My recent research has mostly focused on querying and transforming Cobol and C software. So I was wondering what the chances are for supporting these languages in Semmle ?

Getting at a good Cobol frontend will either take a nice amount of money, or a bigger amount of time and effort I think. You could look at open source projects such as OpenCobol, but these are nowhere near processing realistic (i.e. industrial) Cobol code. You might also want to look at the Grammar Recovery Kit (http://www.cs.vu.nl/grammars/grk/) which has a VS Cobol II grammar. While this still needs something more powerful than a run-of-the-mill yacc/bison, I have been able to put it to good work.

If none of this has you deterred then there is still the question of which version of Cobol to target. One choice would be to go for the Cobol 85 standard or the newer 2002 one (though 85 code is more common, in my experience). But you will also have to consider vendor-specific extensions, in which case you should look at what IBM and Fujitsu-Siemens are offering.

If you do get this going, though, it could trigger a lot of experiments against some real-life "ancient" (and yet thriving) code! (I'll be running a few of them myself, of course. )