2 Sep
2011
2 Sep
'11
5:29 a.m.
On 08/31/2011 10:46 AM, Eduard Moraru wrote:
Hi devs,
I was wondering why we don`t generate source artefacts for snapshots, just
like we do for final versions?
While I do agree that final versions are the important ones, it's often very
annoying to have to manually add sources from the filesystem (assuming you
have cloned locally the git repo) for all the maven modules that you want to
see the code or even the javadoc or method parameter names. For final
releases we don`t have that problem because Eclipse and the Maven plugin
(for example) can automatically download the sources in the background
providing an effortless process. However, while working with snapshots, you
are nagged continuously that the sources can not be found automatically and
you experience a bit of unneeded frustration :). Not only that
Would it be that much of a storage penalty for the extra sources jars to
accompany the snapshot builds?
Would it be that much of a build time penalty for actually generating source
builds for each artefact?
Generating the sources zip shouldn't take that much time, since it's
mostly building the zip, the sources are already available. Generating
Javadocs on the other hand would take some time, since it requires
parsing all the java files and generating a lot of HTML. Given that the
sources jar is enough for developers, I'd say that only the sources jar
should be built.
Still, I think that it's not that relevant that sources for snapshot
jars are not available, since it addresses a very small niche. Most
downstream developers should work based on a released stable version,
for which sources are available. Some of those that work on the trunk
platform should depend not on pre-built jars, but on the real project
based on the trunk sources, which makes it easier to modify the platform
code as well. Sure, it's more time-consuming to set up the proper links
between the projects (and more resource-hungry since loading the tens of
modules that the XWiki platform contains takes a lot of heap space), but
it ends up in a more powerful setup.
I'm +0 for enabling this step, but only after the space problem with the
maven machine is solved for good.
--
Sergiu Dumitriu
http://purl.org/net/sergiu/