Philippe Gerum wrote:
Jan Kiszka wrote:


here are some patches and comments on the simulator "quick"-start
process. I guess the patches are due to latest changes in trunk, and I
cannot guarantee that they are clean - but the simulator compiles now.

Regarding the rest of its build and start process: There are two hidden
requirements, namely gcc < 4 (due to -fwritable-strings)

Normally, availability of --fwritable-strings should be auto-detected by the configure script. So there is something wrong going there, I will look at it.

 and gdb 5.x. A

I've been unlucky with GDB 6.x too, but did not check the latest release; I'm currently using 5.3 when running simulations.

note in sim/README about them would be great. Or do you have a fix for
those issues scheduled?

--fwritable-strings should be fixed. I'm not sure if the requirement for GDB 5.x still applies though.

Well, once you got this far (i.e. the xenoscope is built and an ancient
gdb has been prepared), the next step is to produce a test program. Am I
overseeing some example?

Look into ksrc/skins/{vxworks,psos+}/demos. The Makefile also produces a simulation executable for satch. Then run "xenoscope satch_sim"

 Fortunately, I had that demo I once produced

(last year), but that one was easier to build. I first had to drag in
the libs and includes of both the simulator and a regular xenomai
installation (I think to remember this once worked without this
add-ons). Then I need this additional trick in the simulators include dir:

mkdir asm
ln -s <xeno-inst-dir>/include/asm-sim asm/xenomai

This maps #include <asm/xenomai/system.h> to the correct "arch".
Shouldn't this be created automatically during the simulator
installation? Or is something else broken on my box?

You clearly should not need this trick to compile apps against the simulation environment; the instrumenter (gcic) sets the proper flags. I'm going to rebuild one and let you know.

Confirmed. You don't need this hand-made link.



Xenomai-core mailing list

Reply via email to