Re: automauve

[Brian Jones]

Dalibor Topic <address@hidden> writes:

> Hi Brian,
> 
> --- Brian Jones <address@hidden> wrote:
> > I've written a simple means of presenting Mauve test results for
> > multiple projects over a period of time.  I still have to write a
> > little Perl script to create the necessary gnu plot file that will
> > make a nice graph; the graph is the only piece missing.  All of
> > the web pages are generated statically.  I started out with PHP
> > but found it was too slow for calculating the results, etc.
> > 
> > This can be used to track gcj-<platform> results against each
> > platform if you want or any other vm+platform.  The presentation
> > scheme is based loosely on something similar I saw done for
> > building Jakarta with Ant.  Let me know if you find this useful.
> 
> Thanks, that was another missing bit for people evaluating free java
> VM progress.
> 
> Do you have the resources to include kaffe, sablevm & wonka in your
> test grid?

I'm not sure what this means.  Basically I only have access to
whatever Linux I happen to be running on x86.  I wanted to have some
way of tracking measurable progress at finding and fixing Classpath
bugs.  My plans are to make this available under the GNU GPL so other
parties can use if they so desire to track <vm>-<platform> relative
to each other or to other JVMs.

I wouldn't mind finding a way of easily gathering results from other
VMs/platforms and adding them, but my goal remains to find and fix
bugs in Classpath.

> There is a Perl script that converts the output from mauve into HTML
> in the japhar CVS, but it takes a ton of time & memory. You may want
> to take a look at it if you want to create mauve result pages like
> this :
> http://www.google.de/search?q=cache:Ha1wwPfP7XgC:www.japhar.org/mauve-results-11.html++site:www.japhar.org+japhar&hl=de&ie=UTF-8
> (taken from Google's cache because I can't get to the site at the
> moment).

It reads the Mauve results into a hash of hashes; shouldn't take more
memory than the size of the results file.  I didn't change Japhar's
script and it ran in a few seconds on the result file I have and the
output looks reasonable.

Brian
-- 
Brian Jones <address@hidden>