While attempting to build perl in sbox using an armv5b toolchain I was get a
strange result when the perl debian/rules uses cpio to split up the packages.
cpio would always report 0 blocks at the end. I managed to narrow my failure
to the following steps that I could execute from the command line after I
made sure that debian/tmp was left intact with all files present:
[sbox-ixp: ~/crocodile/work/build/perl-5.8.4/debian/tmp] > sb-conf \
st -f -t sbrsh
[sbox-ixp: ~/crocodile/work/build/perl-5.8.4/debian/tmp] > echo \
"usr/bin/sperl5.8.4" | cpio -vdumpl ../build/perl-suid
/bin/cpio: usr/bin/sperl5.8.4 linked to ../build/perl-suid/usr/bin/sperl5.8.4
../build/perl-suid/usr/bin/sperl5.8.4
0 blocks
[sbox-ixp: ~/crocodile/work/build/perl-5.8.4/debian/tmp] > sb-conf \
st -f -t qemu-armeb
[sbox-ixp: ~/crocodile/work/build/perl-5.8.4/debian/tmp] > echo \
"usr/bin/sperl5.8.4" | cpio -vdumpl ../build/perl-suid
/bin/cpio: usr/bin/sperl5.8.4: Value too large for defined data type
0 blocks
I set myself up to debug cpio under qemu so I could step through the code that
led up to the "Value too large for defined data type" error when running
under qemu. I found that cpio is indirectly calling lstat() which I believe
is a function provided by the host kernel (not sure). So, I'm now wondering
if I am seeing a limitation in qemu or not.
I'm not sure if the "silent" failure seen when using sbrsh is the same issue
or not. Any suggestions on how to debug the sbrsh scenario?
Thanks,
Jon
_______________________________________________
Scratchbox-devel mailing list
[email protected]
http://lists.scratchbox.org/cgi-bin/mailman/listinfo/scratchbox-devel
