[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: gzip: ./tmp/netboot/vmlinuz-2.6.26-1-sparc64: not in gzip format



On Sunday 23 November 2008, Manoj Srivastava wrote:
> On Sun, Nov 23 2008, Frans Pop wrote:
> > On Sunday 23 November 2008, Bastian Blank wrote:
> >> sparc still uses kernel-package for building, so I would not count
> >> on anything.
> >
> > Right. So RC bug there.
>
>         File it, with details.

Even though I really was not planning to, I've just done so (#506711).

I'm again disappointed that nobody else took the trouble to even try to
trace this to its cause [1]. Especially as it turned out to be completely
trivial.

Given the error
    gzip: ./tmp/netboot/vmlinuz-2.6.26-1-sparc64: not in gzip format
and util/tftpboot.sh (based on the fact that Geert fixed it there):

<snip>
kernel=$1
[...]
case "$arch" in
    arm* | i386 | mips | mipsel)
        cp $kernel $tftpimage.tmp
        ;;
    *)
        echo "uncompressing kernel"
        gzip -cd $kernel > $tftpimage.tmp
        ;;
esac
</snip>

So, the kernel gets passed in as the first parameter.

A simple grep on tftpboot.sh leads to config/sparc/netboot.cfg, which has:
tftpboot.sh $(TEMP_KERNEL) $(TEMP)/System.map.gz $(TEMP_INITRD) $(TEMP_BOOT)

A short look at the Makefile shows:
$(TEMP_KERNEL): $(STAMPS)tree-unpack-$(targetstring)-stamp

So it looks like the kernel is simply unpacked from the linux-kernel udeb.

As there have been no obvious relevant changes in D-I SVN but there *has*
been a very recent upload of kernel udebs, the conclusion is simple. Waldi
provided the missing link to kernel-package.

There is absolutely NO NEED that I should be doing this. Next time I
really won't. Of course, if others try, fail, and report what they have
done and where they got stuck, I'll be quite happy to offer suggestions
or, if all else fails, to give it a try myself.

Just to be clear: handling the cleanup of this issue is up to others.
See the BR I filed. The fix Geert has apparently done in his local buildd
checkout is NOT a solution and should be reverted.

Cheers,
FJP

[1] Geert's effort is of course very much appreciated, but only took things
    half way. Without that I would not have looked at all.

Attachment: signature.asc
Description: This is a digitally signed message part.


Reply to: