+1 for CHANGES.txt. Yes, it is largely redundant with information in
Jira, but its advantages are [1] a more consise, user-friendly
description than usually exists in Jira (as illustrated by Yonik); and
[2] it provides a great sanity check when merging changes between
branches, since each commit is accompanied by a description. You can
run 'svn merge ...' then 'svn diff CHANGES.txt' and easily check whether
the merge included the desired changes.
I think the best practice is to include a description of the change both
in CHANGES.txt and in the commit message. The description in the commit
message might be shorter, e.g., only the first sentence from
CHANGES.txt. Commit messages should of course also always include all
relevant Jira issue ids, whether it is the initial commit or a merge
between branches. This is more redundancy, but descriptive commit
messages make it possible to use things 'svn log' without constantly
having to follow references to Jira, and issue ids in commit messages
make it possible to see in Jira all branches and revisions that a patch
is in. These cost a bit at commit time, but make the repository and
Jira much nicer to use.
Doug
Chris Mattmann wrote:
Hi Folks,
My +1 for having a CHANGES.txt file, for all the reasons that Yonik
mentioned below. Also, it, to me, is much easier to maintain (simply doing
it incrementally, with each major update), then figuring out the mysteries
of JIRA ;)
So, I'd like to call for a vote from the committer community. Should we use
a CHANGES.txt file? (please mark an X by your choice)
[ ] Yes
[ ] No
Let's leave the vote open until we have a majority of binding votes.
Thanks!
Cheers,
Chris
On 6/12/07 12:38 PM, "Yonik Seeley" <[EMAIL PROTECTED]> wrote:
On 6/12/07, Sami Siren <[EMAIL PROTECTED]> wrote:
2007/6/12, Jukka Zitting <[EMAIL PROTECTED]>:
On 6/7/07, Chris Mattmann <[EMAIL PROTECTED]> wrote:
I've created a Tika changelog, CHANGES.txt, within the TIKA trunk that we
can use to track changes, akin to Nutch. Typically, when a committer
commits
a patch (in Nutch) that closes out a tracked issue (i.e., with JIRA), then
we add a comment into CHANGES.txt, akin to the following (where "n" is a
number incremented from 1, for each release):
I think it's just duplicating information from the svn log, but if
people think it's a good idea, then I have no problem following suit.
Jira can also provide a list of changes (if all changes are logged
there) with the changelog - or release notes [1] functionality.
[1]
http://issues.apache.org/jira/secure/ReleaseNote.jspa?version=12312013&styleN
ame=Text&projectId=10680&Create=Create
I personally use CHANGES.txt in Lucene and Solr as user-level
documentation / release notes. If it's too minor, an internal
cleanup, a documentation fix, etc, I don't put it in. It also
provides a place to concisely explain the impact of a change (in user
language), rather than requiring the user to try and glean this from a
long JIRA discussion mingled with dev-list discissions.
Examples:
3. SOLR-91: allow configuration of a limit of the number of searchers
that can be warming in the background. This can be used to avoid
out-of-memory errors, or contention caused by more and more searchers
warming in the background. An error is thrown if the limit specified
by maxWarmingSearchers in solrconfig.xml is exceeded. (yonik)
and
3. LUCENE-710: added optional autoCommit boolean to IndexWriter
constructors. When this is false, index changes are not committed
until the writer is closed. This gives explicit control over when
a reader will see the changes. Also added optional custom
deletion policy to explicitly control when prior commits are
removed from the index. This is intended to allow applications to
share an index over NFS by customizing when prior commits are
deleted. (Mike McCandless)
It's much more understandable then a JIRA bug list, but it does take more
work.
If you wish to go this route, you could perhaps defer it until the
first Tika release.
-Yonik
