Re: [ANNOUNCE] GitStats development finished (WRT GSoC)

On Wed, Jul 30, 2008 at 16:22, Johannes Schindelin
[off-list ref] wrote:

On Wed, 30 Jul 2008, Sverre Rabbelier wrote:

quoted

Hehe, don't expect too much of it yet though (the 'bug' module that is),
there won't be any "99% of the bugs come from Dscho" yet :P.

So what should I expect?  What does it show?

Atm I have 4 'metrics' (4 out of the 5 suggested ones), namely:

* Does a commit belong to a specific branch. This can be used, for
example, to mark commits on 'maint' as bugfixes with 99% reliability
(you need to make sure your regexp catches all branches you are
interested in, when Junio splits of '1.5.6.4' of maint, it might list
'1.5.6.4' instead of 'maint', as they both point at the same ref).
* Does a commit fully revert another one (e.g., make commit A, do 'git
revert A', it will detect that commit as reverting A).
* Does the commit message match a certain regexp (e.g., if it contains
the word "fixes", it is likely a fix).
* Does the commit diff match a regexp (e.g., a change from
"test_expect_failure" -> "test_expect_success")
- (not tweaked yet) Does a commit partially revert another one, this
needs some tweaking, otherwise small changes are quickly seen as
similar. (If you have two unrelated one-lines, and you set it to
ignore one difference...)

You have to configure a few things to get it running (for example, you
need to tell it what branch is your maintenance branch, if any). You
can also configure how valued each metric is (you can say "if it's in
'maint', add 5 points, but if it's a revert, add 10, and add only 1
point if it matches this regexp). Configuration is trivially done with
'git config GitStats.key value".

(Will add the above to the bug module's documentation)

If anyone comes up with any more metrics to determine whether
something is a bug it can be easily hooked in.

-- 
Cheers,

Sverre Rabbelier