[Mono-devel-list] Analyzing Subversion logs

Paolo Molaro lupus at ximian.com
Wed Jan 12 11:28:34 EST 2005


On 01/12/05 Paolo Marini wrote:
> Sorry for bothering you with this questions but I am working on my degree
> thesis which deals with Subversion repositories analysis. Now I am tyring to
> analyse the mono project to extract some information but I have found that
> an important number of commits appear really close each other.
> For example, it happened more than 150 times that "miguel" performed a
> commit less than 10 seconds after his previous commit. I am trying to

It depends on the specific files. I think this may happen for two
reasons:
*) importing source code from somewhere else: miguel did most of it
and it may happen that cvs recorded different times as the import proceded
*) some time ago we has some file system corruption on the cvs server
and of course something was wrong with the backup:-) We had the
(mostly) last version of the files, but we lost the history for
quite a few changes in mcs/. To try and get back the history information,
I wrote a perl script that grabbed the changes from the commit list
or something like that and recreated the missing history: this may have
resulted in a large number of changes being committed in fast succession.

I guess that, if you are interested in analyzing subversion repositories,
you should use the info from the repo only starting since the subversion 
switch.
The info extracted from cvs has also a number of issues of skewed data, 
resulting from imports and the like. For example, take the mini/ directory
(had to use cvs annotate on the cvs repo, since svn annotate is unusable):
miguel appears to have written 24K lines (of a total of 80K), but the
lines he wrote in the jit are at best a few hundred. This happens because
he imported the mini sources in cvs: there are many other cases like this.

lupus

-- 
-----------------------------------------------------------------
lupus at debian.org                                     debian/rules
lupus at ximian.com                             Monkeys do it better



More information about the Mono-devel-list mailing list