1 2004-08-18 Martin Quinson
2 Version 0.6.2 (protocol not changed; API changed)
3 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4 gras_error_t), and NULL when not found. Functions expecting a msgtype
5 as argument (msg_wait; msg_send) deal with NULL argument by providing a
6 hopefully usefull message.
7 - Portability to prehistoric sparcs again
9 2004-08-17 Martin Quinson
10 Version 0.6.1 (protocol not changed; ABI not changed)
11 - prealloc some buffers to speed things up
13 2004-08-11 Martin Quinson
14 Version 0.6 (protocol not changed; ABI expended)
15 - The parsing macro can deal with the references, provided that you add
16 the relevant annotations (using GRAS_ANNOTE(size,field_name))
18 2004-08-09 Martin Quinson
19 Version 0.5 (protocol not changed; ABI changed)
20 - Allow to off turn the cycle detection code in data exchange at
21 compilation time. It should be at run time, but I'm short of time (and
22 the config stuff is still broken). That way, we keep dict of of the
23 critical path, which is good because the performance is poor:
24 - search not dichotomial yet
25 - dynar give no way to access their content and memcpy everytime
26 - In composed data description (struct, ref and so on), stop foolness of
27 keeping the subtype's ID, but store the type itself. This keeps sets out
28 of the critical path, which is good since they rely on dynar and
29 dictionnaries. The only loose of that is that we cannot detect the
30 redeclaration of a structure/union with another content (but I'm not sure
31 the code detected this error well before anyway). We still can detect
32 the redefinition discrepancy for the other types.
33 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
34 it breaks the code because of type-punning used all over the place).
35 This breaks on all non-gcc architectures (for now).
37 All those changes (plus the buffer of last time) allow me to gain 2 order
38 of magnitude on cruel tests consisting of 800000 array of integers on two
39 level of a hierarchical structure (200 secondes -> 4 secondes)
42 - the selector of reference must now return the type it points to, not
45 2004-08-06 Martin Quinson
46 Version 0.4 (protocol changed; ABI not changed)
47 - Allow to pass --gras-log argument to processes in simulation mode. Really.
48 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
49 - Implement a buffer transport, and use it by default (it relies on tcp in
50 real life and on sg in simulation).
51 That's a bit hackish since I had a new field to the structure to store
52 its data without interfering with the subtype ones. Inheritance
53 is tricky in C. And that's a kind of reverse inheritance with one class
54 derivating two classes. Or maybe a game with java interfaces. Anyway,
55 that's damn hard in C (at least).
56 Moreover, I got tired while trying to ensure plugin separation and
57 genericity in SG mode. MSG wants me to do weird things, so let's go for
58 cruel hacks (temporarily of course ;).
59 See comment in transport_private.h:71
60 - do not include all the _interface headers in private but in the files
61 which really need them (to cut the compilation time when they are
64 2004-07-26 Martin Quinson
65 Version 0.3 (protocol not changed; ABI changed)
66 - Major overhault of the datadesc interface to simplify it:
67 - shorted the function names:
68 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
69 - add a trivial way to push/pop integers into the cbps without malloc.
70 This allows to make really generic sub_type description, which simply
71 pop their size of the stack.
72 - add a function gras_datadesc_ref_pop_arr() which does what users want
73 most of the time: Declare a dynamic array (which pops its size of the
74 stack) and declare a reference to it. Poor name, but anyway.
75 - kill the post-send callback, add a post-receive one
77 2004-07-23 Martin Quinson
78 Version 0.2 (protocol changed; ABI changed)
79 - add some testing for cpbs in the test cases, and fix some more bugs.
80 This invalidate again the little64 data file, since I cannot regenerate
82 - remove an awfull optimization in the logging stuff, allowing me to:
84 - learn gcc how to check that the argument match the provided format
85 - fix all errors revealed by gcc after that
86 - internal keys of dict are not \0 terminated. Deal with it properly in
87 loggings instead of segfaulting when the user want to see the logs :-/
89 2004-07-22 Martin Quinson
90 - Fix some stupid bug preventing cbps (callback postit) from working
92 2004-07-21 Martin Quinson
93 - Some documentation cleanups
94 - remove the useless last argument of msgtype_declare
95 - rename the Virtu functions to fit into the 'os' namespace
96 - move headers src/include -> src/include/gras/ and stop fooling with
97 gras -> . symbolic link
98 - make distcheck is now successful
100 2004-07-19 Martin Quinson
102 - Build shared library also
103 - Install html doc to the right location
104 - stop removing maintainer files in make clean
105 - build tests only on make check
107 2004-07-13 Martin Quinson
109 - No major issue in previous version => change versionning schema
110 - Re-enable little64 convertion test now that Abdou kindly regenerated the
111 corresponding dataset.
113 2004-07-11 Martin Quinson
115 - Get it working with any kind of structure (we can compute the padding
116 bytes remotely for all the architectures I have access to)
117 - Implement the structure parsing macro (still not quite robust/complete)
118 - Improvement to the remote testing toysuite
120 2004-07-10 Martin Quinson
122 - get ride of a bunch of deprecated macros
123 - actually run the test for two-compliment, not only compile it :-/
124 - test whether the structures get packed (and bail out if yes. Damn.
125 Alignment is a serious matter)
126 - test whether the structures get compacted (but respecting the alignment
127 constraints of each types)
128 - test whether the array fields of structures can straddle alignment boundaries
130 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
131 (took me 2 hours to find that bug, looking at the wrong place)
132 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
133 before sending/receiving and are used to compute the offsets of fields
134 - Given that padding size depend even on compiler options, keep track of
135 alignment and aligned_size only for the current architecture. Not a big
136 deal since we send structure fields one after the other (seems
138 - Add the datastructure used for IEEE paper by the PBIO guys to the test
139 program, let it work on linux/gcc/little32. portability todo.
141 2004-07-08 Martin Quinson
142 - import and improve remote compilation support from FAST
143 - make sure make check works on half a dozen of machines out there
145 2004-07-07 Martin Quinson
146 Let's say it's version 0.0.3 ;)
147 - Implement conversions (yuhu!)
148 - Let it work on solaris (beside conversion, of course)
149 - Stupid me, using rand() to generate the conversion datatests in not wise.
151 2004-07-06 Martin Quinson
152 - Let make dist work, since I'm gonna need it to compile on remote hosts
153 - Let Tests/datadesc_usage write the architecture on which the file was
154 generated as first byte.
155 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
156 architecture descriptions.
157 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
158 on those architectures.
159 - Optimization: send/recv array of scalar in one shoot
161 2004-07-05 Martin Quinson
162 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
164 - Plug a whole bunch of memleaks
165 - each process now have to call gras_{init,exit}. One day, their log
166 settings will be separated
167 - Continue the code factorisation between SG, RL and common in Transport.
169 2004-07-04 Martin Quinson
171 - Redistribution between SG and RL.
172 We wanna have to accept in SG, so move accepted related parts of RL in
173 the common part. (more precisely, the dynar of all known sockets is no
174 more a static in transport.c, but part of the process_data)
176 [gras_stub_generator]
177 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
179 2004-07-03 Martin Quinson
180 - Create a new log channel tbx containing dict, set, log, dynar (to shut
181 them all up in one shot)
183 - Fix the ugly case of reference to dynamic array.
184 - New (semi-public) function gras_datadesc_size to allow the messaging
185 layer to malloc the needed space for the buffer.
187 - gras_socket_close now expect the socket to close (and not its address to
188 put NULL in it after it). This is because the socket passed to handlers
189 is one of their argument (=> not writable).
191 - propagate the interface cleanup from last week in datadesc, ie remove a
192 superfluous level of indirection. User pass adress of variable
193 containing data (both when sending and receiving), and not of a variable
194 being a pointer to the data. Let's say that I like it better ;)
195 The price for that is constructs like "int msg=*(int*)payload" in
196 handlers, but it's a fine price, IMHO.
198 - Let it work in RL (yuhu)
200 2004-06-21 Martin Quinson
202 - porting SG plugin and SG select to new standards (works almost).
203 - plug memleaks and fix bugs around.
206 - cleanup the prototype of data recv and force users to specify when they
207 want to handle references to objects. Test case working even for cycles.
208 - plug memleaks. Valgrind is perfectly ok with this.
210 2004-06-12 Martin Quinson
212 - cleanup the separation between plugin and main code in plugin creation
214 2004-06-11 Martin Quinson
216 - Reput hook for raw sockets, needed for BW experiments
217 - kill a few lines of dead code
218 [Data description] Interface cleanup
219 - gras_datadesc_by_name returns the searched type or NULL.
220 That way, no variable is needed to use a type desc once, which makes
222 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
223 two parameters were strings (field name, type name), leading to
225 [Dicos] Interface cleanup
226 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
227 This is consistant with the dynar API.
229 2004-04-21 Martin Quinson
231 - Porting to new standards.
234 There is no bag anymore, no need to take extra provision to mask the
235 pointers behind "ID".
236 Better splitup of functions between files create/exchange/convert.
237 This is still a bit artificial since convert and receive are so
238 interleaved, but anyway.
240 - add a queued message list to procdata (the ones not matching criteria
242 - factorize some more code between SG and RL wrt procdata
244 - use gras_exit in example to track memleaks
245 - get rid of gs_example now that GS is properly integrated into gras
246 - update run_test to integrate the lastest tests (datadesc)
248 - rename WARNINGn macros to WARNn since it prooved error-prone
250 2004-04-19 Martin Quinson
252 - register init/exit functions within gras module mecanism
253 - send/receive function.
254 Convertion is not implemented, but short-cutted if not needed.
255 struct/array elements are sent one by one (instead of block-wise), but
256 nobody really cares (yet). Get a prototype before optimizing.
257 - tests (using a file socket) for DD send/receive on:
258 - base types: int, float
259 - array: fixed size, string (ie ref to dynamic string)
260 - structure: homogeneous, heterogeneous
261 - chained list, graph with cycle
262 Believe it or not, valgrind is not too unhappy with the results. The
263 cycle happily segfaults, but the others are ok. And I'm sick of pointers
267 - Bugfix when using a filename explicitely (instead of '-')
269 2004-04-09 Martin Quinson
271 - factorize more code between RL and SG in socket creation
272 - Complete the implementation and tests of:
274 o file (only in RL, and mainly for debugging)
276 I lost 3 days to design a portable address resolver, and then decided
277 that the prototype mainly have to run on my box.
278 Addressing portability too early may be like optimizing too early :-/
280 - use gras_init in the Tests instead of the crappy parse_log_opt
281 (the latter function is removed)
282 [Conditional execution]
283 - New functions: gras_if_RL/gras_if_SG (basic support for this)
284 [Code reorganisation]
285 - Get rid of libgrasutils.a since it makes more trouble than it solves.
286 Build examples against the RL library, since there is no way to disable
287 its creation for now.