Next cygwin/powerpc Problem (Was RE: Getting reacquainted -- Cygw in/powerpc-rtems Build Error)

[Ralf Corsepius]

On Tue, 2005-01-04 at 13:03 -0500, Mills, John M. wrote:
> Mark - Thanks for your note.
> 
> -----Original Message-----
> From: Mark Dean [mailto:M.Dean at sstl.co.uk]
>  ...
> Subject: RE: Getting reacquainted -- Cygwin/powerpc-rtems Build Error
> 
> MD-    Looks to me that you are trying exactly the config we're trying to
> use. I've had
> MD- nothing but problems with RTEMS/cygwin. If you get a configuration
> working I'd be REALLY 
> MD- REALLY interested to know what versions of RTEMS and the tools you used
> (and any 
> MD- patches).
> 
> I'll put a summary together.
> 
> I got much further when I rebuilt my tools and RTEMS-4.6.2 with couple of
> minor changes in my scripts:
> 
>  1. I disabled NLS for gcc (probably not an issue),
nls becomes an issue if installing several different versions of GCC in
parallel.

It therefore is disabled in all RTEMS toolchain binaries.

>  2. I did not '--enable-shared' for gcc (this wasn't available to build for
> PPC anyway),
Correct. --enable-shared does not make any sense for RTEMS.

>  3. I enabled 'version-specific-runtime-libs' (probably not an issue yet),
--enable-version-specific-runtime-libs is a must if you want to install
several GCCs in parallel to the same prefix.

It therefore is enabled by all RTEMS toolchain binaries.

>  4. For the RTEMS build I specified '--prefix=<my_toolchain_prefix>' (which
> _was_ probably
>     the real key).
?? I fail to understand.

The prefix can be chosen to almost arbitrary values.


> NEXT PROBLEM:
> All PPC BSP targets now build and install successfully _except_ MVMV2307,
> mcp750, and mtx603e. These targets all fail to build 'tests/tmtests/tm27',
> but they all build OK if this subdirectory is removed from the build list!
> 
> The problems all seem to arise from one included file: <bsp.h>, giving
> errors such as: 
> 
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> +++++++++++
> powerpc-rtems-gcc --pipe -B../../../../../lib/
> -B../../../../../mvme2307/lib/ -specs bsp_specs -qrtems -mcpu=604 -mmultiple
> -mstring -mstrict-align -meabi -msdata=eabi  -O4 -g -Wall -ansi -fasm
> -o o-optimize/tm27.nxe o-optimize/task1.o
> ../../../../../mvme2307/lib/no-dpmem.rel
> ../../../../../mvme2307/lib/no-event.rel
> ../../../../../mvme2307/lib/no-msg.rel ../../../../../mvme2307/lib/no-mp.rel
> ../../../../../mvme2307/lib/no-part.rel
> ../../../../../mvme2307/lib/no-signal.rel
> ../../../../../mvme2307/lib/no-timer.rel
> ../../../../../mvme2307/lib/no-rtmon.rel 
> o-optimize/task1.o(.text+0x29c): In function `Task_1':
> /usr/src/rtems-4.6.2/c/src/tests/tmtests/tm27/task1.c:92: undefined
> reference to `Install_tm27_vector'
> o-optimize/task1.o(.text+0x2b0):/usr/src/rtems-4.6.2/c/src/tests/tmtests/tm2
> 7/task1.c:105: undefined reference to `Cause_tm27_intr'
>  [... similar errors elided ...]
> o-optimize/task1.o(.text+0x430): In function `Isr_handler':
> /usr/src/rtems-4.6.2/c/src/tests/tmtests/tm27/task1.c:257: undefined
> reference to `Clear_tm27_intr'
>  [... elided ...]
> o-optimize/task1.o(.text+0x524):/usr/src/rtems-4.6.2/c/src/tests/tmtests/tm2
> 7/task1.c:268: undefined reference to `Cause_tm27_intr'
> collect2: ld returned 1 exit status
> make[5]: *** [o-optimize/tm27.exe] Error 1
> make[5]: Leaving directory
> `/tmp/build/rtems/powerpc-rtems/c/mvme2307/tests/tmtests/tm27'
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

> >>>> QUESTION: Does this ring a bell for anyone?
Bugs in these BSPs - Probably already fixed in RTEMS-CVS.

> MD-    The weirdest problem I've had is that gcc builds ok but takes 3
> hours!
I guess you are referring to powerpc-rtems-gcc. Yes, building it takes
long, and it probably takes very much longer if building it on an
emulator such as Cygwin.

But where's the actual problem? Unless you are developing on GCC, you
normally build it once and then use it for a longer period. ;)

>  it then won't 
> MD- compile the powerpc-rtems bsp with various error messages.

There is no powerpc-rtems BSP. There are several powerpc-rtems BSPs,
some of which are bugged and ought to be fixed.

Ralf