Re: Two more corrupt .debs found from 2.2_rev0 images
On 20 Aug 2000, Philip Hands wrote:
[...]
> If I were to redo any/all of the CDs, it would only be after I replace
> almost all the hardware on the cdimage.d.o machine, and then check the
> checksums of all the archive mirror.
>
> I may be able to do that tomorrow, if we decide we need to urgently
> regenerate the images, otherwise I'll probably do the hardware upgrade
> Tuesday, and then just wait for 2.2_rev1 to come around.
>
> What do people think? It seems that the images we have are pretty
> much OK, but is that good enough? I think we should aim for
> perfection, and in the past there has been a tendency to say that we
> cannot muck about with things after release, with the end result being
> that our users ended up suffering fixable bugs for much longer than
> was really necessary.
[...]
> Opinions?
I checked the pdksh from i386 Bin-2 against that on ftp.d.o. As I more-or-less
expected, it's a single-bit error:
< 0064d0 c9 59 ee 0b 75 aa 32 3d 94 52 ed dc ee c9 c8 aa (from CD)
---
> 0064d0 c9 59 ee 0b 75 aa 32 3d 94 52 ed dc ee c9 c0 aa (from FTP)
This can only have happened during mkisofs, since Packages.* and md5sum.txt on
the CD have the correct md5sum.
So, IMHO the most correct solution would not be to generate new images, but
to correct this problem at the binary level (where it originated) by simply
toggling the offending bit in the image. Then calculate and publish the new
checksum, with a warning on the "last-minute-notes" page. Since a workaround
for the "problematic CDs" is easily available (download pdksh and dpkg -i,
only 212 kB), I don't think it's worth the trouble of doing anything more.
(I.e. I wouldn't bump the CD's revision for just one incorrect bit; there's
also no need for apt-cdrom/dpkg-multicd etc. to know that this bit was
wrong/is corrected.)
In fact, I've just corrected my private copy of the Bin-2 image (how I love
mc(1)! Offset 0x148554de), the new md5sum of the image is:
be04cd6d17159d66978ad227b26ed17d binary-i386-2.iso
Time to fix: <10 min, most of that used by cp and md5sum, with the guarantee
of not introducing new bugs. I can easily do it on cdimage.d.o, too, if it's
up again.
I don't know yet about the powerpc problems (don't have the images anywhere
handy), but I suspect those can be solved just as easily.
Regards,
Anne Bezemer
Reply to: