httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brian Behlendorf <>
Subject Re: votes on 1.0.2
Date Sun, 11 Feb 1996 21:48:34 GMT
On Sun, 11 Feb 1996, Garey Smiley wrote:
> On Sat, 10 Feb 1996 23:35:11 -0800 (PST) you wrote:
> >100b.os2_port.patch +1
> >  looks fine to me - wish we could have a little more abstraction, but oh
> >  well.
> Any ideas on how this could be done?

Instead of saying "If we have EMX defined, then don't do this setgid command,
because EMX doesn't support Unix groups", in conf.h you say "EMX is defined
as not supporting groups" and in the code you say "if the ability to manage
groups is defined, then do this piece of code, else do this other piece of
code".  In other words, do the conditional switch not on the OS, but on a
certain ability which may be an attribute of mroe than one OS, and then
define in a top-level place what a particular OS can do.  For example: does
NT have the concept of "groups"?  If not, then we want the same condition 
to apply.

Maybe this is unnecessary, but it's something that popped up when I 
looked at the patch.  Similarly for the Next setsockopt patches - I tried 
what the patch suggested for NeXT on BSDI (using a (char *) instead of a 
(const char *)) and it compiled and seemed to run fine.  If we can avoid 
mentioning platforms or introducing conditionals based on platforms in 
the code, the better off we probably are.  It might make the #define 
namespace larger, but it will probably also make the code easier to 
manage, and port.

Isn't there a GNU package that does things like this?  Helps manage 
projects where cross-platform development is a priority?  Hmm.


--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--  http://www.[hyperreal,organic].com/

View raw message