On 21 January 2013 03:30, Lewis John Mcgibbney <[email protected]> wrote: > Hi, > Two questions > - does anyone know why we seem to be running unstable builds most of the > time? This may just be Jenkins...
See my other email. Most of our unstable builds that I have investigated turned out to be transient HTTP errors from the test sites that we use to verify compatibility with some real world applications. It shouldn't be too difficult to setup a Jetty web server for integration tests. > - does anyone know why the builds attempt iteratively to build various > modules? The integration tests are performed in a slightly strange manner, that may have been necessary in the past, but I don't think is necessary with Maven-2.2+ or Maven-3.0+. They are performed using the maven-invoker-plugin that boots up a separate copy of maven using a different pom.xml file as the base file for the reactor on the separate copy. This may cause maven to attempt to build things multiple times. Jenkins on the other hand, may be setup to run the build and ignore failures until the end of the build, which will attempt to compile apache-any23-core, even though apache-any23-mime failed to build. That failure to build apache-any23-mime successfully may mean that it was not deployed, and the apache-any23-core build may be using an out of date copy. I managed to get the apache-any23-mime tests to fail on my machine using your combined patches, where it doesn't fail on my tika-1.2 or spi-extractors branches separately. I will investigate what parts of the patch are missing that should still be there and hopefully get a patch for it today. Cheers, Peter
