[pc386 BSP] ticker.exe issue (CVS head)

[Karel Gardas]

On Mon, 30 Jan 2006, Joel Sherrill wrote:

>
> Taking my time to see what runs and doesn't.  These are the compilers I
> had handy.  I don't have any 3.4.x i386-rtems compilers handy.
>
> RTEMS -- CVS head with serial port console and QEMU.  Same source
> for all.   All use binutils 2.16.1.
>
> WORKS:
> rtems-4.7-i386-rtems4.7-gcc-gcc3.3.5newlib1.12.0-2
> -- prints all time messages

Great! At least something! :-)

> FAILS:
> i386-rtems4.7-gcc (GCC) 4.0.1 (RTEMS 
> gcc-4.0.1-20050810/newlib-1.13.0-20050810-4)
> -- ends with register dump after three task prints
>
> WORKS but...:
>  i386-rtems4.7-gcc (GCC) 4.1.0 20060123 (prerelease)
>  newlib 1.14
> - Weird, prints on 0, 15 and 30 second boundary.  But runs.

Do you mean only TA3 task is running? If so, I've also seen this here -- 
this is if you add some debug print as I sent in the patch in this thread, 
you will get this.

> WORKS but ...:
> i386-rtems4.7-gcc (GCC) 4.2.0 20060124 (experimental)
> newlib 1.14
>  - Weird, prints on 0, 15 and 30 second boundary.  But runs.
>
> Do you get the same failure with all of your compilers?

No, usually 3.4.x bumbs on me with a lot of messages, while 4.x blocks and 
then corrupt VGA RAM with few chars. This is on real k6-2 machine using 
VGA as a console.

But anyway, nothing is able to run complete ticker.exe as it is for 
example the case with PSIM or edb7312 BSPs.

I'm now compiling 4.2-20060128 based toolchain, if it does not work, then 
I will try 3.3.5-based one.

Thanks for the info,
Karel
--
Karel Gardas                  kgardas at objectsecurity.com
ObjectSecurity Ltd.           http://www.objectsecurity.com