Ross Gardler wrote: > 2009/6/12 Gavin <[email protected]>: >> >>> -----Original Message----- >>> From: [email protected] [mailto:[email protected]] On >>> Behalf Of Ross Gardler >>> Sent: Tuesday, 9 June 2009 5:54 AM >>> To: [email protected] >>> Subject: Re: Report Time >>> >> <snip> >> >>> I would also suggest that the project should focus on encouraging >>> Apache projects to use the tool, perhaps a Gump style build and report >>> mechanism should be a focus for the current team. This will raise >>> awareness of the utility of RAT. > > ... > >> One such ability, which I have implemented over the last few days, is to >> make use of RAT. As RAT is already there being built, I decided to automate >> the use of the jar produced from the build. Now, projects can utilise >> Buildbot to include RAT testing on their source/binaries/etc. I > > Cool!
+1 >> RAT reports can be displayed (currently) as txt files on the ASF Buildbot >> website, they can be emailed (though not currently enabled as a fine balance >> needs to be achieved considering per commit triggers). They can also be >> produced on demand via IRC if they want. > > I'd suggest being able to configure the mail to only be sent if there > is a problem. It's realy useful to have RAT say "hey you forgot to add > a licence header to this new file" +1 the feedback from scan is that a message once a month or so is useful even if everything's cool but otherwise only mail when stuff goes wrong >> This is both useful to RAT as well as the projects using it - as RAT project >> can implement new features or fixes based on the results/feedback from the >> projects using it - one commit later to svn and within a minute all projects >> are immediately using the new version of RAT. > > Is there any kind of summary report screen that would help RAT > developers see unintended side effects that cause multiple errors > across projects? that would be useful :-) >> Buildbot itself has many other features and capabilities not yet >> implemented/realised (Nexus is next on my hit list, as well as a place to >> put release candidates (as most folk usually upload that stuff to >> people.apache.org/~username)). > > I'd suggest putting it into the lab (no releases, but the code is what > matters at this stage). +1 >> My hope here is with integration of RAT is that it can raise the profile of >> both RAT and Buildbot within the Apache community. I'm not sure Ross what >> you had in mind for increased awareness but I hope this is an option we can >> explore. > > This is *exactly* what I had in mind. The RAT web site should have a > section saying, "look RAT is used in the Apache Buildbot" etc. +1 the web site is in need of some TLC >> We can talk about ideas on best/different implementation ideas if folks >> agree it is a good idea. > > +1 +1 >> http://ci.apache.org/buildbot.html > > ... > >> A few of those projects I have configured the builds to upload >> websites/documentation/javadoc/Apis - and also RAT reports for a few. >> >> http://ci.apache.org/projects/ > > This shows that RAT is not all that useful right now unless properly > configured. I was intrigued to see that RAT itself has 85 unapproved > licences - http://ci.apache.org/projects/rat/rat-report.txt > > I'd say the first thing the RAT team need to do is get that down to 0. the problem is that a number of these are false positives (everything in target sub directories are just consequences of the build) or intentional (some of the tests for missing headers need to have no headers) detached license information seems like the right way to solve this. so that sounds like an implementation priority. >> One major thing I'd to work on soon -- and I'd like help with if anyone has >> time (if not it will just take longer I guess) is to make use of the XML >> output and so create an xslt to produce nice readable HTML reports to go on >> the site. Do we have a DTD? Do we need one? Is the current XML output >> suitable or will it need tweaking? Has any of this been done already? This >> stuff I can find out soon. > > I'll have to pass on al those questions I'm afraid. there isn't a DTD or schema ATM since the current format is a bit rubbish i'd prefer to move to RDF or xhtml+rdf but that might be a barrier to entry and may be difficult to transform so, maybe it'd be best to start from the other direction: brainstorm about the design of a really good HTML report and then just work backwards. ideas? >> One last thing I though of, I'm sure we can gather info from these reports >> and send it to wherever Clutch needs it in order to implement that extra >> line on clutch you mentioned Ross. > > That would be really cool. Perhaps the simplest thing to do would be > to have your buildbot dump the files in a location that the existing > Clutch code can access. +1 - robert
