2010/2/10 Gav... <ga...@16degrees.com.au>: > > >> -----Original Message----- >> From: Mark Thomas [mailto:ma...@apache.org] >> Sent: Tuesday, 9 February 2010 8:27 PM >> To: Tomcat Developers List >> Cc: builds@apache.org >> Subject: Re: buildbot failure in ASF Buildbot on tomcat-trunk >> >> On 09/02/2010 10:25, build...@apache.org wrote: >> > The Buildbot has detected a new failure of tomcat-trunk on ASF >> Buildbot. >> > Full details are available at: >> > http://ci.apache.org/builders/tomcat-trunk/builds/83 >> > >> > Buildbot URL: http://ci.apache.org/ >> > >> > Buildslave for this Build: bb-vm_ubuntu >> > >> > Build Reason: >> > Build Source Stamp: [branch tomcat/trunk] 907983 >> > Blamelist: markt >> > >> > BUILD FAILED: failed compile >> >> The Eclipse download takes forever which seems to be the root cause of >> a >> number of these failures. Could the build property "jdt.home" be set to >> somewhere that isn't deleted on every ci build? That should speed up >> the >> build process considerably - as well as not downloading a 25MB file >> from >> archive.eclipse.org every time we build. > > Actually, I haven’t set that property at all that I know of. > In Buildbot it can be added as a property at build time so it's no problem. > > Shall I go ahead and do that? > > What value should it be, empty? > > Gav... >
Not jdt.home, but base.path I think. The "base.path" property points to a location, where all dependent libraries are downloaded from the net. The default value of it is base.path=/usr/share/java and as I see from stdio output from buildbot, there it is set to /home/buildslave3/slave3/tomcat-trunk/build/basepath/ The problem is that all those dependencies are deleted before each run and are redownloaded, which wastes time and, with a 40 MB download from eclipse.org, resulted in intermittent failures. Can we avoid deleting those files between runs? In the meanwhile, yesterday I changed dependencies in tomcat-trunk, and now we are downloading a 1.6 MB file instead of 40 MB one. So, the problem will be less noticeable for trunk now, but will still present for "tomcat-6-trunk", where this change is not implemented yet. The "tomcat-6-trunk" buildbot apparently has the same value of "base.path" as the "tomcat-trunk" one. That is, it reuses libraries downloaded by "tomcat-trunk" on its previous run. Thus it was more rare to see errors coming from it. Best regards, Konstantin Kolinko
