Re: [ECOS] Re: Printf function error

[Kelvin Lawson <klawson at ad-holdings dot co dot uk>]

Hi Aniket,

There are a few issues that are fixed for SuperH in the latest sources 
in CVS. To be sure you have all the fixes it might be best to download 
them. However here are some things I had to get round with v2.0:

The new toolchain generated subsections under bss (bss.XXX) - these were 
missed by the SH linker script. See the patch: 
http://sources.redhat.com/ml/ecos-patches/2003-07/msg00087.html

The toolchain also seems to treat common symbols differently. If you are 
building the FreeBSD stack you may have problems linking because of 
icmpstat appearing in multiple places. The workaround for this was to 
declare it in ip_icmp.c and make it an extern for other object files.

Again with the FreeBSD stack, if you're building the httpd monitor 
(monitor.c), you may have to fix some strings that are split over 
multiple lines.

I also had stack alignment problems fixed by:
http://sources.redhat.com/ml/ecos-discuss/2003-04/msg00089.html

Cheers,
Kelvin.


Aniket S Joshi wrote:
> Hi Kelvin,
>
> Thanks for your valuable & Informative reply.
> Sorry for the late response.
> Now I am updating the GCC with new version.
> meanwhile pls send me the details of changes required for ECOS 2.0 so that after GCC updation i will implement that changes in code also.
> Are you still working on Sh-7729 + ECOS combination?
>
> Thanks once again
>
>
> Regards
>
> Aniket
>
>
> > > > Kelvin Lawson <klawson@ad-holdings.co.uk> 02/20/04 08:04PM >>>
>
> Hi Aniket,
>
> I found the same issue with GCC 3.2.1 on SH3. I also found that in some 
> cases it generated bad code with optimisation turned off. Since moving 
> to GCC 3.3.2 I haven't come across any code generation problems. I'd 
> recommend you download the latest sources and rebuild the toolchain.
>
> I used these instructions for building the toolchain:
> 	http://ecos.sourceware.org/build-toolchain.html 
>
> But used the latest version of the sources:
> 	. binutils-2.14
> 	. newlib-1.12.0
> 	. gcc-core-3.3.2
> 	. gcc-g++-3.3.2
> 	. insight-6.0
>
> I had to apply the following patches to GCC and Insight:
> 	. http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13068 
> 	. http://gcc.gnu.org/ml/gcc-patches/2002-07/msg00433.html 
> 	. http://sources.redhat.com/ml/insight/2003-q4/msg00034.html 
>
> I also found some issues with the eCos v2.0 release when using the new 
> sh-elf toolchain. I would use the latest eCos sources from CVS in order 
> to avoid these. I can send you details of what changes are required to 
> v2.0 if you like though.
>
> Cheers,
> Kelvin.
>
>
> John Dallaway wrote:
>
>
> > Hi Aniket
> >
> > Aniket S Joshi wrote:
> >
> >
> > > I am tring to run the example program "towthreads.c" But it is giving 
> > > output as follows
> > >
> > > Entering twothreads' cyg_user_start()
> > > function
> > > Beginning execution; thread data is 0
> > > Beginning execution; thread data is 1
> > > Thread 00000000000000000000000000000000000000..........
> >
> >
> > This is a manifestation of a code optimisation issue in GCC 3.2.1 (and 
> > possibly other versions) targetting sh-elf. The eCosPro Developer's Kit 
> > for SuperH targets ships with a patched toolchain to address this problem.
> >
> > John Dallaway
> > eCosCentric Limited



--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss