1 Version 0.7 (protocol not changed; API changed)
2 2004-09-16 Martin Quinson
3 - Fix some size_t madness on 64bit architectures.
5 2004-09-08 Martin Quinson
6 - Reduce the number of system headers loaded, overload some more system
7 calls (such as malloc to cast the result of the system one, and work
9 - Fix and reintroduce the config support
11 2004-09-07 Martin Quinson
12 - Source code reorganization to allow Arnaud to surf all over there.
13 - Allow to document the logging categories
14 - Remove all uppercase from logging categories and useless cleanup in names.
16 2004-08-18 Martin Quinson
17 Version 0.6.2 (protocol not changed; API changed)
18 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
19 gras_error_t), and NULL when not found. Functions expecting a msgtype
20 as argument (msg_wait; msg_send) deal with NULL argument by providing a
21 hopefully usefull message.
22 - Portability to prehistoric sparcs again
24 2004-08-17 Martin Quinson
25 Version 0.6.1 (protocol not changed; ABI not changed)
26 - prealloc some buffers to speed things up
28 2004-08-11 Martin Quinson
29 Version 0.6 (protocol not changed; ABI expended)
30 - The parsing macro can deal with the references, provided that you add
31 the relevant annotations (using GRAS_ANNOTE(size,field_name))
33 2004-08-09 Martin Quinson
34 Version 0.5 (protocol not changed; ABI changed)
35 - Allow to off turn the cycle detection code in data exchange at
36 compilation time. It should be at run time, but I'm short of time (and
37 the config stuff is still broken). That way, we keep dict of of the
38 critical path, which is good because the performance is poor:
39 - search not dichotomial yet
40 - dynar give no way to access their content and memcpy everytime
41 - In composed data description (struct, ref and so on), stop foolness of
42 keeping the subtype's ID, but store the type itself. This keeps sets out
43 of the critical path, which is good since they rely on dynar and
44 dictionnaries. The only loose of that is that we cannot detect the
45 redeclaration of a structure/union with another content (but I'm not sure
46 the code detected this error well before anyway). We still can detect
47 the redefinition discrepancy for the other types.
48 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
49 it breaks the code because of type-punning used all over the place).
50 This breaks on all non-gcc architectures (for now).
52 All those changes (plus the buffer of last time) allow me to gain 2 order
53 of magnitude on cruel tests consisting of 800000 array of integers on two
54 level of a hierarchical structure (200 secondes -> 4 secondes)
57 - the selector of reference must now return the type it points to, not
60 2004-08-06 Martin Quinson
61 Version 0.4 (protocol changed; ABI not changed)
62 - Allow to pass --gras-log argument to processes in simulation mode. Really.
63 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
64 - Implement a buffer transport, and use it by default (it relies on tcp in
65 real life and on sg in simulation).
66 That's a bit hackish since I had a new field to the structure to store
67 its data without interfering with the subtype ones. Inheritance
68 is tricky in C. And that's a kind of reverse inheritance with one class
69 derivating two classes. Or maybe a game with java interfaces. Anyway,
70 that's damn hard in C (at least).
71 Moreover, I got tired while trying to ensure plugin separation and
72 genericity in SG mode. MSG wants me to do weird things, so let's go for
73 cruel hacks (temporarily of course ;).
74 See comment in transport_private.h:71
75 - do not include all the _interface headers in private but in the files
76 which really need them (to cut the compilation time when they are
79 2004-07-26 Martin Quinson
80 Version 0.3 (protocol not changed; ABI changed)
81 - Major overhault of the datadesc interface to simplify it:
82 - shorted the function names:
83 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
84 - add a trivial way to push/pop integers into the cbps without malloc.
85 This allows to make really generic sub_type description, which simply
86 pop their size of the stack.
87 - add a function gras_datadesc_ref_pop_arr() which does what users want
88 most of the time: Declare a dynamic array (which pops its size of the
89 stack) and declare a reference to it. Poor name, but anyway.
90 - kill the post-send callback, add a post-receive one
92 2004-07-23 Martin Quinson
93 Version 0.2 (protocol changed; ABI changed)
94 - add some testing for cpbs in the test cases, and fix some more bugs.
95 This invalidate again the little64 data file, since I cannot regenerate
97 - remove an awfull optimization in the logging stuff, allowing me to:
99 - learn gcc how to check that the argument match the provided format
100 - fix all errors revealed by gcc after that
101 - internal keys of dict are not \0 terminated. Deal with it properly in
102 loggings instead of segfaulting when the user want to see the logs :-/
104 2004-07-22 Martin Quinson
105 - Fix some stupid bug preventing cbps (callback postit) from working
107 2004-07-21 Martin Quinson
108 - Some documentation cleanups
109 - remove the useless last argument of msgtype_declare
110 - rename the Virtu functions to fit into the 'os' namespace
111 - move headers src/include -> src/include/gras/ and stop fooling with
112 gras -> . symbolic link
113 - make distcheck is now successful
115 2004-07-19 Martin Quinson
117 - Build shared library also
118 - Install html doc to the right location
119 - stop removing maintainer files in make clean
120 - build tests only on make check
122 2004-07-13 Martin Quinson
124 - No major issue in previous version => change versionning schema
125 - Re-enable little64 convertion test now that Abdou kindly regenerated the
126 corresponding dataset.
128 2004-07-11 Martin Quinson
130 - Get it working with any kind of structure (we can compute the padding
131 bytes remotely for all the architectures I have access to)
132 - Implement the structure parsing macro (still not quite robust/complete)
133 - Improvement to the remote testing toysuite
135 2004-07-10 Martin Quinson
137 - get ride of a bunch of deprecated macros
138 - actually run the test for two-compliment, not only compile it :-/
139 - test whether the structures get packed (and bail out if yes. Damn.
140 Alignment is a serious matter)
141 - test whether the structures get compacted (but respecting the alignment
142 constraints of each types)
143 - test whether the array fields of structures can straddle alignment boundaries
145 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
146 (took me 2 hours to find that bug, looking at the wrong place)
147 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
148 before sending/receiving and are used to compute the offsets of fields
149 - Given that padding size depend even on compiler options, keep track of
150 alignment and aligned_size only for the current architecture. Not a big
151 deal since we send structure fields one after the other (seems
153 - Add the datastructure used for IEEE paper by the PBIO guys to the test
154 program, let it work on linux/gcc/little32. portability todo.
156 2004-07-08 Martin Quinson
157 - import and improve remote compilation support from FAST
158 - make sure make check works on half a dozen of machines out there
160 2004-07-07 Martin Quinson
161 Let's say it's version 0.0.3 ;)
162 - Implement conversions (yuhu!)
163 - Let it work on solaris (beside conversion, of course)
164 - Stupid me, using rand() to generate the conversion datatests in not wise.
166 2004-07-06 Martin Quinson
167 - Let make dist work, since I'm gonna need it to compile on remote hosts
168 - Let Tests/datadesc_usage write the architecture on which the file was
169 generated as first byte.
170 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
171 architecture descriptions.
172 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
173 on those architectures.
174 - Optimization: send/recv array of scalar in one shoot
176 2004-07-05 Martin Quinson
177 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
179 - Plug a whole bunch of memleaks
180 - each process now have to call gras_{init,exit}. One day, their log
181 settings will be separated
182 - Continue the code factorisation between SG, RL and common in Transport.
184 2004-07-04 Martin Quinson
186 - Redistribution between SG and RL.
187 We wanna have to accept in SG, so move accepted related parts of RL in
188 the common part. (more precisely, the dynar of all known sockets is no
189 more a static in transport.c, but part of the process_data)
191 [gras_stub_generator]
192 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
194 2004-07-03 Martin Quinson
195 - Create a new log channel tbx containing dict, set, log, dynar (to shut
196 them all up in one shot)
198 - Fix the ugly case of reference to dynamic array.
199 - New (semi-public) function gras_datadesc_size to allow the messaging
200 layer to malloc the needed space for the buffer.
202 - gras_socket_close now expect the socket to close (and not its address to
203 put NULL in it after it). This is because the socket passed to handlers
204 is one of their argument (=> not writable).
206 - propagate the interface cleanup from last week in datadesc, ie remove a
207 superfluous level of indirection. User pass adress of variable
208 containing data (both when sending and receiving), and not of a variable
209 being a pointer to the data. Let's say that I like it better ;)
210 The price for that is constructs like "int msg=*(int*)payload" in
211 handlers, but it's a fine price, IMHO.
213 - Let it work in RL (yuhu)
215 2004-06-21 Martin Quinson
217 - porting SG plugin and SG select to new standards (works almost).
218 - plug memleaks and fix bugs around.
221 - cleanup the prototype of data recv and force users to specify when they
222 want to handle references to objects. Test case working even for cycles.
223 - plug memleaks. Valgrind is perfectly ok with this.
225 2004-06-12 Martin Quinson
227 - cleanup the separation between plugin and main code in plugin creation
229 2004-06-11 Martin Quinson
231 - Reput hook for raw sockets, needed for BW experiments
232 - kill a few lines of dead code
233 [Data description] Interface cleanup
234 - gras_datadesc_by_name returns the searched type or NULL.
235 That way, no variable is needed to use a type desc once, which makes
237 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
238 two parameters were strings (field name, type name), leading to
240 [Dicos] Interface cleanup
241 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
242 This is consistant with the dynar API.
244 2004-04-21 Martin Quinson
246 - Porting to new standards.
249 There is no bag anymore, no need to take extra provision to mask the
250 pointers behind "ID".
251 Better splitup of functions between files create/exchange/convert.
252 This is still a bit artificial since convert and receive are so
253 interleaved, but anyway.
255 - add a queued message list to procdata (the ones not matching criteria
257 - factorize some more code between SG and RL wrt procdata
259 - use gras_exit in example to track memleaks
260 - get rid of gs_example now that GS is properly integrated into gras
261 - update run_test to integrate the lastest tests (datadesc)
263 - rename WARNINGn macros to WARNn since it prooved error-prone
265 2004-04-19 Martin Quinson
267 - register init/exit functions within gras module mecanism
268 - send/receive function.
269 Convertion is not implemented, but short-cutted if not needed.
270 struct/array elements are sent one by one (instead of block-wise), but
271 nobody really cares (yet). Get a prototype before optimizing.
272 - tests (using a file socket) for DD send/receive on:
273 - base types: int, float
274 - array: fixed size, string (ie ref to dynamic string)
275 - structure: homogeneous, heterogeneous
276 - chained list, graph with cycle
277 Believe it or not, valgrind is not too unhappy with the results. The
278 cycle happily segfaults, but the others are ok. And I'm sick of pointers
282 - Bugfix when using a filename explicitely (instead of '-')
284 2004-04-09 Martin Quinson
286 - factorize more code between RL and SG in socket creation
287 - Complete the implementation and tests of:
289 o file (only in RL, and mainly for debugging)
291 I lost 3 days to design a portable address resolver, and then decided
292 that the prototype mainly have to run on my box.
293 Addressing portability too early may be like optimizing too early :-/
295 - use gras_init in the Tests instead of the crappy parse_log_opt
296 (the latter function is removed)
297 [Conditional execution]
298 - New functions: gras_if_RL/gras_if_SG (basic support for this)
299 [Code reorganisation]
300 - Get rid of libgrasutils.a since it makes more trouble than it solves.
301 Build examples against the RL library, since there is no way to disable
302 its creation for now.