[prev in list] [next in list] [prev in thread] [next in thread] 

List:       wine-devel
Subject:    RE: Yet some more config.h
From:       Patrik Stridvall <ps () leissner ! se>
Date:       2000-11-26 11:14:36
[Download RAW message or body]

> On Sat, 25 Nov 2000, Patrik Stridvall wrote:
> 
> > Well Solaris C doesn't have neither const not stdcall
> > and I have after some hacking got a few WineLib applications
> > to run so it is not impossible.
> 
> Yes, but why bother. Solaris tools suck big rocks in general, 
> and can go
> away ASAP AFAIC. 

FYI the version of Solaris C that I use at work is from _1996_
and generates better code and compiles faster than the _latest_
offical version of GNU C.

The reason that we use Solaris C is because mangagement didn't
trust GNU C in 1995 and since it works fine we haven't had
any need to upgrade. For exactly the same reason we still run
Solaris 2.6.

I don't know why you think that Solaris tools sucks.

It is possible one day that GNU C will be the king of compilers
in the same sense that Linux today is the king of OS:s, but
it certainly isn't there yet and will not, I predict, be for
many years yet.

> I mean, what do we have to gain from 
> supported antiquated
> tools when there are better, free alternatives?

But the free alternatives are _not_ better. Cheaper, yes,
Solaris C is quite expensive, but not better.

> But this is a 
> different
> topic... 
> We have to take a decission on whether or not we are going to
> support such tools. If yes, than we _need_ the config file in all
> files.

True.

> If not, we may remove it from a lot of them.

True.
 
> > However winapi_check uses some kind of heuristics to determine
> > whether config.h is needed or not so it is not 100% certain
> > it is always correct.
> 
> There is no need for heuristics. If we adopt a simple rule 
> (if file tests
> any constants, it has to include config.h directly), than we 
> can trivially
> say with certainty if config.h is needed or not.

The heuristics is in the parsing (and semantic analysis) 
not in the actual application of the rules.

For example

#define FOO HAVE_SYS_TIME_H
#if FOO
#include <sys/time.h>
#endif

will fool the current heuristics.

[prev in list] [next in list] [prev in thread] [next in thread]