I know this has been discussed previously, and both Debian and Gentoo
(heh, me), have had issues with GNUstep.sh, it's use of make_services,
and assumptions about the user environment.
Problems became apparent to me, using Gentoo's portage sandox, that
the installation of the GNUstep libraries, which except for
gnustep-make, is dependant on GNUstep.sh, assumes that the use of a
user's "proper" $HOME is available, and ignore's the environment's
"configured" $HOME.
Being able to configure the $HOME environment variable has made the
package management nightmares I was having go away. I've put together
a patch that effects NSUser.m in -base and user_home.c in -make.
I've thought a lot about possible consequences, and the only one I can
come up with is "Is the getenv() function secure enough on all
platforms we support?" For this reason, I ask that anyone please go
over this patch for GNUstep-ness, go over for coding correctness and
security, and don't flame me too bad 'cause I know this was discussed
and supposedly solved already.
Thanks,
__Armando Di Cianno
<base-nsuser-home-fix.patch>
Only in base/Source: .NSUserDefaults.m.swp
diff -ur base-orig/Source/NSUser.m base/Source/NSUser.m
--- base-orig/Source/NSUser.m 2004-06-25 15:30:43.000000000 -0400
+++ base/Source/NSUser.m 2004-07-23 21:47:19.147599472 -0400
@@ -230,13 +230,26 @@
{
NSString *s = nil;
#if !defined(__MINGW__)
- struct passwd *pw;
+ const char *env_list;
[gnustep_global_lock lock];
- pw = getpwnam ([loginName cString]);
- if (pw != 0)
+ env_list = getenv("HOME");
+ if (env_list != 0)
+ {
+ s = [NSString stringWithCString: env_list];
+ }
+ else
{
- s = [NSString stringWithCString: pw->pw_dir];
+ struct passwd *pw;
+
+ pw = getpwnam ([loginName cString]);
+ if (pw != 0)
+ {
+ s = [NSString stringWithCString: pw->pw_dir];
+ }
}
[gnustep_global_lock unlock];
#else
