Re: conditionally compiling C++ code

[Bob Proulx]

Thomas Dickey wrote:
> Peter Eisentraut wrote:
> > I would advise against writing configure scripts that check whether some
> > library or program exists, and if not, automatically disable some feature
> > of the package altogether.  This leads to tons of problems downstream.

I agree.  Make it easy for downstream packagers to get consistent
results.  Most users will see your package through the eyes of your
downstream packagers who mostly volunteer their time to make it easy
for further downstream users to use your package painlessly.

> > Example: For years, we shipped a configure script with the logic that if
> 
> bad example: if the configure script cannot find readline, the program
> can't use it.
> 
> try to think of a plausible example (if you're able).

As I read Peter's post I believe the point was that the tool was
generally considered to require readline functionality.  Not having it
would lead to bug reports from users.  *Automatically* dropping
readline support at configure time was a bad thing.  It was better to
error if readline was not found so as to alert the user that they were
not getting the full feature set they were expecting.  If they *chose*
to drop readline support with a configure option as a conscious
decision then great because then there are no surprises.

Bob