Martin Quinson [Tue, 8 Nov 2011 21:19:42 +0000 (22:19 +0100)]
sanitize the way parsing callbacks of the routing submodule are registered
Martin Quinson [Tue, 8 Nov 2011 21:19:02 +0000 (22:19 +0100)]
One day, we'll have to deal with route creation in sg_platf. That day, we'll need that data structure
Martin Quinson [Tue, 8 Nov 2011 21:10:28 +0000 (22:10 +0100)]
simplify this function
Martin Quinson [Tue, 8 Nov 2011 20:57:55 +0000 (21:57 +0100)]
rename some types for consistency
Martin Quinson [Tue, 8 Nov 2011 20:54:39 +0000 (21:54 +0100)]
that FIXME would be a bad idea to implement
Martin Quinson [Tue, 8 Nov 2011 20:46:55 +0000 (21:46 +0100)]
reindent, no real change
Martin Quinson [Tue, 8 Nov 2011 20:40:14 +0000 (21:40 +0100)]
implement cluster tag on top of sg_platf w/o XML black magic
Martin Quinson [Tue, 8 Nov 2011 20:23:47 +0000 (21:23 +0100)]
Kill the XML-level callbacks to AS_open/close and create some at sg_platf level instead
Martin Quinson [Tue, 8 Nov 2011 17:15:53 +0000 (18:15 +0100)]
simplify that vicious nasty code
Martin Quinson [Tue, 8 Nov 2011 16:04:04 +0000 (17:04 +0100)]
kill dead code
Martin Quinson [Tue, 8 Nov 2011 15:38:54 +0000 (16:38 +0100)]
/me also hates void* pointers
Martin Quinson [Tue, 8 Nov 2011 15:04:22 +0000 (16:04 +0100)]
/me hates globals
Martin Quinson [Tue, 8 Nov 2011 15:00:57 +0000 (16:00 +0100)]
integrate the peer creation in sg_platf properly
Martin Quinson [Tue, 8 Nov 2011 14:46:04 +0000 (15:46 +0100)]
preparse the arguments of peer tag
Martin Quinson [Tue, 8 Nov 2011 14:30:09 +0000 (15:30 +0100)]
improve peer creator: use sg_platf_new_*, not the XML bypassing mechanism
Arnaud Giersch [Thu, 18 Nov 2010 21:58:58 +0000 (22:58 +0100)]
Destroy dax parser after use.
Arnaud Giersch [Mon, 7 Nov 2011 22:04:05 +0000 (23:04 +0100)]
Fix memory leak in xbt_cfg_set_parse().
Arnaud Giersch [Mon, 7 Nov 2011 21:24:44 +0000 (22:24 +0100)]
Kill bogus fixme: xbt_free_f already exists.
Arnaud Giersch [Mon, 7 Nov 2011 22:14:01 +0000 (23:14 +0100)]
Parenthesis should be protected in test parameters.
Martin Quinson [Tue, 8 Nov 2011 14:06:44 +0000 (15:06 +0100)]
further parser cleanups
- kill some dead code related to the old lua console
- do not load a header at the *end* of another header
- solve some header inclusion madness created in this process
Martin Quinson [Tue, 8 Nov 2011 13:52:27 +0000 (14:52 +0100)]
do use the parameters I'm given, not some random globals
Martin Quinson [Tue, 8 Nov 2011 13:29:21 +0000 (14:29 +0100)]
kill dead code
Martin Quinson [Tue, 8 Nov 2011 13:24:05 +0000 (14:24 +0100)]
improve comments
Martin Quinson [Tue, 8 Nov 2011 13:22:01 +0000 (14:22 +0100)]
plug a memleak when the cluster's router_id is not given
Martin Quinson [Tue, 8 Nov 2011 12:43:27 +0000 (13:43 +0100)]
fix constness in parser
Martin Quinson [Tue, 8 Nov 2011 12:15:40 +0000 (13:15 +0100)]
further fields renamings in the parser
Christophe Thiéry [Tue, 8 Nov 2011 10:55:47 +0000 (11:55 +0100)]
Update README of Lua examples
Christophe Thiéry [Tue, 8 Nov 2011 10:38:41 +0000 (11:38 +0100)]
Clean the code of Lua examples.
Reindent, remove useless semicolons, declare variables as local, etc.
Christophe Thiéry [Tue, 8 Nov 2011 09:58:58 +0000 (10:58 +0100)]
Add unversioned chord large deployment files to .gitignore
Also remove '\r' characters from .gitignore
Christophe Thiéry [Tue, 8 Nov 2011 09:48:40 +0000 (10:48 +0100)]
These files have nothing to do here
Martin Quinson [Tue, 8 Nov 2011 11:51:01 +0000 (12:51 +0100)]
further code simplification. This one were nasty
Martin Quinson [Tue, 8 Nov 2011 11:22:00 +0000 (12:22 +0100)]
simplify the code a bit
Martin Quinson [Tue, 8 Nov 2011 11:05:51 +0000 (12:05 +0100)]
Also cleanup the fields of s_sg_platf_link_cbarg_t
This one is not just a renaming:
- kill the unused V_policy_initial_link
- make V_link_sharing_policy (now 'policy') a
e_surf_link_sharing_policy_t (which thus become public)
This should be enough for field renaming for now. But the cleanup is
definitly not finished :-/
Martin Quinson [Tue, 8 Nov 2011 10:48:32 +0000 (11:48 +0100)]
also cleanup the field names of s_sg_platf_router_cbarg_t
Still manual, so atomic commits to avoid issues
Martin Quinson [Tue, 8 Nov 2011 10:42:19 +0000 (11:42 +0100)]
Parser cleanup: simplify structure's fields' names
Damn, my eclipse fail that refactoring! That's manual, pity me...
Martin Quinson [Tue, 8 Nov 2011 10:27:02 +0000 (11:27 +0100)]
Reimplement the lua console on top of sg_platf
I admit it was too early to do that change since sg_platf is not quite
fully separated from the XML parsing yet, but the cruft accumulates
from everywhere, so I have to break stuff a bit to fix the overall
picture. Yeah, the lua tests are not fixed yet.
Martin Quinson [Tue, 8 Nov 2011 08:06:29 +0000 (09:06 +0100)]
kill more empty function chaining in lua console
Martin Quinson [Mon, 7 Nov 2011 23:16:44 +0000 (00:16 +0100)]
Slight cleanups in the lua console
- kill some empty functions chaining
- allow for the creation of routers from lua
(the console is still broken, more profund cleanups will be needed)
Martin Quinson [Mon, 7 Nov 2011 23:10:41 +0000 (00:10 +0100)]
Kill some further empty function chaining.
This breaks the lua console for a while, but this is to be redone with
the new sg_platf anyway.
Martin Quinson [Mon, 7 Nov 2011 22:57:42 +0000 (23:57 +0100)]
Properly integrate the link creation into sg_platf
Christophe Thiéry [Mon, 7 Nov 2011 21:14:27 +0000 (22:14 +0100)]
This structure was not necessary
Christophe Thiéry [Mon, 7 Nov 2011 18:21:29 +0000 (19:21 +0100)]
Lua: don't garbage collect a task after sending it (the receiver will)
Also clean the code of Lua tasks. Show an appropriate error message when
the sender tries to access a task already sent to someone else.
Christophe Thiéry [Mon, 7 Nov 2011 18:20:49 +0000 (19:20 +0100)]
Lua: add a function sglua_copy_value
Christophe Thiéry [Mon, 7 Nov 2011 18:19:24 +0000 (19:19 +0100)]
Add comments about who should or shouldn't call MSG_task_destroy
Christophe Thiéry [Mon, 7 Nov 2011 17:30:46 +0000 (18:30 +0100)]
After a send, the sender should never access the task anymore.
Since processes are in different Lua worlds, the receiver can garbage
collect the task.
Martin Quinson [Mon, 7 Nov 2011 17:05:18 +0000 (18:05 +0100)]
further parser cleanups
Properly integrate the begin and end of platform definition in sg_platf.
New public functions:
- sg_platf_open()
- sg_platf_close()
New internal callback:
- sg_platf_postparse_add_cb(void_f_void_t)
(nobody cares about sg_platf_open(), it's just here for cosmetics in
user code)
Martin Quinson [Mon, 7 Nov 2011 15:40:25 +0000 (16:40 +0100)]
also deprecate ETag_surfxml_router
Martin Quinson [Mon, 7 Nov 2011 15:33:59 +0000 (16:33 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Martin Quinson [Mon, 7 Nov 2011 15:33:11 +0000 (16:33 +0100)]
Further parser cleanups
- Kill useless function chaining
- Deprecate the ETag_host in XML now that it's useless
Navarrop [Mon, 7 Nov 2011 15:06:05 +0000 (16:06 +0100)]
Don't use a fix value of host.
Navarrop [Mon, 7 Nov 2011 14:35:46 +0000 (15:35 +0100)]
Add a source file for model vivaldi. Write vivaldi_get_route.
Update vivaldi platforms (examples/platforms/syscoord/
Navarrop [Mon, 7 Nov 2011 13:39:13 +0000 (14:39 +0100)]
Remove some bprintf, fix memleaks.
Martin Quinson [Mon, 7 Nov 2011 13:11:01 +0000 (14:11 +0100)]
fix the last broken tests by sanitizing the way properties are handled during the parse
Navarrop [Mon, 7 Nov 2011 12:52:30 +0000 (13:52 +0100)]
Resolved merge "merge changes. Strange, it seems to ate pierre commit" and
redo the missing part. Set the struct for get_route of tag cluster.
Navarrop [Mon, 7 Nov 2011 12:50:43 +0000 (13:50 +0100)]
Remove unused function.
Martin Quinson [Mon, 7 Nov 2011 11:14:15 +0000 (12:14 +0100)]
Further parser cleanups
Simply rename a bunch of symbols to make the module logic clear
Martin Quinson [Mon, 7 Nov 2011 11:04:31 +0000 (12:04 +0100)]
integrate properly the router creation in sg_platf module
Martin Quinson [Mon, 7 Nov 2011 10:47:11 +0000 (11:47 +0100)]
plug memleaks: surfxml does not allocate memory, sg_platf (and callbacks) do it on need
Martin Quinson [Mon, 7 Nov 2011 10:41:10 +0000 (11:41 +0100)]
Make the sg_plat_new_host() function public
Martin Quinson [Mon, 7 Nov 2011 10:35:08 +0000 (11:35 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Martin Quinson [Mon, 7 Nov 2011 10:31:15 +0000 (11:31 +0100)]
Continue parser cleanups
Prepare to split the interface of sg_platf in two: one public
interface (include/simgrid/platf.h) will contain the function
available to the user code to create elements while another interface
(src/include/simgrid/platf_interface.h) will contain what internal
modules need to declare callbacks on the element creation events.
For now, the files are still very empty, but I try to not break stuff
even further, so I commit soon.
Christophe Thiéry [Mon, 7 Nov 2011 10:03:09 +0000 (11:03 +0100)]
The parser's behavior has changed, revalidate the output of msg-trace
Christophe Thiéry [Mon, 7 Nov 2011 09:50:42 +0000 (10:50 +0100)]
Add some _XBT_GNUC_UNUSED to avoid compilation warnings with NDEBUG
Arnaud Giersch [Mon, 7 Nov 2011 06:40:43 +0000 (07:40 +0100)]
This include is not needed anymore.
Arnaud Giersch [Sun, 6 Nov 2011 17:18:21 +0000 (18:18 +0100)]
Remove useless code obfuscation.
Arnaud Giersch [Sun, 6 Nov 2011 10:54:58 +0000 (11:54 +0100)]
Fix handling of the backtrace log formatters %b and %B.
The first backtrace entry (_xbt_log_event_log) is not displayed.
Arnaud Giersch [Sun, 6 Nov 2011 10:50:18 +0000 (11:50 +0100)]
Cosmetics: fix comment.
Arnaud Giersch [Sat, 5 Nov 2011 22:53:23 +0000 (23:53 +0100)]
A message changed its category.
Arnaud Giersch [Sat, 5 Nov 2011 22:24:26 +0000 (23:24 +0100)]
Improve xbt_log_layout_format_doit().
Make it simpler and faster.
Not sure about the %b formatter, but who cares?
Arnaud Giersch [Sat, 5 Nov 2011 21:37:12 +0000 (22:37 +0100)]
Improve xbt_log_layout_simple_doit().
Make it simpler and slightly faster.
Arnaud Giersch [Sat, 5 Nov 2011 21:12:11 +0000 (22:12 +0100)]
Avoid (almost) duplicated code in xbt_log_layout_*.
Remove the dynamic log layout function.
The layout functions now try to build the log string in a buffer of given
length, and return success or failure.
The first try is with a buffer of fixed size.
In case of failure, the buffer is expanded, until it is large enough.
With this change, the code is much simpler, and the only penalty is with
very large log strings.
Arnaud Giersch [Sat, 5 Nov 2011 20:54:10 +0000 (21:54 +0100)]
Use fputs() instead of fprintf("%s") in file log appender.
It is more natural, and should be faster.
Arnaud Giersch [Sat, 5 Nov 2011 20:51:52 +0000 (21:51 +0100)]
Remove old debugging cruft.
Rationale: xbt_backtrace_display_current() uses fprintf(stderr) instead
of the logging facility.
Arnaud Giersch [Sat, 5 Nov 2011 20:50:41 +0000 (21:50 +0100)]
Fix the fixme.
And don't short-circuit the logging facility.
Martin Quinson [Sat, 5 Nov 2011 15:41:42 +0000 (16:41 +0100)]
using pointer to scalar is a strange way to return a value...
Martin Quinson [Sat, 5 Nov 2011 15:18:13 +0000 (16:18 +0100)]
Nobody will ever need to override the behavior of config parsing, so simplify the code
Martin Quinson [Sat, 5 Nov 2011 14:27:26 +0000 (15:27 +0100)]
Add a new module allowing to interact with the platform directly
* For now, it's super sparse, but the ongoing cleanup in the parser aims
precisely at populating it.
* Also, kill some oneline brain-dead functions in the lua console
Martin Quinson [Fri, 4 Nov 2011 21:34:04 +0000 (22:34 +0100)]
Call surf_parse_models_setup() from the right level
This function is in charge of initializing the models after all
configs have been parsed. It used to be called in the XML parser
callback, it is now called from the platform creation API, which is
intended to be called either from the XML parser or from the lua
console, or directly from the user code.
Martin Quinson [Fri, 4 Nov 2011 21:21:43 +0000 (22:21 +0100)]
Register the generic <AS> callback soon enough
- It is generic because it calls routing_AS_init() with the arguments
from AX_*
- soon enough is right at the beginning of the parsing process, not
deep after the models are inited
Martin Quinson [Fri, 4 Nov 2011 21:00:57 +0000 (22:00 +0100)]
fixing lua console (brain dead) damages to the parser
Martin Quinson [Fri, 4 Nov 2011 20:43:18 +0000 (21:43 +0100)]
this function should definitely not be called from lua, but from within the parsing logic
Martin Quinson [Fri, 4 Nov 2011 20:33:29 +0000 (21:33 +0100)]
Clean callbacks for the host parsing.
- do not rely on global field, but pass the parameter to the callbacks
- do not show the dynar containing the callbacks to the public, but
provide a specific function to add a callback not needing to see the
dynar
This breaks some tests, but I need to split my changes to see when I
improve the situation, and revert stuff when I degrade the situation.
I start to fix the tests right away.
Martin Quinson [Fri, 4 Nov 2011 19:29:50 +0000 (20:29 +0100)]
putain, ca compile pas cette merde
Martin Quinson [Fri, 4 Nov 2011 15:32:12 +0000 (16:32 +0100)]
merge changes. Strange, it seems to ate pierre commit
Martin Quinson [Fri, 4 Nov 2011 14:54:14 +0000 (15:54 +0100)]
yet another psychotic cleanup
Martin Quinson [Fri, 4 Nov 2011 14:53:46 +0000 (15:53 +0100)]
regenerate with flexml 1.9.2 (fix to my recent changes by ... agier ;)
Navarrop [Fri, 4 Nov 2011 14:20:36 +0000 (15:20 +0100)]
Create a structure to put link up and down for model cluster. Factorize the tag cluster.
Martin Quinson [Fri, 4 Nov 2011 13:32:26 +0000 (14:32 +0100)]
normalize some type names
Martin Quinson [Fri, 4 Nov 2011 13:09:22 +0000 (14:09 +0100)]
remove useless empty parsing callbacks
Martin Quinson [Fri, 4 Nov 2011 12:57:56 +0000 (13:57 +0100)]
kill a bunch of warnings that prevented me from compiling
Navarrop [Fri, 4 Nov 2011 12:47:44 +0000 (13:47 +0100)]
Without ! is better sorry...
Navarrop [Fri, 4 Nov 2011 12:17:15 +0000 (13:17 +0100)]
Add V_cluster_availability_file and V_cluster_state_file to struct for cluster.
Navarrop [Fri, 4 Nov 2011 11:54:57 +0000 (12:54 +0100)]
Do the free into the right file.
Navarrop [Fri, 4 Nov 2011 11:34:09 +0000 (12:34 +0100)]
Get rid of uselessly duplicated code.
Martin Quinson [Fri, 4 Nov 2011 10:59:43 +0000 (11:59 +0100)]
no need to reset a struct that will be freed
Navarrop [Fri, 4 Nov 2011 10:51:00 +0000 (11:51 +0100)]
Replace stupid char* by doubles in the cluster struct to avoid tons of useless printf/scanf.
Navarrop [Fri, 4 Nov 2011 10:43:37 +0000 (11:43 +0100)]
Add new model for cluster tag.
Martin Quinson [Fri, 4 Nov 2011 09:41:34 +0000 (10:41 +0100)]
ensure that the create_resource functions do call strdup by themselves
Also, ensure that the CPU models use surf_resource_new() instead of
dupplicating its code
Martin Quinson [Fri, 4 Nov 2011 08:19:13 +0000 (09:19 +0100)]
useless cosmetics
Martin Quinson [Thu, 3 Nov 2011 17:11:24 +0000 (18:11 +0100)]
damn. Make sure that the models are inited only once when parsing the same file twice, while ensuring that it still work when parsing after a SD_exit()/SD_init() reset (we have vicious test cases, of which 10 are still broken)