On Mon, 28 Feb 2011, Jim Cromie wrote:

> Julia Lawall <julia <at> diku.dk> writes:
> 
> > 
> > I am organizing a session on development tools for the 2011 Linux Plumbers 
> > conference (http://www.linuxplumbersconf.org/2011/).  If you have an idea 
> > of a tool that might be of interest to the Linux developer, please let me 
> > know.  A suggestion of a person who would be a good speaker about the tool 
> > would be especially appreciated.
> > 
> > thanks,
> > julia
> > 
> > 
> 
> judging from the previous response, this is more specific to coccinelle
> than you intended, so its not really relevant to a plumbers-talk (yet)
> 
> Round-tripping with Coccinelle:
> 
> How about deriving .cocci files from patches accepted into Linux ?
> IIRC, some ~350 patches have been done, many by yourself,
> though that number has apparently gone up...
> 
> ...linux-2.6$ git log --grep=[cC]occinelle --oneline |wc
>     814    5350   44470
> 
> It would be interesting to see whether such an enhancement
> would find all existing patches (those documented in the commit msg)
> and whether it would issue the same .cocci files for them.
> Eventually, the quality could meet or exceed that produced by
> most people (coccinelle authors excluded:)
> 
> It sounds genuinely useful too, by detecting hand-coded api-evolution 
> patches, perhaps with just one user converted (or even zero!), and writing a 
> .cocci file that would find and patch all the other use cases.
> This sounds very powerful; at minimum it would lower the learning curve 
> for SmPL, and dramatically reduce the tedious parts of api-evolution 
> and improve the patchset itself (by ensuring a more complete conversion)

It's a pretty hard problem.  We have done some work on inferring semantic 
patches from a few manually developed patches:

# Generic Patch Inference
Jesper Andersen and Julia Lawall
23rd IEEE/ACM International Conference on Automated Software Engineering, 
337-346, L'Aquila, Italy, September 2008. 
http://coccinelle.lip6.fr/papers/andersen_ase08.pdf

We are now looking at inferring semantic patches from simple bug fixing 
patches, but don't yet have any results to report...

thanks,
julia
_______________________________________________
Cocci mailing list
[email protected]
http://lists.diku.dk/mailman/listinfo/cocci
(Web access from inside DIKUs LAN only)

Reply via email to