[BusyBox] Re: Busybox version 0.44 on Strongarm ( ARM processor)

Erik Andersen andersen at lineo.com
Thu Jun 22 19:16:31 UTC 2000


On Thu Jun 22, 2000 at 09:33:38AM -0400, Modi, Al wrote:
> Erik, attached below is my comment and questions below your comments and
> suggestion. Thanks again for your assistance in this.

No problem.  Glad to try and help.

> > Hmm. Can your compiler compile other executables that do run properly? Can
> > you compile up and run ash? Does it help if you change the makefile to not
> > use optimization, i.e.:
>
> Erik, I have compiled linux kernel using this cross-compiler tools and
> several other small application. I can down load ash and try to compile using
> the tools I have and see how it works. I will change the make file to comment
> out the OPTIMIZATION = line on the make file and replace it with OPTIMIZATION
> = . I will try to see whether it helps.

Ok, cool.

> > Well, this definitely looks like the executable isn't coming out correctly.
> > You didn't include bres620, but instead included bres619 (which wasn't what
> > you intended I think). bout620 looks fine.
> 
> I am sorry, I made a mistake. Attached is the bres620 file. This file shows
> that modules cp_mv.c and regexp.c had some warnings, I am not sure whether
> this is normal or not. What do you think about this warning ?

Yes, these warnings are normal.  Don't worry about them.

> > I did just find yesterday that there are some lines in the Makefile that set
> > CFLAGS= instead of CFLAGS+=, and these lines cancelled out these settings. I
> > fixed these in CVS. Perhaps this is causing some problems. Could you try the
> > latest Makefile from CVS?
> 
> I am not sure when you say that "Could you try the latest Makefile from CVS
> ?". I obtained this file from the ftp site
> .ftp://ftp.lineo.com/pub/busybox/busybox-0.44pre5.tar.gz  Today I went to
> this ftp site and I did not find the file busybox-0.44pre5.tar.gz  any more I
> do find the new file busybox-0.45.tar.gz with date of 06/22/00. I will get
> this file and try it. Is there another way to get a busybox tar file ? I
> don't know any other way except to go to the ftp server. Thanks.

There was a new release of BusyBox yesterday, which was why 0.44pre5
disappeared.  busybox-0.45.tar.gz is now the latest released version (and it
has the Makefile fixes I described above).  BusyBox now has a mailing list, bug
tracking, and public CVS access to the latest development source code.  You
probably should subscribe to the busybox mailing list.  For details about CVS
access, see http://busybox.net/cvs.html

> Yesterday, I copied a version of the mount utility that support mounting of
> loop devices on a ramdisk. On my arm development system (brutus board). When
> kernel try to execute rc.sysint script, I got a following error. 
> 
> mount: permission fault on section address = 0xfffffffc, > code 2
> segmentation fault ( core dumped) This error is similar to what I showed with
> busybox. The error was gone when I used ramdisk that has following library
> different in the /lib directory. I did not build mount utility, mount utility
> was provided to me. Does this helps you in any way with the problem I am
> seeing with the busybox ? 

Hmm.  So apps that you did not compile yourself also segfault the same way
busybox does on your arm devel board.  What does 'ldd' show for this mount app?
What does 'ldd' show on busybox when statically linked?  When dynamically
linked?


> Ramdisk /lib contained that provide me the similar error to busybox is as follows:
> 
> ld-2.1.2.so
> ld-linux.so.2 -> ld-2.1.2.so
> ld.so
> libc-2.1.2.so
> libc.so.4 -> libc.so.4.6.27
> libc.so.4.6.27
> libc.so.6 -> libc-2.1.2.so
> 
> Ramdisk /lib contained that eliminated this error is as follows:
> 
> ld-2.1.so
> ld-linux.so.2 -> ld-2.1.so
> ld.so
> libc-2.1.so
> libc.so.4 -> libc.so.4.6.27
> libc.so.4.6.27
> libc.so.6 -> libc-2.1.so

> By looking at this it seems to me that mount utility needed a different set
> of library. Does this helps you any way to assist me with the problem that I
> am seeing ?

Well.  I am currently thinking that somehow the set of libraries you are
compiling against (those provided by your arm cross compiler?) are broken.
When you dynamically link busybox, and place it in the ramdisk with the new set
of libraries that eliminated the mount error, does it work?

> Q.1 When busybox is build with --static switch does it really matter what is
> the content of /lib directory on my ramdisk ? My understanding of --static
> flag is that when modules, utility etc.. are build using --static switch,
> ithe content of /lib directory does not matter. Is this true ? If you have
> knowledge on this could you please explain this to me. Once again thanks in
> advance for all your assistance in this.

Your understanding of --static is correct.

 -Erik

--
Erik B. Andersen   email:  andersen at lineo.com
--This message was written using 73% post-consumer electrons--





More information about the busybox mailing list