[virt-tools-list] [PATCH libosinfo 1/3] Fill out media for all Ubuntu distros

[Zeeshan Ali (Khattak)]

On Thu, Feb 23, 2012 at 1:25 AM, Daniel P. Berrange <berrange at redhat.com> wrote:
> On Thu, Feb 23, 2012 at 12:44:15AM +0200, Zeeshan Ali (Khattak) wrote:
>> On Tue, Feb 21, 2012 at 7:46 PM, Daniel P. Berrange <berrange at redhat.com> wrote:
>> > From: "Daniel P. Berrange" <berrange at redhat.com>
>>
>> I really should have paid more attention before ACKing some of the
>> patches. :) Here is the problem introduced by this one:
>>
>> > ---
>> >  data/oses/ubuntu.xml |  731 ++++++++++++++++++++++++++++++++++++++++++++++++--
>> >  1 files changed, 711 insertions(+), 20 deletions(-)
>> >
>> > diff --git a/data/oses/ubuntu.xml b/data/oses/ubuntu.xml
>> > index 78239f6..fb331a3 100644
>> > --- a/data/oses/ubuntu.xml
>> > +++ b/data/oses/ubuntu.xml
>> > @@ -6,6 +6,35 @@
>> >     <version>4.10</version>
>> >     <vendor>Canonical</vendor>
>> >     <family>Linux</family>
>> > +
>> > +    <media arch="i386">
>> > +      <url>http://old-releases.ubuntu.com/releases/warty/warty-release-install-i386.iso</url>
>> > +      <iso>
>> > +        <volume-id>Ubuntu 4.10 i386</volume-id>
>> > +        <system-id>LINUX</system-id>
>> > +      </iso>
>> > +      <kernel>casper/vmlinuz</kernel>
>> > +      <initrd>casper/initrd.img</initrd>
>> > +    </media>
>> ..
>> > +
>> > +    <media arch="i386" live="true" installer="false">
>> > +      <url>http://old-releases.ubuntu.com/releases/warty/warty-release-live-i386.iso</url>
>> > +      <iso>
>> > +        <volume-id>Ubuntu 4.10 i386</volume-id>
>> > +        <system-id>LINUX</system-id>
>> > +      </iso>
>> > +      <kernel>casper/vmlinuz</kernel>
>> > +      <initrd>casper/initrd.img</initrd>
>> > +    </media>
>>
>> With volume and system ID being identical for both medias, only the
>> first media will be matched against both types of medias. Same goes
>> for other medias you added in this patch. Wonder what could be a
>> solution here though. Perhaps detection code should check if there is
>> multiple matches and it there are, match basename of known URL against
>> that of media's patch?
>
> Yep, the name is about the only difference between them - all the ISO
> metadata is identical. The other thing I considered is to add an MD5
> checksum of the ISO file, which would trivially distinguish them,
> but I fear that would make it too unique.

That might be a very good idea actually (Christophe was meaning to do
that any ways?) but calculation of md5 takes a very long time:

$ time md5sum ~/ISOs/Fedora-16-x86_64-DVD.iso
bb38ea1fe4b2fc69e7a6e15cf1c69c91  /home/zeenix/ISOs/Fedora-16-x86_64-DVD.iso

real	0m23.348s
user	0m9.276s
sys	0m1.654s

so I'll recommend to restrict it to first N bytes, where we could
think/discuss the value of N but 1MiB should be good enough and takes
very little time:

$ time dd if=~/ISOs/Fedora-16-x86_64-DVD.iso bs=1M count=1 2> /dev/null |md5sum
e25ea147176f24239d38a46f501bd25e  -

real	0m0.011s
user	0m0.004s
sys	0m0.008s

Actually if we start using md5sum for detection, we can pretty much
remove all other detection logic/extraction.

-- 
Regards,

Zeeshan Ali (Khattak)
FSF member#5124