jrate-devel Mailing List for jRate

Hi Alex,

On Mon, Mar 20, 2006 at 07:41:13PM +0000, Alex Craddock wrote:
> i am trying to implement a program the uses the eventHandling functions 
> of jRate but i cannot seem to find a way to get it to catch the deadline 
> miss time here is the code i am using

Sorry for the delay in my response.

Right, deadline miss handler support for async events hasn't been
implemented.  (The support for deadline misses for periodic realtime
threads was revamped last year, and just hasn't yet been extended to
event handling yet.)  I can probably get support for this into the CVS
version soon if it's important to you.

Morgan
-- 
Morgan Deters
md...@mo...

Hi Trevor,

On Wed, Mar 22, 2006 at 01:19:04PM -0800, Trevor Harmon wrote:
> This is a user question, not a devel question, but since there isn't  
> a user mailing list for jRate, I'm posting to jRate-devel.

There is, actually, though it's not hosted through sourceforge and
isn't very active.  Subscription info at
http://jrate.sf.net/MailingLists.php

> I know the FAQ states that jRate can run on any POSIX OS, but my  
> understanding is that one needs a true RTOS in order to have, for  
> example, a proper implementation of priority inheritance. A non-RTOS  
> doesn't provide the thread-level guarantees that are necessary for  
> true RT, correct? If so, what is considered to be the best RT kernel  
> for use with jRate? (I'm a student and don't have funding to purchase  
> a commercial OS, so I'm limited to the free alternatives such as  
> rtlinuxfree.com. Are any of these "good enough" for jRate?)

I'll share my experiences and observations.  Many use a standard
vanilla Linux/pthreads setup despite the lack of respect for priority
etc.  This does work for many apps without very strict timing
requirements, and it's the easy route.  However, not only is priority
inheritance not supported in such a setup, priorities aren't always
even respected.  I've seen low-priority threads running in preference
to high-priority ones on both NPTL and LinuxThreads outside of the
usual priority inversion scenario.

Minimally, to support priority inheritance, you need RTNPTL.  As an
additional benefit, RTNPTL is careful in other places to ensure that
high-priority threads always run in preference to low-priority ones.
There are some notes on setting up jRate with RTNPTL in the INSTALL
file in the distribution.  It requires you to build a patched kernel
and patched glibc, then point jRate to it when you ./configure.

Also, there are several patches to the Linux kernel to make it
maximally preemptible and low-latency.  In particular I suggest
looking at Ingo Molnar's realtime-preempt patch.  I've played with
this thing in the lab and verified some really amazing timing
characteristics.  Never played with it on embedded though.
http://people.redhat.com/~mingo/realtime-preempt/

There is also some interest in getting jRate to work with RTAI/fusion.
This hasn't been attempted in earnest yet though AFAIK.  Presumably
jRate will function under RTAI/fusion, but since jRate depends on
signal handling etc. it will only operate in "compatibility" mode (not
hard-RT).  "Porting" jRate to use the subset of POSIX functionality
supported in RTAI/fusion in hard-RT is an outstanding task.  See the
jrate-devel list archives (last August or so) for some discussion of
this.

(In the past, of course, TimeSys kernels were popular.  Less so now,
since they aren't so keen on making it easily downloadable anymore.)

The specific solution you find that works will also depend on support
for your target board too.  This can confuse things considerably,
since you may end up trying to combine kernel patches made for
different kernel versions.

Morgan
-- 
Morgan Deters
md...@mo...

On Mar 28, 2006, at 3:43 AM, Borislav Grigorov wrote:

> I am new to the Real Time programming. Recently I was able to build  
> jRate under Kubuntu Linux and my idea to use Eclipse to write  
> compile and run Real Time programs. Is it possible and can you tell  
> me what I need to do so that eclipse uses all the jRate features.

jRate is just a compiler for RTSJ, and the syntax of RTSJ programs is  
exactly like normal Java programs. So, if you use Eclipse's normal  
Java support, that's pretty much all you need (and all Eclipse can  
offer).

Compilation is another thing, but here again, jRate is just a  
modified version of GCJ. However, I'm not sure what kind of support  
Eclipse has for GCJ, if any.

Trevor

Hi=20
I am new to the Real Time programming. Recently I was able to build jRate=
 under Kubuntu Linux and my idea to use Eclipse to write compile and run =
Real Time programs. Is it possible and can you tell me what I need to do =
so that eclipse uses all the jRate features.

Thank you,=20
Best wishes Borislav Grigorov


-----------------------------------------------------------------
http://Host.GBG.bg - =C1=E8=E7=ED=E5=F1 =F0=E5=F8=E5=ED=E8=FF =E7=E0 =F5=EE=
=F1=F2=E8=ED=E3 =E8 =E4=EE=EC=E5=E9=ED =E8=EC=E5=ED=E0

This is a user question, not a devel question, but since there isn't  
a user mailing list for jRate, I'm posting to jRate-devel.

I know the FAQ states that jRate can run on any POSIX OS, but my  
understanding is that one needs a true RTOS in order to have, for  
example, a proper implementation of priority inheritance. A non-RTOS  
doesn't provide the thread-level guarantees that are necessary for  
true RT, correct? If so, what is considered to be the best RT kernel  
for use with jRate? (I'm a student and don't have funding to purchase  
a commercial OS, so I'm limited to the free alternatives such as  
rtlinuxfree.com. Are any of these "good enough" for jRate?)

Thanks,

Trevor

Hello,

i am trying to implement a program the uses the eventHandling functions 
of jRate but i cannot seem to find a way to get it to catch the deadline 
miss time here is the code i am using



            AperiodicParameters Missrelease = new 
AperiodicParameters(new RelativeTime(0,1),new 
RelativeTime(0,1),turnOmiss,turnOmiss);
            turnO = new AsyncEvent(Missrelease);
            AperiodicParameters turnMiss    = new 
AperiodicParameters(new RelativeTime(1,1),new 
RelativeTime(1,1),turnOmiss,turnOmiss);
           
            TurnOn                 handlerOn     = new TurnOn(new 
PriorityParameters(1),turnMiss,cruise);   
           
            AperiodicParameters turnODeadMiss    = new 
AperiodicParameters(new RelativeTime(0,1),new 
RelativeTime(0,1),turnOmiss,turnOmiss);
       
            activateCC = new ThreadedAsyncEventHandler(new 
PriorityParameters(15),
                                            turnODeadMiss,
                                            null,
                                            null,
                                            null,
                                            handlerOn);
          turnO.addHandler(activateCC);



----------------------------

turnO is of type AsyncEvent
TurnOn extends the RealtimeThread Class
turnOmiss   extends ThreadedAsyncEventHandler ( i have also tried 
BoundAsyncEventHandler)
activateCC is of type AsyncEventHandler
cruise you dont need to worry about, it is just a parameter i pass for 
that class.

any input to this problem would be helpful, in order to test this i use:

   turnO.fire()

which when fired puts the thread to sleep for 1000ms

Thanks for any help

Berto.

Trevor,

On Tue, Mar 14, 2006 at 03:00:38AM -0800, Trevor Harmon wrote:
> On Mar 10, 2006, at 11:20 PM, Morgan Deters wrote:
> 
> >My current plan is to limp along this way until I bring jRate up to
> >GCC 4.x, at which point the problem goes away.  The only real thing
> >keeping me from doing that in the past is that I wanted to put some
> >AIE support in while we're still at 3.3.x -- that's really the last
> >major compiler enhancement necessary -- and then in one effort jRate
> >can fully support all of the 3.3 and 3.4 lines, as well as 4.0.
> 
> What is AIE? I'm not familiar with this term, and Google doesn't seem  
> to be, either.

AsynchronouslyInterruptedException.  Trying to save my wrists by
abbreviating it.  :-)

Morgan
-- 
Morgan Deters
md...@mo...

On Mar 10, 2006, at 11:20 PM, Morgan Deters wrote:

> My current plan is to limp along this way until I bring jRate up to
> GCC 4.x, at which point the problem goes away.  The only real thing
> keeping me from doing that in the past is that I wanted to put some
> AIE support in while we're still at 3.3.x -- that's really the last
> major compiler enhancement necessary -- and then in one effort jRate
> can fully support all of the 3.3 and 3.4 lines, as well as 4.0.

What is AIE? I'm not familiar with this term, and Google doesn't seem  
to be, either.

Trevor

Hi Trevor,

On Fri, Mar 10, 2006 at 05:08:18PM -0800, Trevor Harmon wrote:
> Any thoughts on this? Anyone here using GCC 4 to build jRate?

Yeah, I've always had trouble using GCC 4.x to build GCC 3.3.x.  I've
got a gcc-3.4 binary installed on all my boxen, though, and I've just
been setting CC=gcc-3.4 when I ./configure.

My current plan is to limp along this way until I bring jRate up to
GCC 4.x, at which point the problem goes away.  The only real thing
keeping me from doing that in the past is that I wanted to put some
AIE support in while we're still at 3.3.x -- that's really the last
major compiler enhancement necessary -- and then in one effort jRate
can fully support all of the 3.3 and 3.4 lines, as well as 4.0.

Admittedly, that's probably less useful now than it might have been a
year ago, especially in light of the zillion (approximate figure)
bugfixes made to GCC's Java bits since 3.3.  No one really wants to
use the old 3.3.x line for Java development.

Take care,
Morgan
-- 
Morgan Deters
md...@mo...

I was building jRate on a Fedora Core 4 machine, and I ran into a  
problem that I believe occurs due to an incompatibility between the  
libiberty bundled in jRate and GCC 4, which is shipped with FC4. I  
was able to fix the problem, but it's a bit of a hack, and if there  
are already plans to move to a more recent libiberty -- perhaps  
there's something already in CVS? -- then it's probably best just to  
hold off rather than submit a patch for the issue.

Any thoughts on this? Anyone here using GCC 4 to build jRate?

Trevor

Hello,

On Fri, Mar 10, 2006 at 05:48:59PM +0100, jou...@en... wrote:
> Could you help me solving those problems. I thank you in advance for your
> answer.

Sure.  Most often, this is caused by the program using the standard
Java library (libgcj.so.4) that you have installed system-wide, which
is probably not a jRate one (or the wrong version, or something).
You'll need to set your LD_LIBRARY_PATH to $prefix/lib.  For example,
if you configured jRate to install itself in /opt/jrate, then you
would

  export LD_LIBRARY_PATH=/opt/jrate/lib:$LD_LIBRARY_PATH
  ./a.out

or whatever.  If you don't want to have to specify the LD_LIBRARY_PATH
at runtime, you can make the application prefer certain library
directories with the -rpath linker option, which you can pass to the
linker by using the -Wl option to jRate-gcj.  For example:

  jRate-gcj --main=Main -Wl,-rpath,/opt/jrate/lib *.java

The resulting binary will look in /opt/jrate/lib before /usr/lib,
/lib, /usr/local/lib, etc.  You can use the "ldd" command to see which
objects will be linked in when you run the application.

In a future release, we'll probably rename the jRate version of the
installed libgcj object so this issue doesn't come up.

HTH,
Morgan
-- 
Morgan Deters
md...@mo...

Dear Sir,

I would like to use jrate under a Debian Linux. I had no problem with the
configure/make/make install. Then I tried to compile some of the demos
(HelloWorld and MemoryArea) and I had no error. However, when I try to
execute the output files, it returns the following errors :

concerning HelloWorld:

jRate-gcj --main=3DHelloWorld HelloWorld.java
./a.out
	Segmentation fault


concerning MemoryArea:

jRate-gcj --main=3DMemoryAreaDemo MemoryAreaDemo.java
./a.out
	./a.out: relocation error: ./a.out: undefined symbol:
_ZN5javax8realtime10MemoryArea6class$E


Could you help me solving those problems. I thank you in advance for your
answer.

Best Regards.

JOUAULT Vincent