Re: Verify the integrity of Debian CDs-part 2
On Tue, 2005-05-10 at 23:56 +0000, alrenaud@swissinfo.org wrote:
> Different options in jigdo(or mkisofs)will probably generate
> slightly different .iso files,then different md5sums.If the
> different CD vendors assembling the ISOs use different options
> of the official (published) .iso files(md5sums),would be a
> waste of time trying to verify the integrity of a particular CD
> set.Not talking about potential problems that particular CD
> recording software could raise,or mounting /cdrom in a Linux
> file system.
> So a universal,consistent and straightforward way to verify
> the integrity of a particular set of CDs is needed.Am I wrong?
Generally speaking you are wrong. The verification method using md5sums
works perfectly well. While it is possible to derive different md5sums
on cds which have the same content by using different graft-points
option to mkisofs you will still get consistent results by getting the
md5sum from the same mirror you get the iso image. The same applies to
getting iso images with jigdo.
On very rare occasions you can get synchronisation errors on the jigdo
mirrors, ie you do the download of the jigdos before all of the
uploading is complete.
I am joining this thread late so I apologise if I go over any old
ground.
Firstly you should do an md5sum of the isos you have downloaded. Resolve
any failures before burning any cds.
Only then should you burn anything. If you get a different md5sum, then
your burn is wrong or your measurement method.
Every version of cdrecord that I have used over the past 3 years burns 2
sectors of 2048 bytes of nulls after the actual iso image on the cdr by
default.
My measurement method is this:
1. mount the cdr.
2. use df and get the count of 1k blocks on the cdr.
3. unmount the cdr.
4. Divide the 1k blocks by 2 to get no-of-sectors.
5. Invoke readcd thus:
readcd dev=/dev/hdx sectors=0-<no-of-sectors> f=- | md5sum
The dev option as above will likely work on 2.6.x kernels, for 2.4
kernels use cdrecord -scanbus to get the required bus,target,lun (eg
dev=0,3,0).
>
> Despite this,I am still interested in the published MD5SUM files
> for the weekly snapshots generated in April.In this page:
>
> http://cdimage.debian.org/pub/cdimage-testing/cd/jigdo-area/powerpc/
>
> Is there a place where I can find them?
You might try clicking on the link labelled MD5SUMS on that page.
Bob
Reply to: