Martin Quinson [Tue, 13 Dec 2016 09:35:26 +0000 (10:35 +0100)]
rename the class objects and C types for the As -> NetZone transition
Martin Quinson [Tue, 13 Dec 2016 08:57:00 +0000 (09:57 +0100)]
clang-format changes related to the previous commit
Martin Quinson [Tue, 13 Dec 2016 08:55:07 +0000 (09:55 +0100)]
start renaming AS to NetZone
These things are obviously not Autonomous Systems as in Internet:
they form a hierarchy...
That's an intrusive change ongoing, but we want the best for SimGrid,
don't we?
Martin Quinson [Mon, 12 Dec 2016 21:26:04 +0000 (22:26 +0100)]
continue emptying the surf_routing files
Martin Quinson [Mon, 12 Dec 2016 21:13:27 +0000 (22:13 +0100)]
move some storage-related content out of routing into storage area
Martin Quinson [Mon, 12 Dec 2016 15:35:40 +0000 (16:35 +0100)]
sonar: try harder to not choke on casting pointers into literals
Martin Quinson [Mon, 12 Dec 2016 15:06:15 +0000 (16:06 +0100)]
cosmetics: .name().c_str() becomes .cname()
Martin Quinson [Mon, 12 Dec 2016 14:43:57 +0000 (15:43 +0100)]
sonar: ignore 'Pointers should not be cast to integral types' and bump version to 3.13.91
Martin Quinson [Mon, 12 Dec 2016 14:30:04 +0000 (15:30 +0100)]
don't fool sonar if we can avoid it
Martin Quinson [Mon, 12 Dec 2016 14:23:13 +0000 (15:23 +0100)]
these are redundent tests of which we already have the answer
Martin Quinson [Mon, 12 Dec 2016 14:19:50 +0000 (15:19 +0100)]
don't test pointers after dereferencing them (thanks, sonar)
Martin Quinson [Mon, 12 Dec 2016 09:39:38 +0000 (10:39 +0100)]
try to not disable our logs on sonar, just to see
Martin Quinson [Mon, 12 Dec 2016 07:59:30 +0000 (08:59 +0100)]
revalidate this test now that I updated the corresponding errmsg
Martin Quinson [Mon, 12 Dec 2016 07:57:07 +0000 (08:57 +0100)]
please sonar: dont apply binary operators to signed ints
Martin Quinson [Sun, 11 Dec 2016 21:14:07 +0000 (22:14 +0100)]
improve an error message
Martin Quinson [Sun, 11 Dec 2016 20:55:15 +0000 (21:55 +0100)]
typo
Martin Quinson [Sun, 11 Dec 2016 20:29:50 +0000 (21:29 +0100)]
please sonar, and other small cleanups
- Init members during initialization
- One operation per line at most
- Don't let fields uninited after the constructor
- Obey our naming schema (fields' name end with _)
Martin Quinson [Sun, 11 Dec 2016 20:18:46 +0000 (21:18 +0100)]
kill the s_surf_parsing_link_up_down_t datatype
Martin Quinson [Sun, 11 Dec 2016 20:07:06 +0000 (21:07 +0100)]
cosmetics
Martin Quinson [Sun, 11 Dec 2016 19:56:41 +0000 (20:56 +0100)]
use std::pair instead of s_surf_parsing_link_up_down_t for Cluster private links
Martin Quinson [Sun, 11 Dec 2016 18:22:12 +0000 (19:22 +0100)]
revalidate the surf_usage tests
These tests are meaningless anyway and should be replaced with unit
tests at some point. Hopefully soon.
Martin Quinson [Sun, 11 Dec 2016 17:48:59 +0000 (18:48 +0100)]
No need for the AS creation callback
Martin Quinson [Sun, 11 Dec 2016 17:48:29 +0000 (18:48 +0100)]
My detection of user errors creates some invalid accesses
Martin Quinson [Sun, 11 Dec 2016 13:56:16 +0000 (14:56 +0100)]
rename a field for cosmetics
Martin Quinson [Sun, 11 Dec 2016 13:13:49 +0000 (14:13 +0100)]
greatly simplify the way peers are created in Vivaldi AS
We used to create a mini cluster for each added peer, which is useless
since they are already in a Vivaldi AS. So instead, add the link
up&down to the Vivaldi containing AS.
This way, we can hide again the netcard of the AsImpl, that was made
public only a few commits ago.
Martin Quinson [Sat, 10 Dec 2016 21:55:46 +0000 (22:55 +0100)]
network/coordinates option is now useless
Martin Quinson [Sat, 10 Dec 2016 16:50:02 +0000 (17:50 +0100)]
Detect and diagnose a possible user error
Martin Quinson [Sat, 10 Dec 2016 15:54:48 +0000 (16:54 +0100)]
make our FindSimGrid.cmake more visible (and improve it)
Martin Quinson [Fri, 9 Dec 2016 21:39:00 +0000 (22:39 +0100)]
please sonar, praise sonar
Martin Quinson [Fri, 9 Dec 2016 21:21:28 +0000 (22:21 +0100)]
Use boost instead of xbt in Vivaldi
Martin Quinson [Fri, 9 Dec 2016 21:14:38 +0000 (22:14 +0100)]
use NetCard extension to store vivaldi coordinates
Martin Quinson [Thu, 8 Dec 2016 21:34:14 +0000 (22:34 +0100)]
make NetCards extendable, and use this for NS3 extension
Martin Quinson [Thu, 8 Dec 2016 20:36:07 +0000 (21:36 +0100)]
cosmetics
Martin Quinson [Thu, 8 Dec 2016 20:26:49 +0000 (21:26 +0100)]
use std::vector instead of xbt_dynar_t (in NS3)
This should fix quite a lot of leaks, too.
Martin Quinson [Thu, 8 Dec 2016 19:42:25 +0000 (20:42 +0100)]
NS3 extensions are for netcards, not for hosts
We need to have it for every netcard, be it an host, an AS or an
host-less router.
It's cumbersome to put it onto hosts when possible and onto netcard
for routers and ASes, and then search first in the host and if not
found in the netcards.
Instead, only put it on netcards, and we're set.
Martin Quinson [Thu, 8 Dec 2016 10:29:49 +0000 (11:29 +0100)]
move the NetCard related functions into a NetCard.cpp
Martin Quinson [Thu, 8 Dec 2016 09:21:09 +0000 (10:21 +0100)]
actually, BypassRoute can be made private to AsImpl
Martin Quinson [Thu, 8 Dec 2016 09:03:04 +0000 (10:03 +0100)]
rename AsRoute to BypassRoute, and move it to its own file
Martin Quinson [Thu, 8 Dec 2016 08:48:24 +0000 (09:48 +0100)]
Kill Onelink now that NS3 does not use it anymore
Martin Quinson [Thu, 8 Dec 2016 08:44:47 +0000 (09:44 +0100)]
make an As::onRouteCreation signal so that NS3 does not need getOneLinkRoutes afterward
Martin Quinson [Wed, 7 Dec 2016 19:29:45 +0000 (20:29 +0100)]
kill the last remainings of RoutingPlatf
Martin Quinson [Wed, 7 Dec 2016 18:16:37 +0000 (19:16 +0100)]
cosmetics
Martin Quinson [Wed, 7 Dec 2016 16:13:01 +0000 (17:13 +0100)]
less VM in surf.h
Martin Quinson [Wed, 7 Dec 2016 15:57:11 +0000 (16:57 +0100)]
Kill the fact that several model VM models may be provided.
- VM are a plugin: take it, copy it and change it yourself
- helps removing VMs from the trunk (I demand you to enter this plugin!)
Martin Quinson [Tue, 6 Dec 2016 20:32:29 +0000 (21:32 +0100)]
please sonar and use & when creating function pointers
Martin Quinson [Tue, 6 Dec 2016 20:16:50 +0000 (21:16 +0100)]
cosmetics
Martin Quinson [Tue, 6 Dec 2016 20:14:58 +0000 (21:14 +0100)]
move NetCard to its own header file
Martin Quinson [Tue, 6 Dec 2016 09:04:51 +0000 (10:04 +0100)]
no need for a NetCard and a NetCardImpl when both are in the same file
Augustin Degomme [Tue, 6 Dec 2016 08:32:27 +0000 (09:32 +0100)]
add -no-pie flag to thread sanitizer (see https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=796246)
Martin Quinson [Mon, 5 Dec 2016 23:06:06 +0000 (00:06 +0100)]
please my old friend SonarQube now that it's back
Evgeny Mandrikov [Mon, 5 Dec 2016 19:18:03 +0000 (20:18 +0100)]
SonarQube: upgrade scanner to the latest version 2.8
2.6.1 fixes https://jira.sonarsource.com/browse/SQSCANNER-18
so that warning message are properly displayed in log.
Evgeny Mandrikov [Mon, 5 Dec 2016 18:55:41 +0000 (19:55 +0100)]
SonarQube: remove useless properties "sonar.c.std" and "sonar.cpp.std"
When using build-wrapper this is inferred based on an actual build.
Martin Quinson [Mon, 5 Dec 2016 20:29:37 +0000 (21:29 +0100)]
please sonar: this variable will be overwritten in the next loop cycle
Martin Quinson [Mon, 5 Dec 2016 19:58:48 +0000 (20:58 +0100)]
try to fix a shell syntax error, bummer
Martin Quinson [Mon, 5 Dec 2016 19:45:09 +0000 (20:45 +0100)]
fix the syntax in the sonar configuration (fix #126)
I drop the #126 PR that propose to kill that config item all together
because I really think that sonar should not feel too much concerned
by our logging macros.
Yes they increase the nesting level if you expend them, but nobody
expends them mentaly. We use them as inline functions even if we
cannot implement them as functions for technical reasons.
Martin Quinson [Mon, 5 Dec 2016 19:31:34 +0000 (20:31 +0100)]
Don't show the sonar token in the logs (fix #127)
the token was regenerated, too.
Martin Quinson [Mon, 5 Dec 2016 19:27:06 +0000 (20:27 +0100)]
kill the now empty RoutingPlatf datatype \o/
Martin Quinson [Mon, 5 Dec 2016 08:18:05 +0000 (09:18 +0100)]
move the rootAS from routing_platf into the EngineImpl
Now that we instanciate a s4u::Engine when using the routing, the
surf_usage tests don't have to manually call surf_exit(): it called
atexit() during initialization. Gosh, it took me so long to figure out
why these two tests would break when the rootAs is moved...
Martin Quinson [Mon, 5 Dec 2016 10:03:46 +0000 (11:03 +0100)]
tiny optim
Martin Quinson [Mon, 5 Dec 2016 09:36:07 +0000 (10:36 +0100)]
Convert routing_platf->getRouteAndLatency into s4u::Host->routeTo
The bad thing is that the public API does not expose the surf::Link
type, so there is little you can do about it. For now.
Martin Quinson [Mon, 5 Dec 2016 09:45:02 +0000 (10:45 +0100)]
don't really use netCards in the network model. Keep at the host level
Martin Quinson [Sun, 4 Dec 2016 21:21:06 +0000 (22:21 +0100)]
fix java build
Martin Quinson [Sun, 4 Dec 2016 20:23:15 +0000 (21:23 +0100)]
cleanups in the dict and dynar tests
Martin Quinson [Sun, 4 Dec 2016 20:05:11 +0000 (21:05 +0100)]
stop testing heterogeneous dicts, too
Martin Quinson [Sun, 4 Dec 2016 17:47:09 +0000 (18:47 +0100)]
Deprecate xbt_dict_new()
Please people, use xbt_dict_new_homogeneous(). Non-homogeneous dicts
were a very bad idea that could have only originated in Perl anyway.
Martin Quinson [Sun, 4 Dec 2016 18:26:03 +0000 (19:26 +0100)]
don't pass the free_f parameter to MSG_*_set_property_value()
Martin Quinson [Sun, 4 Dec 2016 17:49:40 +0000 (18:49 +0100)]
Don't store raw objects as host properties, which must be strings
Martin Quinson [Sat, 3 Dec 2016 21:02:20 +0000 (22:02 +0100)]
cosmetics
Martin Quinson [Fri, 2 Dec 2016 22:59:06 +0000 (23:59 +0100)]
update a teshfile that was affected by #123, which Fabien just fixed
Fabien Chaix [Fri, 2 Dec 2016 13:30:57 +0000 (05:30 -0800)]
call updateRemaining when the action is suspended (fix #123)
if not, the LMM is giving the right results, but surf does not notice.
Thus the fix does not touch LMM, only surf_interface.cpp. Note that
suspend has the specificity that we cannot estimate when it will end.
Thus it does not seem appropriate to use the nextOccuringEventLazy
method which is essentially trying to find when will be the next step.
Instead I propose to call updateReaminingLazy from the suspend
function.
Martin Quinson [Fri, 2 Dec 2016 22:27:06 +0000 (23:27 +0100)]
make Boost.test optional, as I fail to find it on appveyor
Martin Quinson [Fri, 2 Dec 2016 08:10:37 +0000 (09:10 +0100)]
plug an obvious memleak, bummer
Martin Quinson [Thu, 1 Dec 2016 21:10:13 +0000 (22:10 +0100)]
Add boost unit_test_framework to our dependencies
Martin Quinson [Thu, 1 Dec 2016 21:11:19 +0000 (22:11 +0100)]
Merge branch 'master' of github.com:simgrid/simgrid
Martin Quinson [Thu, 1 Dec 2016 21:10:47 +0000 (22:10 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Martin Quinson [Thu, 1 Dec 2016 20:40:45 +0000 (21:40 +0100)]
cosmetics
Martin Quinson [Thu, 1 Dec 2016 20:10:09 +0000 (21:10 +0100)]
more informative assert message
Martin Quinson [Thu, 1 Dec 2016 20:09:29 +0000 (21:09 +0100)]
add an EngineImpl, unused so far
Martin Quinson [Mon, 28 Nov 2016 22:36:27 +0000 (23:36 +0100)]
cosmetics (and stop using routing_platf)
Martin Quinson [Mon, 28 Nov 2016 22:26:17 +0000 (23:26 +0100)]
gosh, so much cruft in this code
Martin Quinson [Thu, 1 Dec 2016 19:45:56 +0000 (20:45 +0100)]
Merge pull request #124 from mpoquet/master
[Energy platform parsing] assert + memory leak
Millian Poquet [Wed, 30 Nov 2016 13:08:01 +0000 (14:08 +0100)]
Merge branch 'master' of https://github.com/mpoquet/simgrid
Millian Poquet [Wed, 30 Nov 2016 12:58:18 +0000 (13:58 +0100)]
Energy: Fixed host parsing assertion + mem leak
When parsing the platform file with energy,
using the old watt description syntax
"watt_min:watt_max" instead of
"watt_idle:watt_min:watt_max" led to obscure
OOB error, the error message should now be
clearer.
Also found a little memory leak,
msg_idle was not freed.
Augustin Degomme [Tue, 29 Nov 2016 17:12:45 +0000 (18:12 +0100)]
disable java for MC FreeBSD builds, as it breaks mmalloc
degomme [Tue, 29 Nov 2016 14:09:12 +0000 (15:09 +0100)]
On Freebsd with clang > 3.8, sometimes we have a null name here.
Try to not segfault in this case.
degomme [Tue, 29 Nov 2016 11:56:29 +0000 (12:56 +0100)]
for some reason (const madness?) this lead to wrong names being printed.
Matthieu Volat [Tue, 29 Nov 2016 11:21:46 +0000 (12:21 +0100)]
Switch the event handling in the MC to libevent.
This is a good alternative to provide a clean socket/signal
handling, which allows to finaly enable the MC on FreeBSD too. The
downside is that now libevent>=2 is necessary to build the model
checker.
Thanks to degomme@ for the mentoring ;)
Matthieu Volat [Mon, 28 Nov 2016 14:09:18 +0000 (15:09 +0100)]
Add more libraries to blacklist.
FreeBSD system and LLVM related libraries.
Matthieu Volat [Mon, 28 Nov 2016 14:00:18 +0000 (15:00 +0100)]
BSD/SysV ptrace resume addr is 1, not 0 like in linux.
Martin Quinson [Mon, 28 Nov 2016 21:16:55 +0000 (22:16 +0100)]
improve the debug logs produced by the route seeking mechanism
Martin Quinson [Mon, 28 Nov 2016 20:29:35 +0000 (21:29 +0100)]
remove a ugly misplaced include
Martin Quinson [Mon, 28 Nov 2016 20:27:38 +0000 (21:27 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
degomme [Mon, 28 Nov 2016 13:35:21 +0000 (14:35 +0100)]
Fix umpire issues : MPI_Comm cannot be safely put inside an int
Fix/Ignore some warnings as well
Matthieu Volat [Sun, 27 Nov 2016 19:31:39 +0000 (20:31 +0100)]
More ptrace portability in model checker.
No PTRACE_O_TRACEEXIT equivalent in the original ptrace API, futher
work will indicate if this will be a problem.
Update message a bit when neither linux's or BSD's version are
available.
Matthieu Volat [Fri, 25 Nov 2016 08:36:33 +0000 (09:36 +0100)]
Mark shared libraries private clean pages as non-writable on FreeBSD.
This is how persmissions are with Linux, and later simgrid identifies
mappings based on path and permissions, so the less restistance path
is to present things more linux-like.
Matthieu Volat [Thu, 17 Nov 2016 12:03:32 +0000 (13:03 +0100)]
In FreeBSD map backend, mark stack same as in Linux
Matthieu Volat [Tue, 15 Nov 2016 20:02:45 +0000 (21:02 +0100)]
Add support to retrieve FreeBSD ucontext registers.
Sadly, no universal method as it depend on the CPU and the system
is free to implement whatever storage it see fit. Leave a terminating
error message for the next system to get ported.
Matthieu Volat [Tue, 15 Nov 2016 19:27:14 +0000 (20:27 +0100)]
Add alternative to remap() in PageStore when not available.
Code is a bit simplistic, having way to expand mappings but not
shrink them (so leaving them as they are), but it does the job.
Matthieu Volat [Tue, 8 Nov 2016 10:29:59 +0000 (11:29 +0100)]
Set waitpid() for model-checked flags according to platform.
Martin Quinson [Fri, 25 Nov 2016 09:59:53 +0000 (10:59 +0100)]
use the right name of an option in the doc