[LLVMdev] Advice on CFG pre-analysis
hendrik at topoi.pooq.com
Mon Jun 2 10:54:21 PDT 2008
On Tue, 27 May 2008 21:19:24 -0700, Talin wrote:
> The overall motivation for this, BTW, is my attempt to prove a personal
> theory, which is that most static type systems are guarding against the
> wrong things. Type mismatch errors don't keep programmers up at night;
> null pointers, deadlocks and race conditions do. So my language tries to
> check for null pointer errors at compile time using static types.
The reason type-mismatch errors don't keep programmers up at night is
that the compiler catches them statically. In the old days, when
compilers didn't catch all these errors (I think of using pointer
variables in PL/1) programmers did stay up all night over type mismatches.
So it isn't that the static type systems are guarding against the wrong
errors -- it's that they aren't guarding against enough of them.
More information about the llvm-dev