Re: exception to Paul's Second Rule?

[Noel Yap]

"Paul D. Smith" wrote:
> In one way, in that you need the directory before the target can be
> built, it is a good way.
> 
> But prerequisites are a conflation of _TWO_ concepts, one is order as
> above, but the other is rebuild dependency.  Do you want all the
> contents of a directory to rebuild just because someone touched the
> directory?  Or touched the directory marker file or whatever?  I think
> not...

Hmm, good point.

Practically speaking, the marker file would exist within the build directory 
structure where it's unlikely someone'll be going through.  In the event that 
someone does go in there and do a touch on the file, then technically, by 
definition, they really
would want to rebuild everything.

>   ny> Of course, the hack itself wouldn't be necessary if something like
>   ny> $$(@D) existed for pattern rules.  Would it make sense to have a
>   ny> feature like this?
> 
> Not likely.
> 
> First, it's a totally different thing: the normal $$(@D) is handled as
> the makefile is read in, using some special-case code to set up the
> extra prerequisites.  If it were implemented for pattern rules it would
> have to be done during the walk of the DAG, which would change the
> contents of the DAG depending on the target, etc.  Gross.

I knew there must've been some logical reason why such a feature hasn't been 
implemented; too bad, it would've been nice to have.

Noel