still problem with new/old interrupts on PowerPC and CVS ?

[Joel Sherrill <joel@OARcorp.com>]

Ralf Corsepius wrote:
> On Thu, 2005-08-04 at 23:33 -0600, e.stiebler wrote:
> 
>>Ralf Corsepius wrote:
>>
>>>On Thu, 2005-08-04 at 12:48 -0500, Joel Sherrill  wrote:
>>>
>>>
>>>>Ralf.. libcpu/irq.h is in the staging area of the build directory but 
>>>>not at the install point:
>>>
>>>I could reproduce the problem and believe to have fixed it in CVS-trunk.
>>>
>>>Ralf
>>
>>Looks good, came much further. Hello_World get compiled, but :
>>
>>make[1]: Entering directory `/SHARED1/RTEMS/examples/filerdback'
>>test -d o-optimize || mkdir o-optimize
>>powerpc-rtems-gcc --pipe 
>>-B/SHARED1/RTEMS/rtemsdocs-4.6.99/powerpc-rtems/ss555/lib/ -specs 
>>bsp_specs -qrtems   -g -Wall  -O4 -fno-keep-inline-functions -g 
>>-mcpu=505 -Dmpc555 -Dss555     -c   -o o-optimize/test.o test.c
>>test.c:12:20: error: memory.h: No such file or directory
>>make[1]: *** [o-optimize/test.o] Error 1
>>make[1]: Leaving directory `/SHARED1/RTEMS/examples/filerdback'
>>make: *** [all] Error 1
> 
> OK, we are facing several problems at once.
> 
> 1. memory.h is an BSD'ism (Read it is non-portable), portable code
> should "#include <string.h>", instead.
> I.e. this code should be updated.
> 
> 2. RTEMS memory.h is part of libnetworking, and is only being installed
> when building RTEMS with networking enabled.
> This is a long withstanding defect in RTEMS, I am struggling with for
> quite a while (The problem is far more general than just memory.h).
> 
> 3. The examples suffer from  bit-rot :)
> 
> For the moment, try building the example with the offending "#include
> <memory.h>" removed. If this raises warnings or errors, try replacing it
> with "#include <string.h>"
> 
> AFAIS, repeated_opens/test.c probably will expose the same problem.

I build all the samples with networking enabled 99% of the time.
I suppose I need to start a sweep with POSIX enabled but everything
else off and then with everything optional disabled.

> Ralf

--joel