[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Simulavr-devel] new joel simulavrxx autoconf tarball
From: |
Joel Sherrill |
Subject: |
Re: [Simulavr-devel] new joel simulavrxx autoconf tarball |
Date: |
Fri, 6 Mar 2009 06:15:47 -0600 |
User-agent: |
Thunderbird 2.0.0.19 (X11/20090105) |
Klaus Rudolph wrote:
Hi Joel,
I did a merge to actual cvs on my system and tried to do a
bootstrap/configure but run into trouble:
-------------
with_bfd_path = /home/zfrdh/work/binutils-2.19/build_avr
with_libiberty_path = check
nm: '/home/zfrdh/work/binutils-2.19/build_avr/lib/libbfd.a': No such file
configure: error: /home/zfrdh/work/binutils-2.19/build_avr DOES NOT
appear to be the location of the avr-built lib/libbfd.a or include/bfd.h
or is an incompatable version of the avr libbfd\n Please use the
--with-bfd-path=<path to your avr-built libbfd library>
I configure with a bfd path that has /bfd at the end of it.
./configure --with-bfd-path=/home/joel/work-gnat/svn/avr-work/b/bfd
It should simply fail if it can't find them. I will add that to
the list of issues.
FYI Eric W has tried it on Mingw and Cygwin and I think we are
getting close to it building on Cygwin. It is using BSD sockets
and Mingw does not have BSD sockets so that is beyond my autotools
adventure.
I am close to having it build even if you don't have Tcl and the
AVR cross tools installed. In that case, it can only build the
simulator w/o Tcl support. But that is useful.
-----------
The problem seems to be, that the default build of binutils has no path
"libs". On my build it is ".libs". A copy of libbfd is also direct in bfd.
Could you expand the tests as needed?
I will correct that.
After I fake that in my binutils path I could compile! Fine!
But now I have no simulavr.so and swig was not running.
So no example could run :-(
Whats my problem?
Probably just not having /bfd at the end of the path.
----
After fixing this I will test a bit and then I will write that to cvs.
But actually I am not able to do any test, because no executable is
available for the tests :-)
We are getting there. There are just a lot of machine
and human variations.
Thanks
Klaus
Some improvements since earlier this week. Hoping
it is approaching "good enough". The new version is
at:
http://www.rtems.org/ftp/pub/rtems/people/joel/simulavrxx/
Eric W. tried it on MingW and that lead to these changes:
- --disable-tcl added
- if tcl.h not found, disable tcl
I did the following
- add configure regression suite to autoconf
- runs it for "make check"
- passes all regression tests (no work on my part, just saying)
- detect ddd. If you have it, the various debug scripts
use it. Otherwise, they just invoke gdb directly.
Various things from Knut Schwichtenberg <address@hidden>
- stdiodemo (UART and LCD)
- LCD code improvements to work like real hardware
- added AT Keyboard example
This builds on Fedora 10. Might build elsewhere but
who knows. I only address one issue. :)