aries-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Holly Cummins <holly.k.cumm...@googlemail.com>
Subject JIRA versions for released bundles
Date Fri, 21 Sep 2012 13:42:03 GMT
Hi all,

One last piece of release work I haven't done yet is the JIRA cleanup.
One reason for this is that I don't think I have the JIRA authority to
create new versions. A more fundamental blocker is that I'm not sure
our current JIRA strategy for releases is going to scale. At the
moment we create a JIRA version for each release - that is, for each
released version. So we already have, for example,

	blueprint.annotation.itests-0.3.2		
	blueprint.annotation.impl-0.3.2		
	blueprint.annotation.api-0.3.2

The 1.0.0 batch of releases contained around 86 bundles, which means
86 new versions cluttering up the JIRA list. Once we start releasing
1.0.1 and 1.1.0 versions, the number of versions will grow even more.
And, of course, I hope that doing a release is now much easier, which
means we expect the number of versions would grow pretty fast. :)

One solution would be to continue using normal numbers like 1.0.0 or
1.0.1 to identify the version. This would keep the proliferation of
versions under control. In combination with the component, this would
nearly be enough to uniquely identify the bundle where a problem
occurred.

Is the lack of precision about exactly which bundle is being
referenced a problem? I'm not sure it is, since users reporting bugs
may not be able to pinpoint whether a problem was caused by
org.apache.aries.application.runtime-1.0.0 or
org.apache.aries.application.utils-1.0.0, for example. Some problems
will be related to several bundles, so in these case choosing a
version which references just one bundle will actually be plain wrong.

What do people think?
Holly

Mime
View raw message