On 11/01/2008, Raphael Geissert wrote:
> There "MUST" be something wrong with the package then, how is that
> i386's and amd64's md5sum are exactly the same?
I don't see this that way. There *might* be a problem in your script or
so.
,---[ let's check ]---
| kibi@kitty1:/tmp/grub$ wget -q http://ftp.de.debian.org/debian/pool/main/g/grub/grub_0.97-29_i386.deb
| kibi@kitty1:/tmp/grub$ wget -q http://ftp.de.debian.org/debian/pool/main/g/grub/grub_0.97-29_amd64.deb
| kibi@kitty1:/tmp/grub$ for i in amd64 i386; do ar x grub_0.97-29_$i.deb control.tar.gz; tar xfz control.tar.gz; mv control control.$i; mv md5sums md5sums.$i; rm control.tar.gz; done
| kibi@kitty1:/tmp/grub$ diff -u md5sums.* | diffstat
| md5sums.i386 | 72 +++++++++++++++++++++++++++++------------------------------
| 1 file changed, 36 insertions(+), 36 deletions(-)
`---
Note that the i386 package has the following additional Depends line,
compared to the amd64 one.
| +Depends: libc6 (>= 2.5-5), libncurses5 (>= 5.4-5)
I guess your scripts are somehow assuming that if one has an empty
Depends line, the other has an empty line as well, or something similar.
Back to grub: Unfortunately, there's no amd64 log (source upload along
with the built binaries…), but it might be that ${shlibs:Depends}
weren't computed correctly or so, so that the Depends line was left
empty.
Indeed, checking a cowbuilder build log:
| dpkg-gencontrol: warning: unknown substitution variable ${shlibs:Depends}
In an amd64 chroot, looking closer:
| $ file debian/grub/usr/bin/mbchk
| usr/bin/mbchk: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), for GNU/Linux 2.6.8, statically linked, stripped
That's also the case for the other binaries, so it results apparently
(I'm no grub maintainer at all) correctly in an empty Depends: line.
> Did you notice the "(U)"? ;)
Yes. I actually expected somehow that you could have noticed that it was
about “grub”.
> I'm, again, sorry for those false positives (didn't expect them by
> comparing md5sums of two different architectures).
I'm not blaming because of false positives. I'd expect more common
sense. Either grub is architecture-dependent, being a low-level stuff,
probably written in C (I know, that might sound like a cliché, but…), or
it is just made out of supercowpowered architecture-independent shell
scripts, but then one might wonder a bit. Seen where it belongs in a
boot sequence?
Reviewing such a short list takes some minutes (to compare with the time
you spent on setting up these scripts), using the main measure when it
comes to being “Architecture: all” or “Architecture: any”: its *content*
(but you know that, I've been repeating this from the very beginning).
--
Cyril Brulebois
Attachment:
pgptc6sWOrho3.pgp
Description: PGP signature