Marcus Sundberg wrote:
> Andreas Beck <[EMAIL PROTECTED]> writes:
> > > How about if we #ifdef around that code. Windows will be the only OS
> > > which can use an environment variable. The others would still be hard
> > > coded.
> > Could be done, but I don't really like special solutions ...
> There has been a giiGetConfDir() function in LibGII for a long time
> for exactly this purpose, and I've just added a ggiGetConfDir()
> function to LibGGI. They can be implemented as needed for Win32.
Actually, that is where I made the changes for using an environment
variable, and I did notice that it didn't exist in ggi. Thanks Marcus!
I assume that it will be in a new snapshot.
> The idea I had was that the location of the config dir should be
> stored in the Registry and retrieved from there. The environment
> variable is not good, as in NT you actually have some security
> wrt file permissions. I don't know if NT has the notion of suid
> programs, but we should really do it right from the start as
> that might be added later.
Yeah, you are right. I was hoping to avoid the registry, but it is the
proper place for the config directory to be. Unfortunately, this really
means there has to be an install process besides just unzip/untar.
I suppose an uninstall process is needed also.
> Another alternative is to always put the config file in the Windows
> directory (which I'd think you can query from the OS, right?)
Ick ( Technical Term ! :) ) Enough crap already goes there.
> Speaking of this - John, it would be nice if you also had a look
> at the ggGetUserDir() function in LibGG and did something nice on
> Windows. I know at least NT has the notion of a "home" directory.
I'll take a look.