data file, since I cannot regenerate
146 - remove an awfull optimization in the logging stuff, allowing me to:
147 - understand it again
148 - learn gcc how to check that the argument match the provided format
149 - fix all errors revealed by gcc after that
150 - internal keys of dict are not \0 terminated. Deal with it properly in
151 loggings instead of segfaulting when the user want to see the logs :-/
153 2004-07-22 Martin Quinson
154 - Fix some stupid bug preventing cbps (callback postit) from working
156 2004-07-21 Martin Quinson
157 - Some documentation cleanups
158 - remove the useless last argument of msgtype_declare
159 - rename the Virtu functions to fit into the 'os' namespace
160 - move headers src/include -> src/include/gras/ and stop fooling with
161 gras -> . symbolic link
162 - make distcheck is now successful
164 2004-07-19 Martin Quinson
166 - Build shared library also
167 - Install html doc to the right location
168 - stop removing maintainer files in make clean
169 - build tests only on make check
171 2004-07-13 Martin Quinson
173 - No major issue in previous version => change versionning schema
174 - Re-enable little64 convertion test now that Abdou kindly regenerated the
175 corresponding dataset.
177 2004-07-11 Martin Quinson
179 - Get it working with any kind of structure (we can compute the padding
180 bytes remotely for all the architectures I have access to)
181 - Implement the structure parsing macro (still not quite robust/complete)
182 - Improvement to the remote testing toysuite
184 2004-07-10 Martin Quinson
186 - get ride of a bunch of deprecated macros
187 - actually run the test for two-compliment, not only compile it :-/
188 - test whether the structures get packed (and bail out if yes. Damn.
189 Alignment is a serious matter)
190 - test whether the structures get compacted (but respecting the alignment
191 constraints of each types)
192 - test whether the array fields of structures can straddle alignment boundaries
194 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
195 (took me 2 hours to find that bug, looking at the wrong place)
196 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
197 before sending/receiving and are used to compute the offsets of fields
198 - Given that padding size depend even on compiler options, keep track of
199 alignment and aligned_size only for the current architecture. Not a big
200 deal since we send structure fields one after the other (seems
202 - Add the datastructure used for IEEE paper by the PBIO guys to the test
203 program, let it work on linux/gcc/little32. portability todo.
205 2004-07-08 Martin Quinson
206 - import and improve remote compilation support from FAST
207 - make sure make check works on half a dozen of machines out there
209 2004-07-07 Martin Quinson
210 Let's say it's version 0.0.3 ;)
211 - Implement conversions (yuhu!)
212 - Let it work on solaris (beside conversion, of course)
213 - Stupid me, using rand() to generate the conversion datatests in not wise.
215 2004-07-06 Martin Quinson
216 - Let make dist work, since I'm gonna need it to compile on remote hosts
217 - Let Tests/datadesc_usage write the architecture on which the file was
218 generated as first byte.
219 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
220 architecture descriptions.
221 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
222 on those architectures.
223 - Optimization: send/recv array of scalar in one shoot
225 2004-07-05 Martin Quinson
226 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
228 - Plug a whole bunch of memleaks
229 - each process now have to call gras_{init,exit}. One day, their log
230 settings will be separated
231 - Continue the code factorisation between SG, RL and common in Transport.
233 2004-07-04 Martin Quinson
235 - Redistribution between SG and RL.
236 We wanna have to accept in SG, so move accepted related parts of RL in
237 the common part. (more precisely, the dynar of all known sockets is no
238 more a static in transport.c, but part of the process_data)
240 [gras_stub_generator]
241 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
243 2004-07-03 Martin Quinson
244 - Create a new log channel tbx containing dict, set, log, dynar (to shut
245 them all up in one shot)
247 - Fix the ugly case of reference to dynamic array.
248 - New (semi-public) function gras_datadesc_size to allow the messaging
249 layer to malloc the needed space for the buffer.
251 - gras_socket_close now expect the socket to close (and not its address to
252 put NULL in it after it). This is because the socket passed to handlers
253 is one of their argument (=> not writable).
255 - propagate the interface cleanup from last week in datadesc, ie remove a
256 superfluous level of indirection. User pass adress of variable
257 containing data (both when sending and receiving), and not of a variable
258 being a pointer to the data. Let's say that I like it better ;)
259 The price for that is constructs like "int msg=*(int*)payload" in
260 handlers, but it's a fine price, IMHO.
262 - Let it work in RL (yuhu)
264 2004-06-21 Martin Quinson
266 - porting SG plugin and SG select to new standards (works almost).
267 - plug memleaks and fix bugs around.
270 - cleanup the prototype of data recv and force users to specify when they
271 want to handle references to objects. Test case working even for cycles.
272 - plug memleaks. Valgrind is perfectly ok with this.
274 2004-06-12 Martin Quinson
276 - cleanup the separation between plugin and main code in plugin creation
278 2004-06-11 Martin Quinson
280 - Reput hook for raw sockets, needed for BW experiments
281 - kill a few lines of dead code
282 [Data description] Interface cleanup
283 - gras_datadesc_by_name returns the searched type or NULL.
284 That way, no variable is needed to use a type desc once, which makes
286 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
287 two parameters were strings (field name, type name), leading to
289 [Dicos] Interface cleanup
290 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
291 This is consistant with the dynar API.
293 2004-04-21 Martin Quinson
295 - Porting to new standards.
298 There is no bag anymore, no need to take extra provision to mask the
299 pointers behind "ID".
300 Better splitup of functions between files create/exchange/convert.
301 This is still a bit artificial since convert and receive are so
302 interleaved, but anyway.
304 - add a queued message list to procdata (the ones not matching criteria
306 - factorize some more code between SG and RL wrt procdata
308 - use gras_exit in example to track memleaks
309 - get rid of gs_example now that GS is properly integrated into gras
310 - update run_test to integrate the lastest tests (datadesc)
312 - rename WARNINGn macros to WARNn since it prooved error-prone
314 2004-04-19 Martin Quinson
316 - register init/exit functions within gras module mecanism
317 - send/receive function.
318 Convertion is not implemented, but short-cutted if not needed.
319 struct/array elements are sent one by one (instead of block-wise), but
320 nobody really cares (yet). Get a prototype before optimizing.
321 - tests (using a file socket) for DD send/receive on:
322 - base types: int, float
323 - array: fixed size, string (ie ref to dynamic string)
324 - structure: homogeneous, heterogeneous
325 - chained list, graph with cycle
326 Believe it or not, valgrind is not too unhappy with the results. The
327 cycle happily segfaults, but the others are ok. And I'm sick of pointers
331 - Bugfix when using a filename explicitely (instead of '-')
333 2004-04-09 Martin Quinson
335 - factorize more code between RL and SG in socket creation
336 - Complete the implementation and tests of:
338 o file (only in RL, and mainly for debugging)
340 I lost 3 days to design a portable address resolver, and then decided
341 that the prototype mainly have to run on my box.
342 Addressing portability too early may be like optimizing too early :-/
344 - use gras_init in the Tests instead of the crappy parse_log_opt
345 (the latter function is removed)
346 [Conditional execution]
347 - New functions: gras_if_RL/gras_if_SG (basic support for this)
348 [Code reorganisation]
349 - Get rid of libgrasutils.a since it makes more trouble than it solves.
350 Build examples against the RL library, since there is no way to disable
351 its creation for now.