Re: [p2-dev] creating artifact repo from arbitrary files

[Krzysztof Daniel <kdaniel@xxxxxxxxxx>]

On Thu, 2013-11-28 at 15:41 +0000, Pascal Rapicault wrote:
> Depending on the content of partial profiles, it may get complicated to get them integrated in the real profile
> I would rely on shipping artifacts.jar that can be used by the profile generating mechanism and I would do the same for metadata repo with a content.jar.
Yeah, that was inevitable, but doesn't cause much trouble for me.

> If you want we can try to have a voice chat about the approach you are taking,

That would be good. We can arrange details on #eclipse-dev.

The approach sounded simple:
* load plugins listed in some files as if they were written in the
master, read-only bundles.info file
(https://git.eclipse.org/r/#/c/18429/7/bundles/org.eclipse.equinox.simpleconfigurator/src/org/eclipse/equinox/internal/simpleconfigurator/utils/SimpleConfiguratorUtils.java, 30-119). Format of those files is the same as of bundles.info: (https://git.eclipse.org/r/#/c/18429/7/bundles/org.eclipse.equinox.p2.tests/testData/simpleConfiguratorExtendedTest/extension2/extension2.info)
* make getProfile method return a profile that has bundles from the
previous point installed. That turns out to be impossible (due to the
every attempt of profile manipulation resulting in a new bundles.info
written).

I really wonder whether partial profile (a file containing just units
merged at profile registry read, like http://fpaste.org/58359/59887541/)
wouldn't be the best option.

Chris. 
> 
> Pascal
> 
> -----Original Message-----
> From: p2-dev-bounces@xxxxxxxxxxx [mailto:p2-dev-bounces@xxxxxxxxxxx] On Behalf Of Krzysztof Daniel
> Sent: November-28-13 9:58 AM
> To: p2-dev
> Subject: [p2-dev] creating artifact repo from arbitrary files
> 
> Hi,
> 
> I've found a problem when trying to install things loaded from a custom *.info file into a base profile - P2 requires metadata (which I generate from a running platform) and artifact repository, connected with an artifact key. The point is that I don't have any artifact repository connected with the *.info files, hence my question: what is more appropriate to do - create repositories on the fly (is that even
> possible) or ship entire repos?
> I like the speed of shipping partial profiles, as there is no startup overhead unless dropins are engaged on startup, but shipping repos has its pros and cons - writing second dropins-like mechanism (tracking repo changes etc etc) seems to be a bit overkill.
> 
> 
> --
> Krzysztof Daniel <kdaniel@xxxxxxxxxx>
> Red Hat
> 
> _______________________________________________
> p2-dev mailing list
> p2-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/p2-dev
> _______________________________________________
> p2-dev mailing list
> p2-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/p2-dev


-- 
Krzysztof Daniel <kdaniel@xxxxxxxxxx>
Red Hat