Re: implementing emake's #pragma multi in GNU make?

[Eric Melski]

On 01/03/2013 11:11 AM, Paul Smith wrote:
> On Thu, 2013-01-03 at 05:53 -0800, Jamie Cuesta wrote:
> > http://blog.melski.net/2013/01/01/pragma-multi-and-rules-with-multiple-outputs-in-gnu-make/
> >  describes a (IMHO very significant, perhaps even "gaping") limitation of GNU 
> > make (and offers a (IMHO) clean/elegant solution to same).
>
> http://lists.gnu.org/archive/html/make-alpha/2002-12/msg00000.html
>
> There may have been other similar threads since, I can't remember.
>
> I'm not so thrilled about the idea of having a #pragma concept that
> changes the way normal rules are interpreted.

We chose the #pragma syntax for two reasons.  First, it entirely avoids 
the debate about what syntax to use to indicate multiple output targets. 
 From the linked thread that seems to have been a point of some 
contention, and there's disagreement between different make variants -- 
some use "a + b", some use "[a b]", somebody proposed using something 
like "a b |:" or similar.  All of the alternatives have drawbacks, 
albeit perhaps not anything that would actually impact users in practice 
(eg, what if you actually want to have a target named "+"?).

The second advantage of #pragma syntax is that it is not 
backwards-incompatible.  A makefile using "#pragma multi" will still be 
usable with older versions of GNU make, except that you don't get the 
special multi-output behavior.  In the best case, the alternatives will 
cause older versions of GNU make to emit spurious warnings; in the worst 
case they could cause build failures.  "#pragma multi" is completely 
invisible to versions of GNU make that don't support it.

Best regards,

Eric Melski
Chief Architect
Electric Cloud, Inc.