| Summary: | [releng] TM Athena tests fail due to WinCE | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Tools] Target Management | Reporter: | Martin Oberhuber <mober.at+eclipse> | ||||
| Component: | RSE | Assignee: | Anna Dushistova <anna.dushistova> | ||||
| Status: | CLOSED WONTFIX | QA Contact: | Martin Oberhuber <mober.at+eclipse> | ||||
| Severity: | normal | ||||||
| Priority: | P3 | ||||||
| Version: | 3.2 | ||||||
| Target Milestone: | 3.3.1 | ||||||
| Hardware: | All | ||||||
| OS: | All | ||||||
| Whiteboard: | |||||||
| Bug Depends on: | 315239 | ||||||
| Bug Blocks: | 302196 | ||||||
| Attachments: |
|
||||||
|
Description
Martin Oberhuber
Created attachment 170591 [details] patch to build wince without source. Attached patch builds wince without source, and specifies the RAPI bundle's Eclipse-PlatformFilter in the feature.xml. This is prompted by the statement on bug 312011 comment 12 that it can cause issues when the bundle-filter is more stringent than the feature-filter. I investigated a little, and I think the problem is this: 1. For build purposes, a single "master feature" is created to hold EVERYTHING a product is building. In case of TM, this includes the WinCE feature which is (only works on Windows), and the Local.Terminal feature (works everywhere except on Windows). So in some sense, it is impossible to satisfy all requirements of the TM master feature. 2. The master feature is translated into a master repository. 3. For installing the repo for testing, Athena uses the p2.director application. But p2.director must always install into ONE given os/ws/arch combination. This fails on the TM-master feature since its os/ws/arch combination is unsatisfiable in its completeness. So right now, it is inconsistent that Athena can (and has to) *build* a feature with unsatisfiable dependencies, but it cannot *install* that feature afterwards. The single master feature is required today for signing etc. Proposed solution 1: -------------------- Instead of using p2.director to install for testing, Athena should optionally be capable of using the p2 "repo2runnable" task and have the runnable dropped into dropins. That way, the p2 reconciler works on dropins to pick out those items from the master feature that are actually runnable on the test platform - and there is no need any more for using/installing the single master zip. Proposed solution 2: -------------------- Rather than automgically installing the master feature into the target for testing, Athena should optionally be capable to accept a list of IU's to install from the repo. That way, rather than installing the inconsistent master feature, we could specify a set of consistent child features to test. This is not relevant for TM Helios any more since we cannot adopt Athena for Helios. See bug 302196 comment 5 for details. Bulk moving 3.3 deferred items to 3.3.1 Athena project has been terminated. |