Re: [equinox-dev] [p2] How to use director/mirror app against an RCP as p2 repo

2017-06-26 Thread Roland Grunberg
On Mon, 2017-06-26 at 12:33 +, Carsten Reckord wrote:
> In the IDE, this works directly: In "Install new software", you can use the 
> P2 profile directory inside of another Eclipse installation as your 
> repository location and install from there. E.g. 
> /p2/org.eclipse.equinox.p2.engine/profileRegistry/SDKProfile.profile
> I'm not sure how the director/mirror app handle this - they seem to be more 
> picky and have fewer heuristics. So I could imagine that in order to get it 
> to work, you might need to specify artifact and metadata repositories 
> separately. In that case, I would assume that the profile directory should 
> work as the metadata repo, while the Eclipse installation directory (or your 
> bundle pool location) would be your artifact repo. This is all guess work 
> though, and I didn't try it myself.

If this is the case, then I would guess there should be an
IArtifactRepository implementation for an installation directory, or
simply some artifact repository that is "close enough" to handle the
format since it's not that different. I know that a profile .tar.gz
file can be parsed by some metadata repo implementation so there should
also be enough info in there to reconstruct where the files are
located.

For Fedora (https://github.com/rgrunber/fedoraproject-p2), we needed to
supply plain directories as part of the target platform that may or may
not have osgi bundles, but aren't p2 repositories, so we ended up
creating our own I{Metadata,Artifact}Repository implementations but
that's a little more generic than what you need.

Cheers,
-- 
Roland Grunberg
___
equinox-dev mailing list
equinox-dev@eclipse.org
To change your delivery options, retrieve your password, or unsubscribe from 
this list, visit
https://dev.eclipse.org/mailman/listinfo/equinox-dev


Re: [equinox-dev] P2 Tests

2017-03-01 Thread Roland Grunberg
On Wed, 2017-03-01 at 09:52 +0100, David Weiser wrote:
> What I did so far:
> I added the Eclipse testing framework. I imported
> org.eclipse.core.tests.harness from the eclipse.runtime bundle. To
> solve the error "No available bundle exports package
> 'org.eclipse.equinox.http' I imported the rt.equinox.bundle without
> success (erros still there). 

I remember when I ran the p2 tests, I also couldn't resolve
'org.eclipse.equinox.http' but I wasn't really dealing with that part
of the code. The last repository I could find it in was :

http://download.eclipse.org/releases/indigo/

IU: org.eclipse.equinox.http 1.0.500.v20110413
* java.package/org.eclipse.equinox.http/0.0.0

If you're not dealing with p2.testserver though, you should be able to
just close the project and only run the tests you care about.

> When I run "All p2 tests" I get one error (testCancellation) and 56
> failures (assertInitialized). When I try to run "All p2 UI Tests" it
> is not started at all. It seems that org.eclipse.osgi can't find
> bundles?!
> 
> What am I missing? Any idea what is wrong? 

I always ran the tests with my default target platform, but more
recently, it hasn't worked out so I took p2-mars.target and made some
changes to get it working with neon and oxygen targets. Not sure why
those weren't updated (everything run through maven anyways?). I filed 
https://bugs.eclipse.org/bugs/show_bug.cgi?id=512323 .


Hope this helps,
-- 
Roland Grunberg
___
equinox-dev mailing list
equinox-dev@eclipse.org
To change your delivery options, retrieve your password, or unsubscribe from 
this list, visit
https://dev.eclipse.org/mailman/listinfo/equinox-dev

Re: [equinox-dev] Disabling computeUses in ResolverImpl

2016-10-21 Thread Roland Grunberg
> Disabling uses is usually not a good idea. I would be interested to know why
> you want to do this.

I was observing rather large initial startup times for an Eclipse installation.
There were ~1200 bundles as part of the installation but past installations of
comparable size never took that long.

The main thing that improved startup time was to ensure the installation didn't
have so many overlapping Export-Packages. After a certain point though (seeing
rather large time spent in mergeCandidatePackages() )I got to thinking about
whether the uses directives might be contributing to this. Disabling
computeUses() seemed to do the trick but I'm also aware of how bad this can be
for debugging actual uses constraint issues.

Cheers,
-- 
Roland Grunberg
___
equinox-dev mailing list
equinox-dev@eclipse.org
To change your delivery options, retrieve your password, or unsubscribe from 
this list, visit
https://dev.eclipse.org/mailman/listinfo/equinox-dev


[equinox-dev] Disabling computeUses in ResolverImpl

2016-10-21 Thread Roland Grunberg
Hi,

It seems like in older versions of the Equinox framework it was possible to 
disable uses constraint checking on startup (osgi.resolver.usesMode).

This doesn't seem to be possible any longer but I'm wondering if the same
behaviour could be mimicked by disabling the computeUses() line in
org.apache.felix.resolver.ResolverImpl [1]. Are there additional side
effects to doing such a thing that weren't present before ?

Cheers,
-- 
Roland Grunberg

[1] 
http://git.eclipse.org/c/equinox/rt.equinox.framework.git/tree/bundles/org.eclipse.osgi/felix/src/org/apache/felix/resolver/ResolverImpl.java?h=R4_6_1#n1234
___
equinox-dev mailing list
equinox-dev@eclipse.org
To change your delivery options, retrieve your password, or unsubscribe from 
this list, visit
https://dev.eclipse.org/mailman/listinfo/equinox-dev