Re: Lintian as a static analysis framework
On Fri, Jul 8, 2011 at 9:25 AM, Stefano Zacchiroli wrote:
> I haven't yet chosen the technology for code indexing, but if anyone has
> experience with *multi-language* code indexers, I'd be happy to hear
> from you. I've looked around a bit, but I've found good technologies
> only for specific languages and I really don't want to have to run
> different indexers for different languages and maintain some proxy code
> that queries all of them in parallel. Hence the fallback I had in mind
> was some non-programming-language thing, like dear old Xapian. But I
> welcome better suggestions.
OpenGrok has support for multiple languages (C, C++, Fortran, Java,
Lisp, Perl, PHP, Python, shell, SQL, Tcl, VB and others), using
"analysers":
http://www.norbye.org/jenkins/job/OpenGrok/javadoc/overview-tree.html
> I guess that would clash with the dear old tension among minimizing
> lintian dependencies and being able to do very specific checks. We
> really don't want to have lintian depend on all static analysis tools
> out there and, last time I've checked, lintian maintainers didn't want
> to have conditional tests (i.e. tests active only when a specific other
> package is installed) to maintain test reproducibility. I know lintian
> maintainers have been struggling with this design decision for a while,
> but I don't know the current state of the discussion.
Right. Thats where the profiles/vendor stuff comes in I think. A
default profile runs only the set of tests that are implemented within
lintian. Any tests that run external commands are in a secondary
profile and lintian warns/fails if the deps for that are missing and
are needed.
--
bye,
pabs
http://wiki.debian.org/PaulWise
Reply to: