Re: forcing a rule

[Noel Yap]

Boris Kolpackov wrote:

> I didn't notice significant downside but it is not free either. Patch
> makes make variable-expand implicit rule's prerequisites for every
> time there is a match (but not every time implicit rule is tried).
> Since most rules don't use double expansion this second expansion is 
> just an additional search in a (usually) not very long string for '$'.  
>
> I use it extensively in my build system to build directories and it 
> works really well.

I don't suppose it should break any existing makefiles, either.  Can you 
foresee any situations which will break after installing the patch?

> > Have you submitted the patch to the GNU make project (via savannah.org I 
> > believe)?  
>
> Well, I found this to be pretty useless. I submitted a patch for 
> $(lastword ) which is a very simple patch with a very straightforward
> (IMO) argumentation and there is a precedent ( $(firstword) ). Since
> I submitted it a few months ago I never heard a word back. Also, I saw 
> a far more bizarre features added to make meantime (like SysV $$@ & 
> friends) so it makes me wonder...

I understand that Paul has put a hold on any new functions since he's trying to 
get Guile as _the_ scripting language within makefiles so I can understand why 
lastword, useful as it is, didn't make the cut.

> So I started my own patch-set with hope that if more people find it
> useful we will be able to get it into official make.  

IME with patches for CVS, this has more of a chance (although certainly no 
guarantee) of working if there's a discussion about the feature.

> > Has this feature been discussed on this list?
>
> I haven't seen anybody express interest in such a feature (until now)
> so there was no point in discussion.

There sure is :-)

Paul, in case you've missed this thread, essentially the patch being discussed 
will allow double evaluation of rules so that the following can be done:

  %/.: | $$(dir\ %).

What do you think about the feature?

Thanks,
Noel