Re: Release schedule

[Richard Frith-Macdonald]

On Friday, April 4, 2003, at 11:38  pm, Chris B. Vetter wrote:

> In that particular case, I wasn't even using anything in NSApplication,
> it just got included via <AppKit/AppKit.h>
>
> Apart from that, you will get a hole slew of missing interfaces, syntax
> errors, eg. in NSWindow.h.
>
> So how is it supposed to make sure it's compatible, if some parts
> "inside" depend on parts, that are defined "outside" the scope?
>
> What am I missing?

I'm not sure that you are.

The STRICT_OPENSTEP definition is there for two purposes -
1. To make it easy for developers to produce code which only makes use 
of the
OpenStep  API (without any extensions) by having the preprocessor remove
anything non-standard which the developer might  use by accident.  Thus 
allowing
development of code which should run on OPENSTEP.
2. To mark the headers so that autogsdoc knows which classes/methods 
belong
to the original OpenStep spec, so it can identify them in the generated 
documentation.

Both those aims require careful placement of #ifdefs ... the first 
being more
tricky because (as you noticed) it has to deal with non-standard data 
types being
used for ivars of standard classes.

I've just been through and altered the gui headers enough to get 
applications
to compile with STRICT_OPENSTEP defined.  I was disappointed by the
amount of alterations I had to do that, and Im certain that the 
STRICT_OPENSTEP
stuff is not in use throughout the headers as it should be :-(
Hopefully this is not the case in the base library, where I think it's 
all correct.

For portability issues (use of STRICT_OPENSTEP, STRICT_MACOS_X, port to
windows, runtime compatibility options etc) the GNUstep developers 
*MUST*
depend upon the people using the system to fix problems, as there 
aren't enough
core developers to test all this sort of thing.  With the best will in 
the world, errors
will continually creep in and the people who need a non-standard system 
have to
keep a check on it to stop that sort of creep.

Given  the state I found the headers in ... my impression is that 
nobody is actually
using STRICT_OPENSTEP for the first purpose ... If nobody is trying to 
use it for
producing OPENSTEP compatible applications, should we be bothering with 
the
first usage at all?