keith@nih-csl.UUCP (keith gorlen) (11/17/89)
I've run into a few problems with the andrew distribution kit, Beta
Release 9/22/89, patch level 4, 10/17/89. I'm running on a
SPARCstation-1 with SunOS 4.0.3c, MIT X11R3, NFS, and no AFS. Any
help with these would be appreciated.
-----
First, I tried to build andrew-beta without the DNS resolver by
placing the following in site.h:
#undef RESOLVER_ENV
Link of "nametimeg" failed because it attempts to use resolver anyway:
cc -g -I/usr/andrew/include/atk -I/usr/andrew/include -I//usr/include/X11 -o nametimeg nmtimeg.o
ld: Undefined symbol
_dn_expand
_putlong
_dn_comp
_putshort
*** Error code 1
make: Fatal error: Command failed for target `nametimeg'
Current working directory /nih/unix/sun4_401/usr/local/bin/andrew-beta/overhead/snap2/nametime
I fixed this by building to use /usr/lib/libresolv.a. Need following
in config/site.mcr:
XUTILDIR = /usr/local/bin
RESOLVLIB = /usr/lib/libresolv.a
-----
One of the more annoying and mysterious problems is that 'sh' fails on
the 'install.time' target in some of the atk subdirectories, for
example:
+ rm -f /usr/andrew/dlib/atk/frame.do
+ pwd
sh: 15522 Memory fault - core dumped
*** Error code 139
make: Fatal error: Command failed for target `install.time'
Current working directory /nih/unix/sun4_401/usr/local/bin/andrew-beta/atk/frame
*** Error code 1
make: Fatal error: Command failed for target `dependInstall'
Current working directory /nih/unix/sun4_401/usr/local/bin/andrew-beta/atk/frame
*** Error code 1
make: Fatal error: Command failed for target `dependInstall'
Current working directory /nih/unix/sun4_401/usr/local/bin/andrew-beta/atk
*** Error code 1
make: Fatal error: Command failed for target `dependInstall'
Current working directory /nih/unix/sun4_401/usr/local/bin/andrew-beta
*** Error code 1
make: Fatal error: Command failed for target `World'
However, if I "cd" into the directory where the failure occurred and
do "make Install" manually, it succeeds, and then I can restart the
build manually. Anyone else seen this behaviour?
-----
After completing the build and install, typescript, console, ez, and
the messages demo seem to work fairly well, but "help" types out its
version number and hangs without bringing up a window.
Thanks in advance.
--
Keith Gorlen phone: (301) 496-1111
Building 12A, Room 2033 uucp: uunet!nih-csl!keith
National Institutes of Health Internet: keith@alw.nih.gov
Bethesda, MD 20892nsb@THUMPER.BELLCORE.COM (Nathaniel Borenstein) (11/17/89)
I think that the ITC people have a patch for the nametimeg bug, which I reported earlier. It may even have been in patch 5, 6, or 7, which you haven't got yet. The problem with help hanging is also one that I reported a little while ago; I believe I fixed it by hand, but this too may have a fix in one of the recent patch files. The problem is that help does some calls related to the resolver and WMHOST, even if you have the resolver turned off. Perhaps someone from the ITC can comment on whether or not these actually made it into one of the last three patch files. -- NB
cfe+@ANDREW.CMU.EDU ("Craig F. Everhart") (11/18/89)
The latest beta patch, and the code that will go to MIT, will not try to build nametimeg unless RESOLVER_ENV is defined. ``nametimeg'' is an application that links directly with the Bind resolver, so it's meaningless without it. This one had slipped through the cracks so far--sorry. I don't know what to say about the ``install.time'' target problems. Does ``pwd'' core-dump on your systems? Were you running out of VM at build time? Craig
guy@auspex.auspex.com (Guy Harris) (11/25/89)
>I don't know what to say about the ``install.time'' target problems.
Since the problems are almost certainly not your fault, there's almost
certainly no obligation for you to say anything about them. There is a
bug in the 4.0 Bourne shell, at least on SPARC machines, which persists
in 4.0.3; you might try filing a bug report with Sun and seeing if they
can get a patched version for you.