Re: [tree-ssa] More cleanups, this time for tree-cfg

[Andrew MacLeod <amacleod at redhat dot com>]

On Wed, 2003-07-23 at 14:19, Diego Novillo wrote:
> On Wed, 2003-07-23 at 14:16, Andrew MacLeod wrote:
> 
> > It does make things a little easier to read sometimes, and more
> > importantly removes you from underlying data structure changes... It
> > also helps prevent user errors by missing a bit somewhere. 
> > If things are inlined, you ought not see gratuitous function calls.
> > 
> Yeah, that's what I like about it.  What I waffle about is the fact that
> we can't enforce this automatically.  Oh, well.
> 
> 
Unfortunately we can't, but we can be net.cops on our code when the
facility is there. In particular on things that I expect to be impacted
down the road. Someday we are going to want to have multiple CFGs. we're
going to want to replace the underlying linking structure.

Someday when Im bored Im going to go patrolling for non-safe practices
around iterators. Such as looking for CE nodes. There should be no
references to COMPUND_EXPR nodes outside of the iterator/manipulation
code. If there is, we either need extra functionality, or it can be done
differently. Otherwise we'll never be able to replace the
implementation.

Andrew