Community
Participate
Working Groups
Created attachment 136652 [details] Error downloading dialog I try and update from http://download.eclipse.org/eclipse/updates/3.5-I-builds I use the "Add" button to add it, and then p2 tries to load it. I get 3 different error/failure dialogs, but after 20 to 30 seconds it does populate with entries. I've attached 3 error pictures (and the populated dialog). PW
Created attachment 136653 [details] Not a repo dialog
Created attachment 136654 [details] Second not a repo dialog
Created attachment 136655 [details] Dialog with contents This will show up eventually if you don't touch anything and simply dismiss the problem dialogs. Try and edit the location or dismiss the install dialog and you'll never know the content is there. PW
What build were you upgrading from? RC1?
(In reply to comment #4) > What build were you upgrading from? RC1? This examples was done on I20090520-2000, but it has been reproducible for me on the builds this week (starting with I20090518-2000) PW
I am also still seeing this on I20090520-2000
Ok, it looks very much like bug 276562, but that was fixed in I20090519-2000.
additional comment: There are no errors in my error log. PW
see also bug 276884. I'm not sure if these are related or not. That case involves a silent/unlogged failure but it does not manifest with the error dialog, the UI just stops in its tracks and does not show the site.
From some initial exploration, I wonder if it is just a real transport exception that we're just not handling very well. The site has about 18 builds in it, for a total of about 60MB of metadata, so this is a huge site to load (Kim is going to clean it up today). It could be that a legitimate timeout or transfer failure during this process is happening. When I tested this, I got the error in one out of three attempts, so it isn't consistently reproducible. When I try with a local mirror, I never get the problem.
The fact that the load completes in the end makes me think the underlying exception is coming from a child of the composite repo, and instead of aggregating it somehow the parent repository is letting it get through?
>The fact that the load completes in the end makes me think the underlying >exception is coming from a child of the composite repo, and instead of >aggregating it somehow the parent repository is letting it get through? Yes, that is possible. CompositeMetadataRepository does catch and log ProvisionException thrown when loading children, but it's possible there is a runtime exception coming through. The I-build site has now been cleaned up and I can't reproduce at all any more.
Closing as worksforme.