[oe] libmatthew-java-0.7.1.tar.gz not fetchable

[Frans Meulenbroeks]

2012/3/20 Steffen Sledz <sledz at dresearch-fe.de>:
> On 19.03.2012 17:21, Tom Rini wrote:
>> On Mon, Mar 19, 2012 at 04:44:10PM +0100, Frans Meulenbroeks wrote:
>>> 2012/3/19 Tom Rini <tom.rini at gmail.com>:
>>>> On Mon, Mar 19, 2012 at 12:49 AM, Steffen Sledz <sledz at dresearch-fe.de> wrote:
>>>>> On 15.03.2012 11:32, Steffen Sledz wrote:
>>>>>> libmatthew-java-0.7.1.tar.gz (used in 2011.03-maintenance branch) is no longer fetchable from it's original location (the maintainer provides the latests version only). And i did not found another reliable source.
>>>>>>
>>>>>> Can someone put the tarball into the OE and/or Angstrom mirrors?
>>>>>
>>>>> Ping!
>>>>
>>>> If someone points Khem at a copy, this can happen...
>>>>
>>>> --
>>> google is your friend
>>>
>>> http://pkgs.fedoraproject.org/repo/pkgs/libmatthew-java/libmatthew-java-0.7.1.tar.gz/6a4db221129f230c64a0f937d00bb703/libmatthew-java-0.7.1.tar.gz
>>>
>>> It might be useful to update the recipe itself as well.
>>> Furthermore we might consider running a fetchall and put all fetched
>>> on the mirror.
>>
>> Does pkgs.fedoraproject.org have a policy about forever, ala
>> snapshot.debian.org?  Otherwise, we're just making a problem for
>> ourselves in the future.  WRT an everything mirror, I think bandwidth
>> usage is a concern there, ala what I think happened with the angstrom
>> everything mirror.

I don't know about the policy fedora has for retaining packages.
There might also be a copy on debian. Google gave a fair number of hits.

Wrt the bandwidth problem: fair
Then again, I'm not sure how much oe classic is build nowadays and how
much load this would really mean.

In any case we could run a fetchall and store the results somewhere.
Then we at least have a copy handy if upstream does not have the
sources any more.
>
> Khem has put the tarball on the mirror. But this does not fix our problem.
>
> Fetching from the original url downloads an XHTML error page and stores it under the name of the archive. This of course results in a checksum mismatch. :(
>
> So in my opinion changing the recipe to the fedoraproject url seems to be the best way.
>
There is yet another option. We could also host the sources somewhere
on oe and adapt the recipe to read from this oe dir.
The difference with a mirror is that this also works if there is no
mirror configured (and ofc "somewhere" could be the mirror dir).

Frans