Re: Sparse and the Google Summer of Code 2008

[Josh Triplett <josh@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 1990 bytes --]

Derek M Jones wrote:
>>> I know, last year was not all that successful, but maybe
>>> there will be more applications this year. I've got some
>>> ideas about abstract interpretation for one. I can't promise
>>> that I'll choose sparse, there should be lots of interesting
>>> projects this year, but there are other people as well.
>> At least one other person has asked about Sparse and SoC, which
>> suggests that some interest exists.  I'll work on getting the
>> org forms filled out today.
>>
>> If anyone else on the Sparse list has an interest in mentoring, please
>> let me know as soon as possible.
> 
> The ideas so far proposed are major undertakings.  The sort of
> thing a summer student will only scratch the surface of.

To clarify, I wouldn't expect a student project to do everything
necessary to complete any of the areas I mentioned.  I wouldn't expect
complete handling of contexts, or a complete port of an extensive test
suite.  (And the comment about code generation falls under "sarcastic
offhand remark", not "project suggestion"; anyone wanting to do that
should scale back their expectations significantly unless they have
lots of compiler experience.)  On the other hand, a student project
could reasonably add one or two additional things to context checking,
or do a proof-of-concept port of a handful of tests with some
underlying infrastructure.

> I think it would be useful to analyze the Linux bug history,
> looking for known faults and their fixes, that are currently not
> detected by Sparse, but which look like they are amenable to static
> detection.
> If more faults are required there is always the BSD bug list.
> 
> If the existing faults were categorized it would give some idea
> of the kinds of constructs that ought to be searched for.

Agreed entirely, and that sounds like an excellent project possibility.
Any additional useful warning in Sparse represents a good contribution.

- Josh Triplett


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]