Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cross-project-issues-dev] feature dependency version incompatibility

Hi Ed-


Thanks for (yet another) patient explanation, I really appreciate you taking the time to write it up.

I took another look at my troubleshooting and it seems that I may have misread the log files.

I thought I was seeing a “only one of the following can be installed at once” error.

I guess this type of error occurs when “singleton:=true” is present in MANIFEST.MF.

However, now that I went back to the log files I see that it was a different error which seems to have been caused by my local configuration.


Christoph, now I understand why you were questioning the error and you were right, sorry about that!


Alexander, thanks for logging and I’d like to do it for 2020-09, but I guess it is less urgent than I thought.


Thanks all and apologies for raising a mostly false alarm.

Tony Homer


From: <cross-project-issues-dev-bounces@xxxxxxxxxxx> on behalf of Ed Merks <ed.merks@xxxxxxxxx>
Reply-To: Cross project issues <cross-project-issues-dev@xxxxxxxxxxx>
Date: Sunday, June 14, 2020 at 12:33 AM
To: "cross-project-issues-dev@xxxxxxxxxxx" <cross-project-issues-dev@xxxxxxxxxxx>
Subject: Re: [cross-project-issues-dev] feature dependency version incompatibility



Whether or not two plugins can be installed simultaneously depends on whether they are declared to be a singleton in the MANIFEST.MF:

  Bundle-SymbolicName: org.eclipse.emf.common;singleton:=true

I believe most (all?) of the Orbit bundles are not singletons.

The repository analyzer reports that I generate for SimRel show information about "duplicates".  E.g., in this Installable Units section of the following link, there is a "Duplicates" radio button to show only the duplicates:

There are 44 duplicates for the coming release.  Apparently batik is super popular with 5 versions; the platform appears to be using the latest...

One of the goals of SimRel was to reduce this to no duplicates.  But given there is really no resource to effectively manage the overall SimRel process, that issue is ignored. 

It's challenging just to get the teams to fix more serious problems though those are stamped out now with only the following two minor (trivially fixable) issues as hold-outs:

Eliminating duplicates would certainly be very good, but I expect some ocean boiling would be involved to actually make that happen.  There's lots of old content being contributed...

It is possible to install everything from SimRel into a single beast:

That does actually launch and it looks to be in better shape for 2020-06 than in the past.  In that beast I can see the two versions:


On 13.06.2020 18:45, Homer, Tony wrote:

Thanks for responding, Christoph.
I think I got your point about sensible version ranges, but as you mentioned the features define strict versions.  Conflicting might be the wrong word to use, but the point is that some features specify javax.annotation 1.3.5.qualifier while others specify javax.annotation 1.2.0.qualifier, so AFAIK these features have conflicting dependency requirements and cannot be installed at the same time.  I was hoping someone would tell me that I am mistaken about this (
On 6/12/20 , 11:06 PM, "cross-project-issues-dev-bounces@xxxxxxxxxxx on behalf of Christoph Läubrich" <cross-project-issues-dev-bounces@xxxxxxxxxxx on behalf of laeubi@xxxxxxxxxxxxxx> wrote:
    Was is selected in the future depends on the version available at build 
    If a never version at runtime is used/allowed depends on the imports in 
    the Bundle itself.
    So if proper imports and use clauses are defined it could be that all 
    works fine even with different (why they are conflicting?) versions.
    In a perfect world, all bundles would use import package with the lower 
    bound of the lowest working version (e.g.  1.2.0) and an upper bound of 
    2.0.0 then it would be possible to upgrade up until the next major release.
    The problem is, that features does not support version ranges and thus 
    will still reference the version from the build (or the one forced into 
    the feature.xml).
    Am 13.06.20 um 07:56 schrieb Homer, Tony:
    > Some 2020-06 features now require javax.annotation 1.3.5 (Jersey Common 
    > from Orbit and bundles that depend on it such as Linux Tools Docker) 
    > while others (such as org.eclipse.e4.rcp) still require javax.annotation 
    > 1.2.0.
    > Is it possible to reconcile these dependency chains without rebuilding 
    > the features, so that an eclipse application can use several of these 
    > conflicting features from SimRel 2020-06?  As far as I know it is not, 
    > but I’d love to learn that I am wrong.
    > e4.rcp does not seem to have any version restriction defined (0.0.0).
    > e4.rcp is resolving javax.annotation from the updates repo at build 
    > time, which provides 1.2.0.
    > What contributes these third-party dependencies to the updates repo?
    > Shouldn’t third-party dependencies be resolved from Orbit instead of 
    > from the updates repo?
    > How do I update one of these dependencies that is being provided by the 
    > updates repo, such as javax.annotation?
    > In any case, it seems that it must be too late to fix this for 2020-06, 
    > but these should get reconciled before 2020-09.
    > I’m interested in helping with this work, but I’ll need the information 
    > I asked for above and could use some direction about which project to 
    > log the bug in for tracking.
    > Thanks!
    > Tony Homer
    > P.S. Is this an appropriate discussion for cross-project-issues dev?  I 
    > don’t want to spam the list so please let me know if this is off-topic 
    > and how I should communicate instead.  Thanks!
    > _______________________________________________
    > cross-project-issues-dev mailing list
    > cross-project-issues-dev@xxxxxxxxxxx
    > To unsubscribe from this list, visit
    cross-project-issues-dev mailing list
    To unsubscribe from this list, visit
cross-project-issues-dev mailing list
To unsubscribe from this list, visit

Back to the top