1 # Hints for the PowerUX operating system running on Concurrent (formerly
2 # Harris) NightHawk machines. Written by Tom.Horsley@mail.ccur.com
4 # Note: The OS is fated to change names again to PowerMAX OS, but this
5 # PowerUX file should still work (I wish marketing would make up their mind
8 # This config uses dynamic linking and the Concurrent C compiler. It has
9 # been tested on Power PC based 6000 series machines running PowerUX.
11 # Internally at Concurrent, we use a source management tool which winds up
12 # giving us read-only copies of source trees that are mostly symbolic links.
13 # That upsets the perl build process when it tries to edit opcode.h and
14 # embed.h or touch perly.c or perly.h, so turn those files into "real" files
15 # when Configure runs. (If you already have "real" source files, this won't
18 if [ -x /usr/local/mkreal ]
22 for j in embed.h opcode.h perly.h perly.c
26 ( cd $i ; /usr/local/mkreal $j ; chmod 666 $j )
32 # We DO NOT want -lmalloc or -lPW, we DO need -lgen to follow -lnsl, so
33 # fixup libswanted to reflect that desire (also need -lresolv if you want
34 # DNS name lookup to work, which seems desirable :-).
36 libswanted=`echo ' '$libswanted' ' | sed -e 's/ malloc / /' -e 's/ PW / /' -e 's/ nsl / nsl gen resolv /'`
38 # We DO NOT want /usr/ucblib in glibpth
40 glibpth=`echo ' '$glibpth' ' | sed -e 's@ /usr/ucblib @ @'`
42 # Yes, csh exists, but doesn't work worth beans, if perl tries to use it,
43 # the glob test fails, so just pretend it isn't there...
47 # Need to use Concurrent cc for most of these options to be meaningful (if you
48 # want to get this to work with gcc, you're on your own :-). Passing
49 # -Bexport to the linker when linking perl is important because it leaves
50 # the interpreter internal symbols visible to the shared libs that will be
51 # loaded on demand (and will try to reference those symbols).
55 ccdlflags='-Zlink=dynamic -Wl,-usys_nerr -Wl,-Bexport'
58 # Configure sometime finds what it believes to be ndbm header files on the
59 # system and imagines that we have the NDBM library, but we really don't.
60 # There is something there that once resembled ndbm, but it is purely
61 # for internal use in some tool and has been hacked beyond recognition
62 # (or even function :-)
66 # There is a bug in memcmp (which I hope will be fixed soon) which sometimes
67 # fails to provide the correct compare status (it is data dependant), so just
68 # pretend there is no memcmp...
72 # Due to problems with dynamic linking (which I also hope will be fixed soon)
73 # you can't build a libperl.so, the core has to be in the static part of the
78 # PowerMAX OS has support for a few different kinds of filesystems. The
79 # newer "xfs" filesystem does *not* report a reasonable value in the
80 # 'nlinks' field of stat() info for directories (in fact, it is always 1).
81 # Since xfs is the only filesystem which supports partitions bigger than
82 # 2gig and you can't hardly buy a disk that small anymore, xfs is coming in
83 # to greater and greater use, so we pretty much have no choice but to
84 # abandon all hope that number of links will mean anything.
88 # Misc other flags that might be able to change, but I know these work right.