# Info from Johann Klasek <[email protected]>
# Merged by Andy Dougherty <[email protected]>
# Last revised Tue Mar 16 19:12:22 EET 1999 by
# Jarkko Hietaniemi <[email protected]>
# uname -a looks like
# DomainOS newton 10.4.1 bsd4.3 425t
# We want to use both BSD includes and some of the features from the
# /sys5 includes.
ccflags="$ccflags -A cpu,mathchip -I`pwd`/apollo -I/usr/include -I/sys5/usr/include"
# When Apollo runs a script with "#!", it sets argv[0] to the script name.
toke_cflags='ccflags="$ccflags -DARG_ZERO_IS_SCRIPT"'
# These adjustments are necessary (why?) to compile malloc.c.
freetype='void'
i_malloc='undef'
malloctype='void *'
# This info is left over from perl4.
cat <<'EOF' >&4
Some tests may fail unless you use 'chacl -B'. Also, op/stat
test 2 may fail occasionally because Apollo doesn't guarantee
that mtime will be equal to ctime on a newly created unmodified
file. Finally, the sleep test will sometimes fail. See the
sleep(3) man page to learn why.
See hints/apollo.sh for hints on running h2ph.
And a note on ccflags:
Lastly, while -A cpu,mathchip generates optimal code for your DN3500
running sr10.3, be aware that you should be using -A cpu,mathlib_sr10
if your perl must also run on any machines running sr10.0, sr10.1, or
sr10.2. The -A cpu,mathchip option generates code that doesn't work on
pre-sr10.3 nodes. See the cc(1) man page for more details.
-- Steve Vinoski
EOF
# Running h2ph, on the other hand, presents a challenge.
#The perl header files have to be generated with following commands
#sed 's|/usr/include|/sys5/usr/include|g' h2ph >h2ph.new && chmod +x h2ph.new
#(set cdir=`pwd`; cd /sys5/usr/include; $cdir/h2ph.new sys/* )
#(set cdir=`pwd`; cd /usr/include; $cdir/h2ph * sys/* machine/*)
#The SYS5 headers (only sys) are overlayed by the BSD headers. It seems
#all ok, but once I am going into details, a lot of limitations from
#'h2ph' are coming up. Lines like "#define NODEV (dev_t)(-1)" result in
#syntax errors as converted by h2ph.
# Generally, h2ph might need a lot of help.
|
