Martin Quinson [Tue, 8 Oct 2019 07:30:56 +0000 (09:30 +0200)]
Add a TODO in the code
Martin Quinson [Tue, 8 Oct 2019 06:57:17 +0000 (08:57 +0200)]
please clang-static by initializing p_idle only once
Arnaud Giersch [Tue, 8 Oct 2019 13:43:48 +0000 (15:43 +0200)]
Try to understand where/when space is missing on travis+windows.
Frederic Suter [Wed, 9 Oct 2019 12:32:33 +0000 (14:32 +0200)]
change handling of nested <prop> declarations.
Before: distinctions between zone props and other props with a boolean
and a single set of properties at a time.
Issue: As host can declare props and disks that in turn may declare
props, it enforces a declaration order: disks then props.
Now: stack a new property set when parsing a opening tag to store props
declared at this level. When closing a tag, we assign the set at the
top the stack where it has to and remove it from the stack.
Martin Quinson [Tue, 8 Oct 2019 05:52:01 +0000 (07:52 +0200)]
deal with rounding errors on 32bits
Martin Quinson [Mon, 7 Oct 2019 21:48:13 +0000 (23:48 +0200)]
more informative error message
Martin Quinson [Mon, 7 Oct 2019 21:44:55 +0000 (23:44 +0200)]
java: obey to sonar, use nio.file.delete for better error messages
Martin Quinson [Mon, 7 Oct 2019 20:24:27 +0000 (22:24 +0200)]
fix makedist
Martin Quinson [Mon, 7 Oct 2019 20:14:39 +0000 (22:14 +0200)]
Further doc improvements
This fixes https://github.com/simgrid/simgrid/pull/330
Martin Quinson [Mon, 7 Oct 2019 19:42:36 +0000 (21:42 +0200)]
tesh 'expect signal' can now accept more than one potential signal
This is mainly for MacOSX which may raise a SIGBUS on stack overflow
instead of SIGSEGV as on Linux.
Martin Quinson [Mon, 7 Oct 2019 18:52:37 +0000 (20:52 +0200)]
cosmetics in the doc of the Host energy plugin
Martin Quinson [Mon, 7 Oct 2019 15:43:01 +0000 (17:43 +0200)]
cosmetics: cpu_load is not supposed to be >1 anymore
Martin Quinson [Mon, 7 Oct 2019 15:40:37 +0000 (17:40 +0200)]
remove a now useless conversion script
The C++ code will hint how to do the conversion. That may be time
consuming for someone wanting to convert a large file manually, but
I'm not sure of how long we'd need that script around. And I think
that nobody has such a long XML file (but the PR authors) so that
should be OK.
Martin Quinson [Mon, 7 Oct 2019 15:27:28 +0000 (17:27 +0200)]
host_energy: rework the code converting the old syntax to the new one
Write it very similarly to the python script provided in the MR.
Just because I don't trust myself with this conversion code.
Martin Quinson [Mon, 7 Oct 2019 14:35:08 +0000 (16:35 +0200)]
Remove another occurence of the now removed 'release' cmake variable
Martin Quinson [Mon, 7 Oct 2019 08:35:58 +0000 (10:35 +0200)]
small sonar cleanup
Augustin Degomme [Mon, 7 Oct 2019 08:01:09 +0000 (10:01 +0200)]
update changelog
Frederic Suter [Mon, 7 Oct 2019 06:59:26 +0000 (08:59 +0200)]
improve output (consistent with similar s4u example)
Martin Quinson [Sun, 6 Oct 2019 22:28:34 +0000 (00:28 +0200)]
Also attach our signal handler to SIGBUS, because MacOSX sometimes raises it on stack overflow
Martin Quinson [Sun, 6 Oct 2019 22:12:32 +0000 (00:12 +0200)]
Merge similar signals, and other cleanups
This change set is a bit large and some bits are brutal, but it
escalated quickly and I went out of patience to finish it.
Martin Quinson [Sun, 6 Oct 2019 21:54:15 +0000 (23:54 +0200)]
Improve the doc of plugins
Martin Quinson [Sun, 6 Oct 2019 09:48:20 +0000 (11:48 +0200)]
sonar: useless cosmetics
Martin Quinson [Sat, 5 Oct 2019 23:53:35 +0000 (01:53 +0200)]
fix that java example
Martin Quinson [Sat, 5 Oct 2019 23:49:42 +0000 (01:49 +0200)]
New doc section on plugins
Martin Quinson [Sat, 5 Oct 2019 21:50:01 +0000 (23:50 +0200)]
Revert "Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug"
The SimGrid way to request non-optimized builds is
-Denable_compile_optimizations=OFF
I agree that this is not very user-friendly, and your commit looked
nice, but it actually broke the build as enable_debug was set to false
in release mode. As a result, -DNDEBUG was passed to the compilation,
removing all xbt_assert() and leading to many compilation issues since
our code does not compile without the asserts.
This reverts commit
a4d4e2488e05031945fa3f1dd887e34c05cfa948.
Martin Quinson [Sat, 5 Oct 2019 21:36:58 +0000 (23:36 +0200)]
kill the now unused 'release' cmake config variable
Martin Quinson [Sat, 5 Oct 2019 21:14:43 +0000 (23:14 +0200)]
Install a conversion path for our users wrt host_energy plugin
This is related to https://github.com/simgrid/simgrid/issues/189
Martin Quinson [Sat, 5 Oct 2019 21:12:13 +0000 (23:12 +0200)]
Proposition of a Release content
Martin Quinson [Sat, 5 Oct 2019 20:23:01 +0000 (22:23 +0200)]
Document recent changes
Martin Quinson [Thu, 26 Sep 2019 22:27:06 +0000 (00:27 +0200)]
Julien EMMANUEL [Fri, 4 Oct 2019 14:08:34 +0000 (16:08 +0200)]
Fix a doc error about actors (Tutorial_algorithms)
Martin Quinson [Sat, 5 Oct 2019 19:57:28 +0000 (21:57 +0200)]
Nitpicking the reviews
MiB is for mebibyte, 1024^2 bytes and other unimportant changes.
Gene Cooperman [Fri, 4 Oct 2019 20:34:25 +0000 (16:34 -0400)]
Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
Gene Cooperman [Wed, 2 Oct 2019 21:27:29 +0000 (17:27 -0400)]
Minor polishing English: Configuring_Simgrid.rst
Augustin Degomme [Thu, 3 Oct 2019 23:24:08 +0000 (01:24 +0200)]
oops
Augustin Degomme [Thu, 3 Oct 2019 19:15:57 +0000 (21:15 +0200)]
timer on windows WSL seems less precise and was counting 0ns for some events... which were then skipped as < 1e-6s, and the trace changed
Augustin Degomme [Thu, 3 Oct 2019 14:26:50 +0000 (16:26 +0200)]
add test for call location with ti traces/replay/adjustment/sleep/computing=0.
Idea is to replay twice a trace with an adjustment file meant to speedup the sleep (and only it, to avoid variability).
It's clearly not realistic, but timings should be reproducible (or not..).
Augustin Degomme [Thu, 3 Oct 2019 14:24:01 +0000 (16:24 +0200)]
replay : allow to really use --cfg=smpi/simulate-computation with replay.
Remove it from the teshes that had it wrongly activated.
It should not break, as it should have been useless before.
Augustin Degomme [Thu, 3 Oct 2019 13:36:23 +0000 (15:36 +0200)]
Add handling of sleep in replay.
We can trace sleep events with --cfg=tracing/smpi/sleeping. Now we can replay them as well.
Even if 1 second is 1 second, let's allow to accelerate/slow these down artificially with smpi/comp-adjustment-file.
This should allow to test the feature.
Augustin Degomme [Thu, 3 Oct 2019 12:57:19 +0000 (14:57 +0200)]
changelog
Augustin Degomme [Thu, 3 Oct 2019 12:57:12 +0000 (14:57 +0200)]
Have trace-call-location work with filenames and not full paths.
It's much easier this way to use adjustment files, and name collisions in a given code are not common.
If it's still needed, previous behaviour can be restored using "smpi/trace-call-use-absolute-path" option
Augustin Degomme [Thu, 3 Oct 2019 10:11:57 +0000 (12:11 +0200)]
Add call_location to ti-trace and replay as well.
A new "location" event in the trace is printed
Martin Quinson [Sun, 29 Sep 2019 16:41:47 +0000 (18:41 +0200)]
Merge branch 'rework-energy-plugin' into 'master'
Rework energy plugin
See merge request simgrid/simgrid!14
Martin Quinson [Fri, 27 Sep 2019 22:01:57 +0000 (00:01 +0200)]
Merge branch 'master' into 'rework-energy-plugin'
# Conflicts:
# include/simgrid/plugins/energy.h
Arnaud Giersch [Fri, 27 Sep 2019 07:17:49 +0000 (09:17 +0200)]
Fix warning: `content' may be used uninitialized.
Augustin Degomme [Thu, 26 Sep 2019 15:28:44 +0000 (17:28 +0200)]
as said in the test : The MPI standard says that there is no way to use MPI_Request_free safely with receive requests
Arnaud Giersch [Thu, 26 Sep 2019 14:33:56 +0000 (16:33 +0200)]
Fix build on openindiana.
On openindiana, getopt() is declared in <stdio.h>, <stdlib.h>, and <unistd.h>.
Include all of them to avoid build errors when it's #define'd to smpi_getopt.
Frederic Suter [Thu, 26 Sep 2019 14:19:55 +0000 (16:19 +0200)]
[sonar] fix some bugs and smells related to disk addition
Arnaud Giersch [Thu, 26 Sep 2019 13:30:59 +0000 (15:30 +0200)]
Fix bug # + whitespace cleanup.
Arnaud Giersch [Thu, 26 Sep 2019 13:21:58 +0000 (15:21 +0200)]
Update .mailmap.
Arnaud Giersch [Thu, 26 Sep 2019 13:03:27 +0000 (15:03 +0200)]
[sonar] Make variable private.
Arnaud Giersch [Thu, 26 Sep 2019 08:20:59 +0000 (10:20 +0200)]
[sonar] Put parentheses around macro parameters.
Arnaud Giersch [Thu, 26 Sep 2019 07:57:35 +0000 (09:57 +0200)]
[sonar] Avoid const_cast.
Arnaud Giersch [Thu, 26 Sep 2019 07:51:52 +0000 (09:51 +0200)]
There's no more need for ActorPtr there.
Arnaud Giersch [Wed, 25 Sep 2019 19:35:42 +0000 (21:35 +0200)]
Use plain pointers in internal structures, and hopefully fix more memory leaks.
Arnaud Giersch [Wed, 25 Sep 2019 19:00:42 +0000 (21:00 +0200)]
Kill unused vector.
Arnaud Giersch [Wed, 25 Sep 2019 19:47:04 +0000 (21:47 +0200)]
Prefer sh to bash.
Frederic Suter [Wed, 25 Sep 2019 18:14:55 +0000 (20:14 +0200)]
introduces <disk> and s4u::Disk
Frederic Suter [Wed, 25 Sep 2019 18:14:27 +0000 (20:14 +0200)]
Merge branch 'master' into disk
Frederic Suter [Wed, 25 Sep 2019 18:14:13 +0000 (20:14 +0200)]
Merge branch 'master' of https://framagit.org/simgrid/simgrid
Martin Quinson [Wed, 25 Sep 2019 16:11:23 +0000 (18:11 +0200)]
regenerate smpi_extended_traces.h and friends after the addition of MPI_Bsend and friends
Martin Quinson [Wed, 25 Sep 2019 16:09:59 +0000 (18:09 +0200)]
Kill TRACE_smpi_finalize() and have the container unregistered as soon as the actor leaves
Of course this requires to revalidate some tesh files, as the ending
date is impacted by this change.
Frederic Suter [Wed, 25 Sep 2019 15:36:52 +0000 (17:36 +0200)]
Merge branch 'master' into disk
Frederic Suter [Wed, 25 Sep 2019 15:36:39 +0000 (17:36 +0200)]
Merge branch 'master' of https://framagit.org/simgrid/simgrid
Augustin Degomme [Wed, 25 Sep 2019 13:31:39 +0000 (15:31 +0200)]
remove trace finalize from process ending, cleanup is done by trace_end anyway.
Martin Quinson [Wed, 25 Sep 2019 11:00:50 +0000 (13:00 +0200)]
fix a memleak coming from cyclic references in smart pointers
Before that change, the SMPI extension of actors got a smart reference
onto the S4U actor. That prevented S4U actors from being garbage
collected, since a SMPI extension is freed (and its smart references
released) only when the S4U actor is destroyed.
The change of this commit simply replace smart references with dumb
pointers to break the deadlock in the reference cycle, so that s4u
actors get collected after use.
Martin Quinson [Mon, 23 Sep 2019 19:30:57 +0000 (21:30 +0200)]
document last change
Frederic Suter [Tue, 24 Sep 2019 15:57:26 +0000 (17:57 +0200)]
Merge branch 'master' into disk
Frederic Suter [Tue, 24 Sep 2019 15:56:56 +0000 (17:56 +0200)]
Merge branch 'master' of https://framagit.org/simgrid/simgrid
Augustin Degomme [Tue, 24 Sep 2019 15:41:53 +0000 (17:41 +0200)]
warning --
Augustin Degomme [Tue, 24 Sep 2019 14:27:09 +0000 (16:27 +0200)]
move newly implemented calls out of the 'not yet implemented' section
Augustin Degomme [Tue, 24 Sep 2019 14:26:45 +0000 (16:26 +0200)]
changelog update
Augustin Degomme [Tue, 24 Sep 2019 14:18:25 +0000 (16:18 +0200)]
activate bsend tests.
Modify one to stop spending too much time sleeping as we test with bogus platforms.
Augustin Degomme [Tue, 24 Sep 2019 14:16:13 +0000 (16:16 +0200)]
add MPI_Bsend, MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach.
fun fact: we are not actually using the provided buffer in SMPI, as we use our own buffers internally for all detached calls (bsend means detached).
Martin Quinson [Mon, 23 Sep 2019 19:29:04 +0000 (21:29 +0200)]
Merge branch 'master' into 'master'
Improving the performance of the NS-3 bindings by:
See merge request simgrid/simgrid!18
bcamus [Mon, 23 Sep 2019 16:38:27 +0000 (18:38 +0200)]
Improving the performance of the NS-3 bindings by:
+ Stop moving NS-3 forward when there is no flow, and synchronizing NS-3 with SimGrid only when a new flow starts instead.
+ Cancelling the Stop event we scheduled in NS-3 if the simulator stops before the event, to prevent untimely stops.
Martin Quinson [Mon, 23 Sep 2019 09:48:42 +0000 (11:48 +0200)]
Start preparing for a release
Martin Quinson [Mon, 23 Sep 2019 09:36:28 +0000 (11:36 +0200)]
Don't mandate C++14 as some build daemons still don't have it
Martin Quinson [Mon, 23 Sep 2019 09:04:31 +0000 (11:04 +0200)]
fix make dist (again and again)
Martin Quinson [Mon, 23 Sep 2019 08:49:42 +0000 (10:49 +0200)]
MC: switch to xxhash as a (fast) hashing function.
Mandates C++14 but should be faster according to
https://aras-p.info/blog/2016/08/02/Hash-Functions-all-the-way-down/
Martin Quinson [Mon, 23 Sep 2019 07:55:56 +0000 (09:55 +0200)]
Add a fast hash function (mandates C++14 for MC)
This comes from
https://github.com/RedSpah/xxhash_cpp/commit/
a310676e1514b49d78f0c9ced0678efea4f82fd9
This requires C++14:
xxhash.hpp:288:40: error: variable declaration in a constexpr function is a C++14 extension
So bump the dependencies of the Model Checker to C++14. That's an
intrusive move, but MC is experimental anyway, and C++14 isn't that
modern anymore.
Martin Quinson [Mon, 23 Sep 2019 00:36:33 +0000 (02:36 +0200)]
fix make dist
Martin Quinson [Mon, 23 Sep 2019 00:34:34 +0000 (02:34 +0200)]
Make this test more verbose to improve debuggability
Martin Quinson [Mon, 23 Sep 2019 00:06:47 +0000 (02:06 +0200)]
Merge the two wifi tests together
Martin Quinson [Sun, 22 Sep 2019 23:19:48 +0000 (01:19 +0200)]
Merge the two wifi platform files
And rename the result to simply 'wifi.xml'
Martin Quinson [Sun, 22 Sep 2019 23:09:02 +0000 (01:09 +0200)]
modernize these platform files
Martin Quinson [Sun, 22 Sep 2019 22:55:54 +0000 (00:55 +0200)]
integrate the wifi unit tests
Martin Quinson [Sun, 22 Sep 2019 22:31:59 +0000 (00:31 +0200)]
reduce the verbosity of this test and increase its effectiveness
This is merely a crash test
Martin Quinson [Fri, 20 Sep 2019 19:37:58 +0000 (21:37 +0200)]
Remove the Codacy badge since we unplugged that project
Augustin Degomme [Fri, 20 Sep 2019 14:37:53 +0000 (16:37 +0200)]
add missing platform file in the dist
Martin Quinson [Fri, 20 Sep 2019 13:38:37 +0000 (15:38 +0200)]
Document recent changes
Augustin Degomme [Fri, 20 Sep 2019 13:28:31 +0000 (15:28 +0200)]
add NS3 files to the dist even when NS3 is not configured
Martin Quinson [Fri, 20 Sep 2019 12:57:08 +0000 (14:57 +0200)]
Merge branch 'master' into 'master'
NS-3: unblock the number of communications + detect when several flows are simultaneously finishing
See merge request simgrid/simgrid!17
Frederic Suter [Fri, 20 Sep 2019 10:27:58 +0000 (12:27 +0200)]
mention closing of GH#207
Frederic Suter [Fri, 20 Sep 2019 08:24:06 +0000 (10:24 +0200)]
Merge branch 'master' into disk
Frederic Suter [Fri, 20 Sep 2019 07:44:25 +0000 (09:44 +0200)]
rework a bit and typo fixes
Frederic Suter [Thu, 19 Sep 2019 16:14:27 +0000 (18:14 +0200)]
update documentation
Frederic Suter [Thu, 19 Sep 2019 14:55:43 +0000 (16:55 +0200)]
Merge branch 'master' into disk
Frederic Suter [Thu, 19 Sep 2019 14:54:56 +0000 (16:54 +0200)]
fix formatting
bcamus [Thu, 19 Sep 2019 14:21:12 +0000 (16:21 +0200)]
adding a test to check if the NS-3 bindings can handle flows that simultaneously end