Bugzilla – Full Text Bug Listing
| Summary: | problems building p2 platform zips | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Eclipse Project] Platform | Reporter: | Kim Moir <kim.moir> | ||||
| Component: | Releng | Assignee: | Platform-Releng-Inbox <platform-releng-inbox> | ||||
| Status: | RESOLVED FIXED | QA Contact: | |||||
| Severity: | normal | ||||||
| Priority: | P3 | CC: | avijayr, dj.houghton, john.arthorne, konstantin, mknauer, ngc2997, pascal, remy.suen, tjwatson, tomcatuser.13 | ||||
| Version: | 3.4 | ||||||
| Target Milestone: | 3.4 M6 | ||||||
| Hardware: | PC | ||||||
| OS: | Windows XP | ||||||
| Whiteboard: | |||||||
| Attachments: |
|
||||||
|
Description
Kim Moir
Can you give a link to the metadata repository that this install is running against? Is it the testUpdates repository? testUpdates-N on dizzy Moving to releng as this is being handled by Kim Here is a summary of the current state: - We currently build the Eclipse SDK, then platform binary, then platform SDK. - The Eclipse SDK is fine, but the platform binary and platform SDK are broken - If we switch the order of execution (platform binary, then Eclipse SDK, then platform SDK), then the first one to be built is good, and the remaining builds are broken. - So, there is some interference from the previous invocation of the generator that is contaminating the subsequent invocations. Comparing the Eclipse SDK in good state (run first), to Eclipse SDK in bad state (run second), we can see: - The root IU is missing all the bundle CUs in the broken state. For example, it is missing CU for org.eclipse.equinox.simpleconfigurator, org.eclipse.equinox.p2.reconciler.dropins, etc - These CUs are connected to the product IU on line 158 of Generator: productContents.rootIUs.add(iterator.next()); - We haven't been able to track down any further why these CUs are missing. Another suspicious line is 926 in Generator.java result[i] = factory.getBundleDescription(bundleLocations[i]); The call factory.getBundleDescription can return null in various cases - error reading the manifest or other error creating bundle description. Many of these exceptions are completely swallowed so we wouldn't see any trace of them, especially in BundleDescriptionFactory#loadManifest. In this case no bundle IU would be created in Generator#generateBundleIUs, and if the bundle IU is not created then the bundle CU would not be created, and therefore it would be missing from the root IU. However, the bundle IU is still present in the repository from a previous invocation of the generator so it otherwise looks fine. Anyway, that's another possible avenue for investigation, but unless we can actually step through the generator it is very hard to narrow down further. Created attachment 94142 [details]
patch
Here is a patch which does a bit better logging.
I didn't release it yet.
*** Bug 224915 has been marked as a duplicate of this bug. *** I just logged https://bugs.eclipse.org/bugs/show_bug.cgi?id=225029 which might be similar. If this is the case, PLEASE fix and release M6.1 - I am unable to work with 3.4 M6 due to the issue Thanks! *** Bug 225029 has been marked as a duplicate of this bug. *** *** Bug 225257 has been marked as a duplicate of this bug. *** Is there a distro of 3.4 M6 which I can use? I am unable to work with M6 on Windows due to this issue. A platform build should be available tomorrow. *** Bug 225410 has been marked as a duplicate of this bug. *** Can this be closed now that the M6 platform zips are available? The updated binary zips were released earlier today and should be replicated on mirrors by tomorrow. The milestone update site has been updated with the corrected metadata. Has the fixed version already landed on http://download.eclipse.org/eclipse/downloads/drops/S-3.4M6-200803301350/index.php? I downloaded that version (which now reports all test results ok, as opposed to a few days ago) earlier this morning (around 1000 CEST), yet it still crashes immediately when it is started. (In reply to comment #16) I tested the eclipse-platform-3.4M6-linux-gtk-x86_64.tar.gz download and was able to start it without problems. Regarding comment #16, what did you platform and zip did you download from this build page http://download.eclipse.org/eclipse/downloads/drops/S-3.4M6-200803301350/index.php (In reply to comment #18) Well, I've downloaded the eclipse-SDK-3.4M6-linux-gtk.tar.gz package for my Ubuntu 7.10 machine.. > Well, I've downloaded the eclipse-SDK-3.4M6-linux-gtk.tar.gz package for my
> Ubuntu 7.10 machine..
This bug never existed for the SDK zips - only the platform zips. I suggest entering a new bug report.
(In reply to comment #20) Hm, isn't https://bugs.eclipse.org/bugs/show_bug.cgi?id=225029 something similar? That one has been marked as a duplicate of this.. (In reply to comment #21) > (In reply to comment #20) May be your crash relate to read-only installation? (bug 216066, bug 225257) (In reply to comment #22) > [..] > May be your crash relate to read-only installation? (bug 216066, bug 225257) Yes, that sounds reasonable. My eclipse installation resides in /opt/eclipse, where a 'normal' user doesn't have write access.. |