1 2004-09-29 Martin Quinson
2 - Re-enable raw sockets.
3 Created by gras_socket_{client,server}_ext;
4 Used with gras_raw_{send,recv}
6 It should allow to kill the last bits of gras first version.
8 This is not completely satisfactory yet (dupplicate code with
9 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
12 - Simplify transport plugin interface by not passing any argument to
13 _server and _client, but embeeding them in the socket struct directly.
15 2004-09-28 Martin Quinson
16 - Finish the port to AIX.
17 autoconf was my problem (segfault within the malloc replacement
18 function. No idea why)
20 2004-09-16 Martin Quinson
21 - Fix some size_t madness on 64bit architectures.
23 2004-09-08 Martin Quinson
24 - Reduce the number of system headers loaded, overload some more system
25 calls (such as malloc to cast the result of the system one, and work
27 - Fix and reintroduce the config support
29 2004-09-07 Martin Quinson
30 - Source code reorganization to allow Arnaud to surf all over there.
31 - Allow to document the logging categories
32 - Remove all uppercase from logging categories and useless cleanup in names.
34 2004-08-18 Martin Quinson
35 Version 0.6.2 (protocol not changed; API changed)
36 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
37 gras_error_t), and NULL when not found. Functions expecting a msgtype
38 as argument (msg_wait; msg_send) deal with NULL argument by providing a
39 hopefully usefull message.
40 - Portability to prehistoric sparcs again
42 2004-08-17 Martin Quinson
43 Version 0.6.1 (protocol not changed; ABI not changed)
44 - prealloc some buffers to speed things up
46 2004-08-11 Martin Quinson
47 Version 0.6 (protocol not changed; ABI expended)
48 - The parsing macro can deal with the references, provided that you add
49 the relevant annotations (using GRAS_ANNOTE(size,field_name))
51 2004-08-09 Martin Quinson
52 Version 0.5 (protocol not changed; ABI changed)
53 - Allow to off turn the cycle detection code in data exchange at
54 compilation time. It should be at run time, but I'm short of time (and
55 the config stuff is still broken). That way, we keep dict of of the
56 critical path, which is good because the performance is poor:
57 - search not dichotomial yet
58 - dynar give no way to access their content and memcpy everytime
59 - In composed data description (struct, ref and so on), stop foolness of
60 keeping the subtype's ID, but store the type itself. This keeps sets out
61 of the critical path, which is good since they rely on dynar and
62 dictionnaries. The only loose of that is that we cannot detect the
63 redeclaration of a structure/union with another content (but I'm not sure
64 the code detected this error well before anyway). We still can detect
65 the redefinition discrepancy for the other types.
66 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
67 it breaks the code because of type-punning used all over the place).
68 This breaks on all non-gcc architectures (for now).
70 All those changes (plus the buffer of last time) allow me to gain 2 order
71 of magnitude on cruel tests consisting of 800000 array of integers on two
72 level of a hierarchical structure (200 secondes -> 4 secondes)
75 - the selector of reference must now return the type it points to, not
78 2004-08-06 Martin Quinson
79 Version 0.4 (protocol changed; ABI not changed)
80 - Allow to pass --gras-log argument to processes in simulation mode. Really.
81 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
82 - Implement a buffer transport, and use it by default (it relies on tcp in
83 real life and on sg in simulation).
84 That's a bit hackish since I had a new field to the structure to store
85 its data without interfering with the subtype ones. Inheritance
86 is tricky in C. And that's a kind of reverse inheritance with one class
87 derivating two classes. Or maybe a game with java interfaces. Anyway,
88 that's damn hard in C (at least).
89 Moreover, I got tired while trying to ensure plugin separation and
90 genericity in SG mode. MSG wants me to do weird things, so let's go for
91 cruel hacks (temporarily of course ;).
92 See comment in transport_private.h:71
93 - do not include all the _interface headers in private but in the files
94 which really need them (to cut the compilation time when they are
97 2004-07-26 Martin Quinson
98 Version 0.3 (protocol not changed; ABI changed)
99 - Major overhault of the datadesc interface to simplify it:
100 - shorted the function names:
101 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
102 - add a trivial way to push/pop integers into the cbps without malloc.
103 This allows to make really generic sub_type description, which simply
104 pop their size of the stack.
105 - add a function gras_datadesc_ref_pop_arr() which does what users want
106 most of the time: Declare a dynamic array (which pops its size of the
107 stack) and declare a reference to it. Poor name, but anyway.
108 - kill the post-send callback, add a post-receive one
110 2004-07-23 Martin Quinson
111 Version 0.2 (protocol changed; ABI changed)
112 - add some testing for cpbs in the test cases, and fix some more bugs.
113 This invalidate again the little64 data file, since I cannot regenerate
115 - remove an awfull optimization in the logging stuff, allowing me to:
116 - understand it again
117 - learn gcc how to check that the argument match the provided format
118 - fix all errors revealed by gcc after that
119 - internal keys of dict are not \0 terminated. Deal with it properly in
120 loggings instead of segfaulting when the user want to see the logs :-/
122 2004-07-22 Martin Quinson
123 - Fix some stupid bug preventing cbps (callback postit) from working
125 2004-07-21 Martin Quinson
126 - Some documentation cleanups
127 - remove the useless last argument of msgtype_declare
128 - rename the Virtu functions to fit into the 'os' namespace
129 - move headers src/include -> src/include/gras/ and stop fooling with
130 gras -> . symbolic link
131 - make distcheck is now successful
133 2004-07-19 Martin Quinson
135 - Build shared library also
136 - Install html doc to the right location
137 - stop removing maintainer files in make clean
138 - build tests only on make check
140 2004-07-13 Martin Quinson
142 - No major issue in previous version => change versionning schema
143 - Re-enable little64 convertion test now that Abdou kindly regenerated the
144 corresponding dataset.
146 2004-07-11 Martin Quinson
148 - Get it working with any kind of structure (we can compute the padding
149 bytes remotely for all the architectures I have access to)
150 - Implement the structure parsing macro (still not quite robust/complete)
151 - Improvement to the remote testing toysuite
153 2004-07-10 Martin Quinson
155 - get ride of a bunch of deprecated macros
156 - actually run the test for two-compliment, not only compile it :-/
157 - test whether the structures get packed (and bail out if yes. Damn.
158 Alignment is a serious matter)
159 - test whether the structures get compacted (but respecting the alignment
160 constraints of each types)
161 - test whether the array fields of structures can straddle alignment boundaries
163 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
164 (took me 2 hours to find that bug, looking at the wrong place)
165 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
166 before sending/receiving and are used to compute the offsets of fields
167 - Given that padding size depend even on compiler options, keep track of
168 alignment and aligned_size only for the current architecture. Not a big
169 deal since we send structure fields one after the other (seems
171 - Add the datastructure used for IEEE paper by the PBIO guys to the test
172 program, let it work on linux/gcc/little32. portability todo.
174 2004-07-08 Martin Quinson
175 - import and improve remote compilation support from FAST
176 - make sure make check works on half a dozen of machines out there
178 2004-07-07 Martin Quinson
179 Let's say it's version 0.0.3 ;)
180 - Implement conversions (yuhu!)
181 - Let it work on solaris (beside conversion, of course)
182 - Stupid me, using rand() to generate the conversion datatests in not wise.
184 2004-07-06 Martin Quinson
185 - Let make dist work, since I'm gonna need it to compile on remote hosts
186 - Let Tests/datadesc_usage write the architecture on which the file was
187 generated as first byte.
188 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
189 architecture descriptions.
190 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
191 on those architectures.
192 - Optimization: send/recv array of scalar in one shoot
194 2004-07-05 Martin Quinson
195 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
197 - Plug a whole bunch of memleaks
198 - each process now have to call gras_{init,exit}. One day, their log
199 settings will be separated
200 - Continue the code factorisation between SG, RL and common in Transport.
202 2004-07-04 Martin Quinson
204 - Redistribution between SG and RL.
205 We wanna have to accept in SG, so move accepted related parts of RL in
206 the common part. (more precisely, the dynar of all known sockets is no
207 more a static in transport.c, but part of the process_data)
209 [gras_stub_generator]
210 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
212 2004-07-03 Martin Quinson
213 - Create a new log channel tbx containing dict, set, log, dynar (to shut
214 them all up in one shot)
216 - Fix the ugly case of reference to dynamic array.
217 - New (semi-public) function gras_datadesc_size to allow the messaging
218 layer to malloc the needed space for the buffer.
220 - gras_socket_close now expect the socket to close (and not its address to
221 put NULL in it after it). This is because the socket passed to handlers
222 is one of their argument (=> not writable).
224 - propagate the interface cleanup from last week in datadesc, ie remove a
225 superfluous level of indirection. User pass adress of variable
226 containing data (both when sending and receiving), and not of a variable
227 being a pointer to the data. Let's say that I like it better ;)
228 The price for that is constructs like "int msg=*(int*)payload" in
229 handlers, but it's a fine price, IMHO.
231 - Let it work in RL (yuhu)
233 2004-06-21 Martin Quinson
235 - porting SG plugin and SG select to new standards (works almost).
236 - plug memleaks and fix bugs around.
239 - cleanup the prototype of data recv and force users to specify when they
240 want to handle references to objects. Test case working even for cycles.
241 - plug memleaks. Valgrind is perfectly ok with this.
243 2004-06-12 Martin Quinson
245 - cleanup the separation between plugin and main code in plugin creation
247 2004-06-11 Martin Quinson
249 - Reput hook for raw sockets, needed for BW experiments
250 - kill a few lines of dead code
251 [Data description] Interface cleanup
252 - gras_datadesc_by_name returns the searched type or NULL.
253 That way, no variable is needed to use a type desc once, which makes
255 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
256 two parameters were strings (field name, type name), leading to
258 [Dicos] Interface cleanup
259 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
260 This is consistant with the dynar API.
262 2004-04-21 Martin Quinson
264 - Porting to new standards.
267 There is no bag anymore, no need to take extra provision to mask the
268 pointers behind "ID".
269 Better splitup of functions between files create/exchange/convert.
270 This is still a bit artificial since convert and receive are so
271 interleaved, but anyway.
273 - add a queued message list to procdata (the ones not matching criteria
275 - factorize some more code between SG and RL wrt procdata
277 - use gras_exit in example to track memleaks
278 - get rid of gs_example now that GS is properly integrated into gras
279 - update run_test to integrate the lastest tests (datadesc)
281 - rename WARNINGn macros to WARNn since it prooved error-prone
283 2004-04-19 Martin Quinson
285 - register init/exit functions within gras module mecanism
286 - send/receive function.
287 Convertion is not implemented, but short-cutted if not needed.
288 struct/array elements are sent one by one (instead of block-wise), but
289 nobody really cares (yet). Get a prototype before optimizing.
290 - tests (using a file socket) for DD send/receive on:
291 - base types: int, float
292 - array: fixed size, string (ie ref to dynamic string)
293 - structure: homogeneous, heterogeneous
294 - chained list, graph with cycle
295 Believe it or not, valgrind is not too unhappy with the results. The
296 cycle happily segfaults, but the others are ok. And I'm sick of pointers
300 - Bugfix when using a filename explicitely (instead of '-')
302 2004-04-09 Martin Quinson
304 - factorize more code between RL and SG in socket creation
305 - Complete the implementation and tests of:
307 o file (only in RL, and mainly for debugging)
309 I lost 3 days to design a portable address resolver, and then decided
310 that the prototype mainly have to run on my box.
311 Addressing portability too early may be like optimizing too early :-/
313 - use gras_init in the Tests instead of the crappy parse_log_opt
314 (the latter function is removed)
315 [Conditional execution]
316 - New functions: gras_if_RL/gras_if_SG (basic support for this)
317 [Code reorganisation]
318 - Get rid of libgrasutils.a since it makes more trouble than it solves.
319 Build examples against the RL library, since there is no way to disable
320 its creation for now.