Until all the plugins are loaded, we cannot start the discovery process, so if the list view is requested before that, just return the list of legacy instances as an approximation.
The current plugin code shouldn't expect the DescriptorList to be fully populated until all the plugins are loaded anyway, so this partial visibility isn't any different from what they currently see.
OTOH, this prevents NPE in ExtensionFinder.Sezpoz when it tries to get the uber classloader.
git-svn-id: https://hudson.dev.java.net/svn/hudson/trunk/hudson/main@15606 71c3de6d-444a-0410-be80-ed276b4c234a
- extensions are now stored in ExtensionList.
- discovery is extensible, but by default we use Sezpoz for annotation-based injection. This requires JDK6 for build.
- a care has been taken to preserve backward compatibility. DescriptorList and manually adding to it continues to work.
- AdministrativeMonitor is converted to use the new approach entirely, as a guinea pig of this feature.
- Descriptors for View is partially switched to this new approach, where some classes register automatically while others do manually. This tests the compatibility aspect of this feature.
A pluggable discovery is an attempt to embrace Plexus integration work, although for that to really work we need more coding.
Next stop, more tests.
git-svn-id: https://hudson.dev.java.net/svn/hudson/trunk/hudson/main@15601 71c3de6d-444a-0410-be80-ed276b4c234a
Previously, Hudson only tried to keep successful builds, meaning it would discard the last stable build
if there was a long run of unstable builds. This made .../lastStableBuild/... permalinks useless.
Artifact archiver logic with "latest only" also improved more generally:
1. Discards old artifacts at the start of the build. It might as well save the disk space sooner rather than later.
2. Rather than deleting all old artifacts prior to some point, now keeps the latest representatives
of different stability classes of build: always keeps last stable if available, and also last successful if newer,
and (just in case) also the last failed if that is newer. Formerly it behaved oddly: during a run of failed builds,
all of their artifacts would be kept, though these are unlikely to be useful.
Also introducing tests for all this functionality, which does not seem to have been tested before.
git-svn-id: https://hudson.dev.java.net/svn/hudson/trunk/hudson/main@15587 71c3de6d-444a-0410-be80-ed276b4c234a
https://www.dev.java.net/svn/hudson/branches/managed-windows-slave
........
r15420 | kohsuke | 2009-02-17 22:54:43 -0800 (Tue, 17 Feb 2009) | 1 line
added a work in progress
........
r15421 | kohsuke | 2009-02-17 23:01:47 -0800 (Tue, 17 Feb 2009) | 1 line
added copyright
........
r15422 | kohsuke | 2009-02-17 23:07:50 -0800 (Tue, 17 Feb 2009) | 1 line
forgot to copy this
........
r15540 | kohsuke | 2009-02-21 21:00:29 -0800 (Sat, 21 Feb 2009) | 1 line
moving WMI support into another library, and making more progress
........
r15541 | kohsuke | 2009-02-21 22:19:43 -0800 (Sat, 21 Feb 2009) | 3 lines
making more progress.
We need an UI to reinstall a service
........
r15556 | kohsuke | 2009-02-22 11:15:12 -0800 (Sun, 22 Feb 2009) | 1 line
allows null
........
r15559 | kohsuke | 2009-02-22 11:29:08 -0800 (Sun, 22 Feb 2009) | 1 line
added a new mode of launcher that accepts one TCP/IP connection and use that for communication
........
r15560 | kohsuke | 2009-02-22 11:39:52 -0800 (Sun, 22 Feb 2009) | 1 line
more bug fixes and stabilization
........
r15577 | kohsuke | 2009-02-22 18:15:51 -0800 (Sun, 22 Feb 2009) | 2 lines
improved the form field databinding by allowing @field to be on <f:entry>.
This enables the convention-over-configuration for pointing to the help file.
........
git-svn-id: https://hudson.dev.java.net/svn/hudson/trunk/hudson/main@15579 71c3de6d-444a-0410-be80-ed276b4c234a