| Summary: | [releng] TM-master-nightly includes some unsigned features and bundles | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Tools] Target Management | Reporter: | Martin Oberhuber <mober.at+eclipse> | ||||
| Component: | RSE | Assignee: | dsdp.tm.rse-inbox <tm.rse-inbox> | ||||
| Status: | RESOLVED WORKSFORME | QA Contact: | David Dykstal <ddykstal.eclipse> | ||||
| Severity: | critical | ||||||
| Priority: | P3 | CC: | dmcknigh | ||||
| Version: | 3.5 | ||||||
| Target Milestone: | 3.5 RC1 | ||||||
| Hardware: | PC | ||||||
| OS: | Windows 7 | ||||||
| Whiteboard: | |||||||
| Attachments: |
|
||||||
*** Bug 404199 has been marked as a duplicate of this bug. *** Correction: The problem is not on our end - the org.eclipse.tm.repo.zip is good and all signed. The problem was, that when I installed the tm.repo.zip into my Eclipse Platform, the Kepler Repository was also enabled. Therefore some features and bundles - those which had the same id / qualifier as the new ones - were pulled from the Kepler simultaneous repo or its mirrors, instead of the tm.repo.zip. And that included some unsigned ones. I think that this behavior of p2 is very unexpected: Why does it download something from the Internet, when my repository is all local ? This makes the install somewhat non-deterministic. I've filed https://bugs.eclipse.org/bugs/show_bug.cgi?id=408208 against Equinox/p2 to investigate the remote download even if things are local. |
Created attachment 231059 [details] Screenshot Build ID: tm-master-nightly #896 from Hudson (15-May-2013) - installed on eclipse-platform-4.3M7-win32-x86_64 - Running on Oracle Java 7u21 x86_64 on Win7 Installing the latest TM drop from hudson, Eclipse warns about some unsigned features and bundles. Looking at the timestamp of these bundles and features, they are very old. I assume that the "features" problem will be fixed automatically when version numbers are updated for Kepler. But the old, unsigned "bundles" may be a problem even if they are really unchanged. Maybe the base repository against which the CBI build compares (and pulls old contents if unchanged) is problematic since unsigned ? Marking as CRITICAL since the Release Train doesn't allow us to deliver unsigned content.