Install problem(s) with recent CVS

[Chris B . Vetter]

Hi,

I just (Mon, 17:20 PST) made a CVS update of GNUstep, moved my
former installation to a different location, to re-install
everything, and stumbled upon the following:

1) Problem with installation path

:0> pwd
/usr/local/src/GNUstep/CVS/gnustep/dev-apps/libobjc
:0> gmake install
[...]
if [ -f shared_obj/ix86/freebsd/gnu-gnu-gnu/libobjc.so.1.0.1 ];
then \
  /usr/bin/install -c
shared_obj/ix86/freebsd/gnu-gnu-gnu/libobjc.so.1.0.1 \

/usr/local/GNUstep/System/Libraries/ix86/freebsd/ix86/freebsd/gnu-gnu-gnu
; \
 (cd
/usr/local/GNUstep/System/Libraries/ix86/freebsd/ix86/freebsd/gnu-gnu-gnu;
  rm -f libobjc.so libobjc.so.1; ln -s libobjc.so.1.0.1
libobjc.so.1;
  ln -s libobjc.so.1 libobjc.so; ) \
fi
[...]
:0> cd /usr/local/GNUstep/System/Libraries/ix86/freebsd/
:0> ls -al
total 4
drwxr-xr-x  4 root  wheel  - 512 Oct 22 17:18 ./
drwxr-xr-x  3 root  wheel  - 512 Oct 22 17:17 ../
drwxr-xr-x  2 root  wheel  - 512 Oct 22 17:17 gnu-gnu-gnu/
drwxr-xr-x  3 root  wheel  - 512 Oct 22 17:18 ix86/
:0> ls -al ix86/freebsd/gnu-gnu-gnu
total 96
drwxr-xr-x  2 root  wheel  -   512 Oct 22 17:18 ./
drwxr-xr-x  3 root  wheel  -   512 Oct 22 17:18 ../
lrwxr-xr-x  1 root  wheel  -    12 Oct 22 17:18 libobjc.so@ ->
libobjc.so.1
lrwxr-xr-x  1 root  wheel  -    16 Oct 22 17:18 libobjc.so.1@ ->
libobjc.so.1.0.1
-rwxr-xr-x  1 root  wheel  - 96024 Oct 22 17:18 libobjc.so.1.0.1*

$GNUSTEP_LIBRARIES already seems to include the CPU and OS, so
$GNUSTEP_TARGET_LDIR will add them and the backend to the
installation path.

Of course, a re-make of core/gnustep-make will not find libobjc.

2) gnustep-base

configure ignores --disable-procfs. If --dis/enable-procfs is
omitted, configure will still use /procfs - which makes this
option kind of unneccessary and confusing.

LONG_LONG_MAX should actually be LLONG_MAX, as defined in
/usr/include/limits.h (Linux, Sun) and
/usr/include/machine/limits.h (*BSD).

HP/UX (10.20) defines LONG_LONG_MAX directly, in
/usr/include/limits.h

--enable-openssl should also check for libcipher, which has
setkey(3), encrypt(3), des_setkey(3) and des_cipher(3) - else
(at least on BSD) ld will puke about
  "WARNING!  setkey(3) not present in the system!"
  "WARNING!  des_setkey(3) not present in the system!"
  "WARNING!  encrypt(3) not present in the system!"
  "WARNING!  des_cipher(3) not present in the system!"

I've also noticed a couple of "variable xyz possibly clobbered
by longjump or vfork", but haven't checked those out yet - and
about tempnam(3) and mktemp(3) possibly used unsafely.

Maybe make should work -s(ilent) so we'll have time to read all
those warnings and notes.

-- 
Chris