Rebuilding RTEMS

Ralf Corsepius corsepiu at faw.uni-ulm.de
Wed Feb 14 10:58:27 UTC 2001


Eric Norum wrote:
> 
> Angelo Fraietta wrote:
> >
> > It appears that the linkcmds file is sometimes not being copied into
> > build-rtems/i386-rtems/pc386/lib directory during the make. I can copy it
> > over by hand and complete the build, but I shouldn't really have to.
> > Somewhere during the build process it must copy this file from the
> > appropriate startup directory. If that part fails, maybe the build should
> > stop there??
> > When is the file cpoied over?
> >
> 
> AHA!
> This is exactly what's been happening here, too, for the gen68360 and
> mvme167 targets.
> 
>
> Are you building on a Linux host with your RTEMS source being mounted
> from another machine via NFS?  I've found that the problem never occurs
> when I'm building with the source on a local disk. 
I am wondering if you can build any source with make.

> When I build with
> the source on an NFS mount the linkcmds (and *only* the linkcmds)
> doesn't get copied across and then the build dies when trying to build
> the first example application. 
The dependency rules being used for linkcmds isn't special at all.
Similar rules are used in almost any arbitrary Makefile in any
aribrary package.

Similar rules are used all over the place in RTEMS sources, with
linkcmds not being treated differently from many headers, bsp_specs
and *.a files.

> It seems to be due to some strange
> interaction between make and NFS.
Not unlikely. Different NFS implementations are known to have
problems in their interaction and sometimes to require special nfs
configurations on client and server side in general (esp. Linux <->
Irix/Solaris). 

All we could do from RTEMS's side is not to pickup linkcmds from the
preinstalled directories, but to pick it up from inside of the BSP's
source directory directly. However, I fear then the timestamp issue
will popup elsewhere in your environment and will not really help
you.

Ralf 

-- 
Ralf Corsepius 
Forschungsinstitut fuer Anwendungsorientierte Wissensverarbeitung
(FAW)
Helmholtzstr. 16, 89081 Ulm, Germany     Tel: +49/731/501-8690
mailto:corsepiu at faw.uni-ulm.de           FAX: +49/731/501-999  
http://www.faw.uni-ulm.de



More information about the users mailing list