1 Version 0.7 (protocol not changed; API changed)
2 2004-09-08 Martin Quinson
3 - Reduce the number of system headers loaded, overload some more system
4 calls (such as malloc to cast the result of the system one, and work
6 - Fix and reintroduce the config support
8 2004-09-07 Martin Quinson
9 - Source code reorganization to allow Arnaud to surf all over there.
10 - Allow to document the logging categories
11 - Remove all uppercase from logging categories and useless cleanup in names.
13 2004-08-18 Martin Quinson
14 Version 0.6.2 (protocol not changed; API changed)
15 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
16 gras_error_t), and NULL when not found. Functions expecting a msgtype
17 as argument (msg_wait; msg_send) deal with NULL argument by providing a
18 hopefully usefull message.
19 - Portability to prehistoric sparcs again
21 2004-08-17 Martin Quinson
22 Version 0.6.1 (protocol not changed; ABI not changed)
23 - prealloc some buffers to speed things up
25 2004-08-11 Martin Quinson
26 Version 0.6 (protocol not changed; ABI expended)
27 - The parsing macro can deal with the references, provided that you add
28 the relevant annotations (using GRAS_ANNOTE(size,field_name))
30 2004-08-09 Martin Quinson
31 Version 0.5 (protocol not changed; ABI changed)
32 - Allow to off turn the cycle detection code in data exchange at
33 compilation time. It should be at run time, but I'm short of time (and
34 the config stuff is still broken). That way, we keep dict of of the
35 critical path, which is good because the performance is poor:
36 - search not dichotomial yet
37 - dynar give no way to access their content and memcpy everytime
38 - In composed data description (struct, ref and so on), stop foolness of
39 keeping the subtype's ID, but store the type itself. This keeps sets out
40 of the critical path, which is good since they rely on dynar and
41 dictionnaries. The only loose of that is that we cannot detect the
42 redeclaration of a structure/union with another content (but I'm not sure
43 the code detected this error well before anyway). We still can detect
44 the redefinition discrepancy for the other types.
45 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
46 it breaks the code because of type-punning used all over the place).
47 This breaks on all non-gcc architectures (for now).
49 All those changes (plus the buffer of last time) allow me to gain 2 order
50 of magnitude on cruel tests consisting of 800000 array of integers on two
51 level of a hierarchical structure (200 secondes -> 4 secondes)
54 - the selector of reference must now return the type it points to, not
57 2004-08-06 Martin Quinson
58 Version 0.4 (protocol changed; ABI not changed)
59 - Allow to pass --gras-log argument to processes in simulation mode. Really.
60 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
61 - Implement a buffer transport, and use it by default (it relies on tcp in
62 real life and on sg in simulation).
63 That's a bit hackish since I had a new field to the structure to store
64 its data without interfering with the subtype ones. Inheritance
65 is tricky in C. And that's a kind of reverse inheritance with one class
66 derivating two classes. Or maybe a game with java interfaces. Anyway,
67 that's damn hard in C (at least).
68 Moreover, I got tired while trying to ensure plugin separation and
69 genericity in SG mode. MSG wants me to do weird things, so let's go for
70 cruel hacks (temporarily of course ;).
71 See comment in transport_private.h:71
72 - do not include all the _interface headers in private but in the files
73 which really need them (to cut the compilation time when they are
76 2004-07-26 Martin Quinson
77 Version 0.3 (protocol not changed; ABI changed)
78 - Major overhault of the datadesc interface to simplify it:
79 - shorted the function names:
80 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
81 - add a trivial way to push/pop integers into the cbps without malloc.
82 This allows to make really generic sub_type description, which simply
83 pop their size of the stack.
84 - add a function gras_datadesc_ref_pop_arr() which does what users want
85 most of the time: Declare a dynamic array (which pops its size of the
86 stack) and declare a reference to it. Poor name, but anyway.
87 - kill the post-send callback, add a post-receive one
89 2004-07-23 Martin Quinson
90 Version 0.2 (protocol changed; ABI changed)
91 - add some testing for cpbs in the test cases, and fix some more bugs.
92 This invalidate again the little64 data file, since I cannot regenerate
94 - remove an awfull optimization in the logging stuff, allowing me to:
96 - learn gcc how to check that the argument match the provided format
97 - fix all errors revealed by gcc after that
98 - internal keys of dict are not \0 terminated. Deal with it properly in
99 loggings instead of segfaulting when the user want to see the logs :-/
101 2004-07-22 Martin Quinson
102 - Fix some stupid bug preventing cbps (callback postit) from working
104 2004-07-21 Martin Quinson
105 - Some documentation cleanups
106 - remove the useless last argument of msgtype_declare
107 - rename the Virtu functions to fit into the 'os' namespace
108 - move headers src/include -> src/include/gras/ and stop fooling with
109 gras -> . symbolic link
110 - make distcheck is now successful
112 2004-07-19 Martin Quinson
114 - Build shared library also
115 - Install html doc to the right location
116 - stop removing maintainer files in make clean
117 - build tests only on make check
119 2004-07-13 Martin Quinson
121 - No major issue in previous version => change versionning schema
122 - Re-enable little64 convertion test now that Abdou kindly regenerated the
123 corresponding dataset.
125 2004-07-11 Martin Quinson
127 - Get it working with any kind of structure (we can compute the padding
128 bytes remotely for all the architectures I have access to)
129 - Implement the structure parsing macro (still not quite robust/complete)
130 - Improvement to the remote testing toysuite
132 2004-07-10 Martin Quinson
134 - get ride of a bunch of deprecated macros
135 - actually run the test for two-compliment, not only compile it :-/
136 - test whether the structures get packed (and bail out if yes. Damn.
137 Alignment is a serious matter)
138 - test whether the structures get compacted (but respecting the alignment
139 constraints of each types)
140 - test whether the array fields of structures can straddle alignment boundaries
142 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
143 (took me 2 hours to find that bug, looking at the wrong place)
144 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
145 before sending/receiving and are used to compute the offsets of fields
146 - Given that padding size depend even on compiler options, keep track of
147 alignment and aligned_size only for the current architecture. Not a big
148 deal since we send structure fields one after the other (seems
150 - Add the datastructure used for IEEE paper by the PBIO guys to the test
151 program, let it work on linux/gcc/little32. portability todo.
153 2004-07-08 Martin Quinson
154 - import and improve remote compilation support from FAST
155 - make sure make check works on half a dozen of machines out there
157 2004-07-07 Martin Quinson
158 Let's say it's version 0.0.3 ;)
159 - Implement conversions (yuhu!)
160 - Let it work on solaris (beside conversion, of course)
161 - Stupid me, using rand() to generate the conversion datatests in not wise.
163 2004-07-06 Martin Quinson
164 - Let make dist work, since I'm gonna need it to compile on remote hosts
165 - Let Tests/datadesc_usage write the architecture on which the file was
166 generated as first byte.
167 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
168 architecture descriptions.
169 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
170 on those architectures.
171 - Optimization: send/recv array of scalar in one shoot
173 2004-07-05 Martin Quinson
174 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
176 - Plug a whole bunch of memleaks
177 - each process now have to call gras_{init,exit}. One day, their log
178 settings will be separated
179 - Continue the code factorisation between SG, RL and common in Transport.
181 2004-07-04 Martin Quinson
183 - Redistribution between SG and RL.
184 We wanna have to accept in SG, so move accepted related parts of RL in
185 the common part. (more precisely, the dynar of all known sockets is no
186 more a static in transport.c, but part of the process_data)
188 [gras_stub_generator]
189 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
191 2004-07-03 Martin Quinson
192 - Create a new log channel tbx containing dict, set, log, dynar (to shut
193 them all up in one shot)
195 - Fix the ugly case of reference to dynamic array.
196 - New (semi-public) function gras_datadesc_size to allow the messaging
197 layer to malloc the needed space for the buffer.
199 - gras_socket_close now expect the socket to close (and not its address to
200 put NULL in it after it). This is because the socket passed to handlers
201 is one of their argument (=> not writable).
203 - propagate the interface cleanup from last week in datadesc, ie remove a
204 superfluous level of indirection. User pass adress of variable
205 containing data (both when sending and receiving), and not of a variable
206 being a pointer to the data. Let's say that I like it better ;)
207 The price for that is constructs like "int msg=*(int*)payload" in
208 handlers, but it's a fine price, IMHO.
210 - Let it work in RL (yuhu)
212 2004-06-21 Martin Quinson
214 - porting SG plugin and SG select to new standards (works almost).
215 - plug memleaks and fix bugs around.
218 - cleanup the prototype of data recv and force users to specify when they
219 want to handle references to objects. Test case working even for cycles.
220 - plug memleaks. Valgrind is perfectly ok with this.
222 2004-06-12 Martin Quinson
224 - cleanup the separation between plugin and main code in plugin creation
226 2004-06-11 Martin Quinson
228 - Reput hook for raw sockets, needed for BW experiments
229 - kill a few lines of dead code
230 [Data description] Interface cleanup
231 - gras_datadesc_by_name returns the searched type or NULL.
232 That way, no variable is needed to use a type desc once, which makes
234 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
235 two parameters were strings (field name, type name), leading to
237 [Dicos] Interface cleanup
238 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
239 This is consistant with the dynar API.
241 2004-04-21 Martin Quinson
243 - Porting to new standards.
246 There is no bag anymore, no need to take extra provision to mask the
247 pointers behind "ID".
248 Better splitup of functions between files create/exchange/convert.
249 This is still a bit artificial since convert and receive are so
250 interleaved, but anyway.
252 - add a queued message list to procdata (the ones not matching criteria
254 - factorize some more code between SG and RL wrt procdata
256 - use gras_exit in example to track memleaks
257 - get rid of gs_example now that GS is properly integrated into gras
258 - update run_test to integrate the lastest tests (datadesc)
260 - rename WARNINGn macros to WARNn since it prooved error-prone
262 2004-04-19 Martin Quinson
264 - register init/exit functions within gras module mecanism
265 - send/receive function.
266 Convertion is not implemented, but short-cutted if not needed.
267 struct/array elements are sent one by one (instead of block-wise), but
268 nobody really cares (yet). Get a prototype before optimizing.
269 - tests (using a file socket) for DD send/receive on:
270 - base types: int, float
271 - array: fixed size, string (ie ref to dynamic string)
272 - structure: homogeneous, heterogeneous
273 - chained list, graph with cycle
274 Believe it or not, valgrind is not too unhappy with the results. The
275 cycle happily segfaults, but the others are ok. And I'm sick of pointers
279 - Bugfix when using a filename explicitely (instead of '-')
281 2004-04-09 Martin Quinson
283 - factorize more code between RL and SG in socket creation
284 - Complete the implementation and tests of:
286 o file (only in RL, and mainly for debugging)
288 I lost 3 days to design a portable address resolver, and then decided
289 that the prototype mainly have to run on my box.
290 Addressing portability too early may be like optimizing too early :-/
292 - use gras_init in the Tests instead of the crappy parse_log_opt
293 (the latter function is removed)
294 [Conditional execution]
295 - New functions: gras_if_RL/gras_if_SG (basic support for this)
296 [Code reorganisation]
297 - Get rid of libgrasutils.a since it makes more trouble than it solves.
298 Build examples against the RL library, since there is no way to disable
299 its creation for now.