Question to OAR

[Ralf Corsepius]

On Tue, 2003-11-18 at 00:44, Dieter Schaefer wrote:
> > > Digging into that problem myself, I found the following: 
> > > * target=m68k-rtems, rtemsbsp=mcf5206elite
> > > * target=arm-rtems, rtemsbsp=edb7312 + vegaplus
> > > builds Ok, including tests 
> > > * powerpc-rtems, rtemsbsp=eth_comm + mbx8xx
> > > fails to build.
> > > Now, guess why ... If you need a hint, don't hesitate to ask ...
> > > Of course, It's fixed now.
> > And ... will you tell us?
> 
> sure, my hack is attached. Pls feel free to use it any way you like.
> .../cpukit/score/cpu/powerpc/Makefile.am 
> The problem for CygWin was that LIB in the line 
> $(LIB): $(OBJS) 
> was undefined. Linux was obviously more tolerant and skipped it.
Hmm, this is odd.

Though such make-rules are known to be non-portable, they are supposed
to be supported by gmake.

Are you sure to be using gmake?

What does
make --version
rsp.
gmake --version
report?

> This is because of modification to take care of powerpc old/new exceptions 
> I guess. All other targets have LIB= ... defined, because they have cpu.c 
> in there directories. For powerpc this is in the bsp directories.
Exactly.

> I know, about 2 years ago it was introduced. Is it still necessary 
> today to have old/new exceptions?
Unfortunately yes. We still have two competing exception processing
implementations for the powerpc.

>  Maybe things could be merged together 
> and unified - just a question.
Well, ... theoretically, yes, but I prefer keeping them separate, until
one these implementations dies out ... 
(Unifying them more or less would mean to concatenate the files and
enclosing them into #ifdefs)

> > RTEMS is an OpenSource project and if you want it to be improved
> > collaborate and contribute, or stay away.
> OpenSource means taking and giving ... I agree with that ...
Glad to hear this. 

Ralf