Re:Concrete Semantics
by nickpsecurity on Friday January 17, 2020 @03:40PM
(#22079011)
Even that is getting automation as I long ago predicted. I think the SPARK-type proofs will be easier to automate this way, too, since they're more mechanical.
https://blog.acolyer.org/2019/08/23/learning-to-prove-theore...
Some introduction https://blog.adacore.com/using-coq-to-verify-spark-2014-code (from 2014, so probably much improved since).
And some examples : http://docs.adacore.com/spark2014-docs/html/ug/gnatprove_by_...
Not the best way to learn manual proof, but I've seen it used very efficiently.
But then if course you're faced with the difficulty of increasing automation : the more you automate, the more it is difficult to have the 'muscle memory' (and lower cognitive load) that you get when flexing the manual proof often...