This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
To make arithmetic on tainted dualvars work properly requires that
[perl5.git] / hints / apollo.sh
CommitLineData
c2960299 1# Info from Johann Klasek <jk@auto.tuwien.ac.at>
1973dd96 2# Merged by Andy Dougherty <doughera@lafayette.edu>
7269eced
GS
3# Last revised Tue Mar 16 19:12:22 EET 1999 by
4# Jarkko Hietaniemi <jhi@iki.fi>
f1ca563b 5
c2960299
AD
6# uname -a looks like
7# DomainOS newton 10.4.1 bsd4.3 425t
8
9# We want to use both BSD includes and some of the features from the
10# /sys5 includes.
7269eced
GS
11ccflags="$ccflags -A cpu,mathchip -I`pwd`/apollo -I/usr/include -I/sys5/usr/include"
12
13# When Apollo runs a script with "#!", it sets argv[0] to the script name.
14toke_cflags='ccflags="$ccflags -DARG_ZERO_IS_SCRIPT"'
c2960299
AD
15
16# These adjustments are necessary (why?) to compile malloc.c.
17freetype='void'
18i_malloc='undef'
19malloctype='void *'
20
21# This info is left over from perl4.
68dc0745 22cat <<'EOF' >&4
f1ca563b
LW
23Some tests may fail unless you use 'chacl -B'. Also, op/stat
24test 2 may fail occasionally because Apollo doesn't guarantee
25that mtime will be equal to ctime on a newly created unmodified
26file. Finally, the sleep test will sometimes fail. See the
27sleep(3) man page to learn why.
28
c2960299
AD
29See hints/apollo.sh for hints on running h2ph.
30
f1ca563b
LW
31And a note on ccflags:
32
33 Lastly, while -A cpu,mathchip generates optimal code for your DN3500
34 running sr10.3, be aware that you should be using -A cpu,mathlib_sr10
35 if your perl must also run on any machines running sr10.0, sr10.1, or
36 sr10.2. The -A cpu,mathchip option generates code that doesn't work on
37 pre-sr10.3 nodes. See the cc(1) man page for more details.
38 -- Steve Vinoski
39
40EOF
c2960299
AD
41
42# Running h2ph, on the other hand, presents a challenge.
43
44#The perl header files have to be generated with following commands
45
46#sed 's|/usr/include|/sys5/usr/include|g' h2ph >h2ph.new && chmod +x h2ph.new
47#(set cdir=`pwd`; cd /sys5/usr/include; $cdir/h2ph.new sys/* )
48#(set cdir=`pwd`; cd /usr/include; $cdir/h2ph * sys/* machine/*)
49
50#The SYS5 headers (only sys) are overlayed by the BSD headers. It seems
51#all ok, but once I am going into details, a lot of limitations from
52#'h2ph' are coming up. Lines like "#define NODEV (dev_t)(-1)" result in
53#syntax errors as converted by h2ph.
54
55# Generally, h2ph might need a lot of help.