Re: Complex compiler/linker tests

[Ralf Wildenhues]

Hi Liviy,

* Liviu Nicoara wrote on Fri, Jul 01, 2005 at 05:28:27PM CEST:
> Ralf Wildenhues wrote:
> >
> >Surely.  But your example has a trivial different solution: don't write
> >code which depends on static inlines to be collapsed.  
> 
> Writing the least common denominator code would eliminate the expressing 
> of interesting and useful features of the language in our programs.

Sure.

> The example I provided is legal C++ code. Testing whether or not the 
> compiler/linker emit/link the correct code should be a function of the 
> infrastructure.

Acknowledged.  Without trying to argue your specific case again: in some
cases, it might make more sense to not test a non-universal feature, but
forego it instead, in order to avoid blowing up the infrastructure (or
maybe for other reasons).  But this is just a dumb-down general comment.

> I am reading the autoconf, etc., m4 files trying to get a feeling on how 
> the internal macros are written and what functionality is provided in 
> them. So far I have not seen anything to offer support for linking a 
> library. Do you know off the top of your head if such functionality is 
> in there?

Library creation (and linking against) is supported with GNU Libtool.
Its macros are huge, though, and you most likely won't need all of that.
Bruno Haible has written some macros to link portably against libraries.
They can be found in gnulib, for example, as part of the `iconv' module.

Regards,
Ralf