Re: [Groff] additions to autoconf files

[Bernd Warken]

On Thu, Jun 06, 2002 at 10:15:03PM +0100, Ralph Corderoy wrote:
> 
> > > The current policy is to add a test only if there is a failure on a
> > > particular system.  Today, there are many standards like POSIX which
> > > define a lot of header files, and most operating systems follow it.
> > > Older platforms are dying so I really don't see a reason to blow up
> > > the configure script with those tests.
> > 
> > That is not a policy, but trial and error, and error, and error,...
> 
> History shows that it isn't though.  It's a policy of only fixing
> problems that exist.  Any errors that do crop up are considered, tested
> for, and worked around.  Who has complained that groff-getopt.h was
> missing?  

groff-getopt.h is a local include, so it does not belong to the test
list, alright; but the other ones seem to be all system includes - so
why not test them?

> Or that they had X11 but not X11/Xaw/Dialog.h?

No one complained so far because gxditview is neither compiled nor
installed so far.  In order to get this running, the tests are
mandatory.
> 
> There's no point in it doing the test unless something takes note of the
> result.  And what is the code to do if some of those header files are
> missing?

What do you mean?  In many places in the groff sources, there are `#ifdef's
on limit.h.  If something isn't there you must provide it yourself - the
variables from the autoconf test ensure that the right branch is selected
automatically.

I really missed a test for type `bool' when I rewrote the intermediate
output parser.  But as God did not want to add a clean test I had to
plunder around with devilish work-arounds ]:-[ 
The more with unit32.  It's so illogical, stupid, arbitrary misuse of
power, guys.


Moreover, who the faq urged me to write groffer with those stupid ancient
plunder shells instead of using nice arrays from bash?  Why aren't we
allowed to use clean ANSI C++?  The group asked for compatibility - now
do it!

Bernd Warken