4.9 Testing Request

[Robert S. Grimes]

Ralf Corsepius wrote:
> On Wed, 2008-09-17 at 21:11 -0400, Robert S. Grimes wrote:
>   
>> I'm having trouble installing the 4.9-testing code with 4.10 tools.  It
>> builds just fine, but fails during the installation.
>>
>> Here is the end of the build output:
>>
>>  /usr/bin/install -c -m 644
>> '../../../../../../rtems-cvs-20080911/c/src/../../cpukit/posix/inline/rtems/posix/barrier.inl'
>> '/usr/rtems/4.9/powerpc-rtems/virtex/lib/include/rtems/posix/barrier.inl'
>>  /usr/bin/install -c -m 644
>> '../../../../../../rtems-cvs-20080911/c/src/../../cpukit/posix/inline/rtems/posix/rwlock.inl'
>> '/usr/rtems/4.9/powerpc-rtems/virtex/lib/include/rtems/posix/rwlock.inl'
>>  /usr/bin/install -c -m 644
>> '../../../../../../rtems-cvs-20080911/c/src/../../cpukit/posix/inline/rtems/posix/spinlock.inl'
>> '/usr/rtems/4.9/powerpc-rtems/virtex/lib/include/rtems/posix/spinlock.inl'
>> test -z "/usr/rtems/4.9/powerpc-rtems/virtex/lib" || /bin/mkdir -p
>> "/usr/rtems/4.9/powerpc-rtems/virtex/lib"
>>  /usr/bin/install -c -m 644 'libposix.a'
>> '/usr/rtems/4.9/powerpc-rtems/virtex/lib/libposix.a'
>>  powerpc-rtems-ranlib '/usr/rtems/4.9/powerpc-rtems/virtex/lib/libposix.a'
>> /bin/bash: line 4: powerpc-rtems-ranlib: command not found
>>     
>
>
>   
>> Seems it can't find powerpc-rtems-ranlib, yet it does exist, and in the
>> same directory as the other tools, so the path is correct.  Any ideas?
>>     
>
> Which --target did you pass to configure?
>
> powerpc-rtems-ranlib indicates you to have used --target=powerpc-rtems
>   
Yes, you are correct.

> This should be --target=powerpc-rtems4.9 (to use the powerpc-rtems4.9
> toolchain) rsp. --target=powerpc-rtems4.10 (to use the powerpc-rtems4.10
> toolchain).
>   
Is this true in an absolute sense, or by convention?  Let me explain 
what I mean.

First, I've never used more than one set of tools, one target, or even 
one RTEMS version at a time, and what I've used has worked fine.  Now, I 
am attempting to use two different environments (my existing 4.8.0-based 
and the 4.9-testing configurations), so maybe my approach is not "best 
practice".  Anyway, my 4.8.0 environment was set up using 
--prefix=/usr/rtems/4.8 for the install point, and 
--target=powerpc-rtems when configuring the tools and BSP.  So, for 
example, the compiler was named "powerpc-rtems-gcc", and it lives in 
"/usr/rtems/4.8/bin".  My PATH environment variable is set to 
"/usr/rtems/4.8/bin:...", and RTEMS_MAKEFILE_PATH is set to 
"/usr/rtems/4.8/powerpc-rtems/virtex".

This works fine, so I was trying to set up the tools and 4.9-testing 
using the same approach:

  target =>  <cpufamily>-rtems
  installpoint => /usr/rtems/<version>
  tools => /usr/rtems/<version>/<target>

So different versions of tools (and I guess, different targets too) are 
stored in separate locations, and the choice as to which to use is based 
on the PATH environment setting.  This is the same approach suggested 
for the selection of BSP via the RTEMS_MAKEFILE_PATH variable.

You seem to suggest I should name the tools with the version number, 
which I don't have a strong opinion about.  You guys do a lot more of 
this multi-version, multi-target stuff than I do, so I'm happy to use 
"best practices" - once I understand them, of course!  :-)

But is that the cause of my problem?  Wouldn't things have broken long 
before, when powerpc-rtems4.10-gcc was not found?

I must be mistaken in my understanding here...

> Background: Somewhat oversimplified, the TARGET specified in
> --target=TARGET is the toolchain prefix of the toolchain to be used.
>   
Yes, I understood that.
> Ralf
>
>   
Thanks,
-Bob