intel driver will only compile with gcc

[Kean Johnston]

> hw/kdrive/ uses named initialisers pretty heavily.  Does SCO's compiler
> actually lack named initialisers?  Do you have any plans to add it?
The UnixWare compiler has both named initializers and variadic macro
support (although the intel driver currently uses a gcc-ism for them
not the way they are defined to work in C99). The OSR5 C compiler does
not, but gcc is used to compile on OSR5 anyway. Ther problem is the last
version of GCC that was supported on OSR5 was 2.95.3, and since leaving
SCO I have not had the time to maintain gcc in my own time, so thats
very likely where it will stay :(

> Also, we're not talking about C99: we're talking about two very specific
> features, that happen to be in C99.  gcc is not a full C99 compiler, yet
> it (and a few others, including Sun's compiler, and Intel's) still
> manage to support it.
True, I used the term C99 too broadly. But since those two features are
both C99 features I used the term for brevity. The fact that gcc (or any
other compiler) supports the features is (or IMHO should be) moot. We
(sorry, let me be pricise ... *I*) however am faced with the difficult
task of maintaining an old platform. The fact that X.org no longer
compiles out of the box on OSR5 and UW7 is a regression from 6.9. Using
these features, or allowing or encouraging their use, will force that
to forever remain the status quo. There are other more portable
alternatives, albeit far less aesthetically pleasing ones. But I hope
that portability wins over legibility, as the former is of more use to
more people than the latter.

I appologise if previous mails in this thread have lead to heated
discussions or upset anyone. Nothing is further from my intent. I
really hope I can persuade you based on technical reasons why those
two clauses are a bad idea. I would be more than happy to take that
discussion off line, or to a landline, or whatever format you deem
appropriate.

Kean