Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 269111

Summary: [launcher] start levels don't get added when launching without a configurator
Product: [Eclipse Project] PDE Reporter: Chris Aniszczyk <caniszczyk>
Component: UIAssignee: Chris Aniszczyk <caniszczyk>
Status: RESOLVED FIXED QA Contact:
Severity: critical    
Priority: P3 CC: bh, jeffmcaffer, phil.kursawe, ulf
Version: 3.5   
Target Milestone: 3.5 M7   
Hardware: PC   
OS: Mac OS X - Carbon (unsup.)   
Whiteboard:
Attachments:
Description Flags
mylyn/context/zip
none
org.eclipse.pde.patch
none
org.eclipse.pde.patch
none
mylyn/context/zip none

Description Chris Aniszczyk CLA 2009-03-17 20:02:10 EDT
We've seen issues where start levels aren't ending up in the config.ini when launching without any configurator.
Comment 1 Chris Aniszczyk CLA 2009-03-18 17:00:34 EDT
adding context... this problem sucks
Comment 2 Chris Aniszczyk CLA 2009-03-18 17:01:15 EDT
Created attachment 129291 [details]
mylyn/context/zip
Comment 3 Ulf Holm Nielsen CLA 2009-03-25 09:03:58 EDT
I can attest that this is rather frustrating. Both wether to start and the level are not exported to the config.ini file. Do you need a testcase or is it obvious where the problem is?
Comment 4 Chris Aniszczyk CLA 2009-03-25 11:44:59 EDT
So the workaround for this problem is to simply use simple configurator.

I would also grab this week's i-build when it comes out as it has a lot of fixes around start levels and product definitions.
Comment 5 Philipp Kursawe CLA 2009-03-28 21:42:32 EDT
Chris see also #270366. I think this needs to be solved. Will test with newest I-Build tomorrow
Comment 6 Chris Aniszczyk CLA 2009-03-29 21:34:41 EDT
*** Bug 270366 has been marked as a duplicate of this bug. ***
Comment 7 Chris Aniszczyk CLA 2009-03-29 21:35:35 EDT
(In reply to comment #5)
> Chris see also #270366. I think this needs to be solved. Will test with newest
> I-Build tomorrow

I am working on this problem but there's a lot of code changes in critical areas so it's going to take awhile. For now, the workaround is to include simpleconfigurator and use those facilities. 
Comment 8 Chris Aniszczyk CLA 2009-03-30 17:49:54 EDT
looking to target next week's i-build with a fix for this.

PDE was a much happier place when there were no bundles with multiple versions :/
Comment 9 Chris Aniszczyk CLA 2009-04-10 12:48:27 EDT
Created attachment 131513 [details]
org.eclipse.pde.patch

Consolidated some of the launching code.

This patch needs some further work and testing before I release it.

I'm targetting next week's i-build.
Comment 10 Chris Aniszczyk CLA 2009-04-13 10:08:28 EDT
Created attachment 131648 [details]
org.eclipse.pde.patch

Here is the final and tested patch.

I dream of a future where we have even more simplified launching code along with unit tests ;)
Comment 11 Chris Aniszczyk CLA 2009-04-13 10:08:32 EDT
Created attachment 131649 [details]
mylyn/context/zip
Comment 12 Chris Aniszczyk CLA 2009-04-13 10:09:18 EDT
done.

> 20090413

I'd appreciate if people would pickup this week's i-build and test this stuff out in their various usage scenarios.
Comment 13 Brett Hackleman CLA 2009-04-29 14:17:20 EDT
Using: I20090428-0100

Sorry it's taken me so long to pick up the I-builds and test out the new fixes.  I did find one problem with exported products that include bundles configured for the default start level.

When the product is exported, the config.ini has:

osgi.bundles=com.bandxi.application@0:start, <snip>

Launching the product fails with:

java.lang.IllegalArgumentException: The requested start level of 0 is invalid.  The value must be a positive integer.

I'm assuming the fix is to not include the ":<start_level>" for bundles using the default start level when generating the config.ini file.  When running the product from within the workbench, the generated osgi.bundles property does not include the start level for these bundles.

Thanks!
Comment 14 Chris Aniszczyk CLA 2009-04-29 14:41:37 EDT
Brett, this is a publisher issue and is being tracked in bug 273961. It should make M7