Re: PR19578: noreturn vs. function pointer propagation part 2

[Diego Novillo <dnovillo at redhat dot com>]

Richard Sandiford wrote:

> Are you saying you actually preferred the idea of marking the blocks?
> If so, I'll try to do it like that instead.
Yes.  My apologies for not making that clear.

The easy approach was to just end blocks at any indirect call, but this 
will cause programs with lots of indirect calls to have more basic 
blocks than necessary.  This may increase compile time unnecessarily.

If you don't find significant compile time differences, then let's leave 
your patch in.  It's safe enough.


Thanks.  Diego.