| |
![Karim Sharif's Gravatar profile Gravatar of Karim Sharif](//www.gravatar.com/avatar/2d147a8bf76ebbc84a49ec32bab8620b?d=identicon&s=32) | Posted by Karim Sharif in reply to Barry Pederson | Permalink Reply |
|
Karim Sharif ![Karim Sharif's Gravatar profile Gravatar of Karim Sharif](//www.gravatar.com/avatar/2d147a8bf76ebbc84a49ec32bab8620b?d=identicon&s=80)
Posted in reply to Barry Pederson
| I couldn't agree more (even though I prefer Linux :-).
In general it's the POSIX standard that these OS's are
associated with that make the system software and
languages on them so portable in the first place.
So, in my opinion, kernel specific stuff belongs as
sort of a 'sub' version of the standard, POSIX.Linux
vs. Linux.POSIX...
Karim
In article <aq74cf$g81$1@digitaldaemon.com>, Barry Pederson says...
>
>I've been fooling around with getting DLI to compile and run under FreeBSD, and have had some minor success in getting a simple "hello world" type program to compile and run.
>
>One thing I noticed was the use of the version identifiers "linux" and "Linux" in the Phobos library to identify non-Win32 code. That seems a bit shortsighted to assume Linux is the only other platform D will run on.
>
>I'd suggest perhaps defining (in the D specs) "POSIX" to be the standard version identifier for code that applies to that family of platforms, which would include Linux, Free/Net/OpenBSD, Cygwin, MacOSX and so on.
>
>A Linux compiler could still enable the "Linux" version identifier in addition to "POSIX", in case you needed to get at Linux-specific bits of code. Presumably FreeBSD would enable ["FreeBSD", "POSIX"] for versions, Cygwin would do ["Cygwin", "POSIX"] .....
>
>I think this would be worth deciding on sooner rather than later - before the habit of using "Linux" for non-Linux specific things gets too ingrained.
>
> Barry
>
|