1 2004-07-21 Martin Quinson
3 - Some documentation cleanups
4 - remove the useless last argument of msgtype_declare
5 - rename the Virtu functions to fit into the 'os' namespace
7 2004-07-19 Martin Quinson
9 - Build shared library also
10 - Install html doc to the right location
11 - stop removing maintainer files in make clean
12 - build tests only on make check
14 2004-07-13 Martin Quinson
16 - No major issue in previous version => change versionning schema
17 - Re-enable little64 convertion test now that Abdou kindly regenerated the
18 corresponding dataset.
20 2004-07-11 Martin Quinson
22 - Get it working with any kind of structure (we can compute the padding
23 bytes remotely for all the architectures I have access to)
24 - Implement the structure parsing macro (still not quite robust/complete)
25 - Improvement to the remote testing toysuite
27 2004-07-10 Martin Quinson
29 - get ride of a bunch of deprecated macros
30 - actually run the test for two-compliment, not only compile it :-/
31 - test whether the structures get packed (and bail out if yes. Damn.
32 Alignment is a serious matter)
33 - test whether the structures get compacted (but respecting the alignment
34 constraints of each types)
35 - test whether the array fields of structures can straddle alignment boundaries
37 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
38 (took me 2 hours to find that bug, looking at the wrong place)
39 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
40 before sending/receiving and are used to compute the offsets of fields
41 - Given that padding size depend even on compiler options, keep track of
42 alignment and aligned_size only for the current architecture. Not a big
43 deal since we send structure fields one after the other (seems
45 - Add the datastructure used for IEEE paper by the PBIO guys to the test
46 program, let it work on linux/gcc/little32. portability todo.
48 2004-07-08 Martin Quinson
49 - import and improve remote compilation support from FAST
50 - make sure make check works on half a dozen of machines out there
52 2004-07-07 Martin Quinson
53 Let's say it's version 0.0.3 ;)
54 - Implement conversions (yuhu!)
55 - Let it work on solaris (beside conversion, of course)
56 - Stupid me, using rand() to generate the conversion datatests in not wise.
58 2004-07-06 Martin Quinson
59 - Let make dist work, since I'm gonna need it to compile on remote hosts
60 - Let Tests/datadesc_usage write the architecture on which the file was
61 generated as first byte.
62 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
63 architecture descriptions.
64 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
65 on those architectures.
66 - Optimization: send/recv array of scalar in one shoot
68 2004-07-05 Martin Quinson
69 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
71 - Plug a whole bunch of memleaks
72 - each process now have to call gras_{init,exit}. One day, their log
73 settings will be separated
74 - Continue the code factorisation between SG, RL and common in Transport.
76 2004-07-04 Martin Quinson
78 - Redistribution between SG and RL.
79 We wanna have to accept in SG, so move accepted related parts of RL in
80 the common part. (more precisely, the dynar of all known sockets is no
81 more a static in transport.c, but part of the process_data)
84 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
86 2004-07-03 Martin Quinson
87 - Create a new log channel tbx containing dict, set, log, dynar (to shut
88 them all up in one shot)
90 - Fix the ugly case of reference to dynamic array.
91 - New (semi-public) function gras_datadesc_size to allow the messaging
92 layer to malloc the needed space for the buffer.
94 - gras_socket_close now expect the socket to close (and not its address to
95 put NULL in it after it). This is because the socket passed to handlers
96 is one of their argument (=> not writable).
98 - propagate the interface cleanup from last week in datadesc, ie remove a
99 superfluous level of indirection. User pass adress of variable
100 containing data (both when sending and receiving), and not of a variable
101 being a pointer to the data. Let's say that I like it better ;)
102 The price for that is constructs like "int msg=*(int*)payload" in
103 handlers, but it's a fine price, IMHO.
105 - Let it work in RL (yuhu)
107 2004-06-21 Martin Quinson
109 - porting SG plugin and SG select to new standards (works almost).
110 - plug memleaks and fix bugs around.
113 - cleanup the prototype of data recv and force users to specify when they
114 want to handle references to objects. Test case working even for cycles.
115 - plug memleaks. Valgrind is perfectly ok with this.
117 2004-06-12 Martin Quinson
119 - cleanup the separation between plugin and main code in plugin creation
121 2004-06-11 Martin Quinson
123 - Reput hook for raw sockets, needed for BW experiments
124 - kill a few lines of dead code
125 [Data description] Interface cleanup
126 - gras_datadesc_by_name returns the searched type or NULL.
127 That way, no variable is needed to use a type desc once, which makes
129 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
130 two parameters were strings (field name, type name), leading to
132 [Dicos] Interface cleanup
133 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
134 This is consistant with the dynar API.
136 2004-04-21 Martin Quinson
138 - Porting to new standards.
141 There is no bag anymore, no need to take extra provision to mask the
142 pointers behind "ID".
143 Better splitup of functions between files create/exchange/convert.
144 This is still a bit artificial since convert and receive are so
145 interleaved, but anyway.
147 - add a queued message list to procdata (the ones not matching criteria
149 - factorize some more code between SG and RL wrt procdata
151 - use gras_exit in example to track memleaks
152 - get rid of gs_example now that GS is properly integrated into gras
153 - update run_test to integrate the lastest tests (datadesc)
155 - rename WARNINGn macros to WARNn since it prooved error-prone
157 2004-04-19 Martin Quinson
159 - register init/exit functions within gras module mecanism
160 - send/receive function.
161 Convertion is not implemented, but short-cutted if not needed.
162 struct/array elements are sent one by one (instead of block-wise), but
163 nobody really cares (yet). Get a prototype before optimizing.
164 - tests (using a file socket) for DD send/receive on:
165 - base types: int, float
166 - array: fixed size, string (ie ref to dynamic string)
167 - structure: homogeneous, heterogeneous
168 - chained list, graph with cycle
169 Believe it or not, valgrind is not too unhappy with the results. The
170 cycle happily segfaults, but the others are ok. And I'm sick of pointers
174 - Bugfix when using a filename explicitely (instead of '-')
176 2004-04-09 Martin Quinson
178 - factorize more code between RL and SG in socket creation
179 - Complete the implementation and tests of:
181 o file (only in RL, and mainly for debugging)
183 I lost 3 days to design a portable address resolver, and then decided
184 that the prototype mainly have to run on my box.
185 Addressing portability too early may be like optimizing too early :-/
187 - use gras_init in the Tests instead of the crappy parse_log_opt
188 (the latter function is removed)
189 [Conditional execution]
190 - New functions: gras_if_RL/gras_if_SG (basic support for this)
191 [Code reorganisation]
192 - Get rid of libgrasutils.a since it makes more trouble than it solves.
193 Build examples against the RL library, since there is no way to disable
194 its creation for now.