[1899] in NetBSD-Development
Re: Problems building SIPB-Athena on a -current ELF system
daemon@ATHENA.MIT.EDU (Greg Hudson)
Fri Aug 6 10:21:11 1999
To: "Charles M. Hannum" <root@ihack.net>
Cc: netbsd-dev@MIT.EDU
In-Reply-To: Your message of "Fri, 06 Aug 1999 02:29:04 EDT."
<199908060629.CAA13944@bill-the-cat.mit.edu>
Date: Fri, 06 Aug 1999 10:20:51 EDT
From: Greg Hudson <ghudson@MIT.EDU>
> 1) build.sh needs a `-l' option; it loses without it. (The `exec' on
> line 113 blows up because $logfile is empty.)
This was a bug, which was fixed in the Athena mainline. I imported
the fix.
> 2) When building directly in a checked out CVS tree, you need to give
> `-c' to do.sh; this option is likewise not mentioned anywhere.
This shouldn't be the case; -c does something different (use tools,
libraries, and include files from the destination directory).
> 3) The `-b' option to do.sh is not mentioned
There isn't one. do.sh always builds into the current directory.
> nor is it mentioned that build.sh also accepts `-s' and `-d'.
There isn't any documentation of the build.sh options. I'll fix that
later. I've added documentation for do -c and do -n in the Athena
mainline.
> 4) discuss needs to be installed before olc is built, so the
> configure script can find dspipe, but this is not the order
> they're done in normally.
This is a bug in olc. (bert carefully included a hardcoded value for
use when $athena is true, and then managed to make the code ignore
it.) I'll fix.
> 5) third/afsbin has some prebuilt executables and libraries in it,
> which I had to update from my ELF AFS build tree.
This will probably remain the case; building AFS from source isn't
very practical when different platforms are using different versions.
(Not sure what to do about 6. Your fix sounds as good as what I might
come up with.)
> 7) third/kpasswd decides to use /usr/include/kerberosIV, rather than
> /.srvd/usr/athena/include/kerberosIV, and loses because
> KADM_INSECURE_PW isn't defined in the NetBSD native version. (Lots
> of other programs use the wrong include directory, but for the most
> part they seem to at least compile.)
I know what's wrong here; I'll fix it somehow. (ATHENA_KRB4_CHECK is
making the wrong assumptions.)
(Hopefully Dan will fix 8.)
> 9) third/bash and third/ispell fail to build completely because of
> makeinfo lossage. (third/bash fails -I the readline/doc directory,
> and can't some files. third/ispell just has an improperly formed
> Texinfo file.)
If a build winds up running makeinfo, something is wrong. Athena is
constantly plagued by the problem of Makefiles containing rules to
rebuild sources, since CVS will reorder the dates on the source files
when you do an import and checkout. Since we only fix these problems
as we notice them, new ones crop up from time to time, particularly in
source-sipb.
If anyone has a good answer, I'm all ears.
I'll look into fixing the bash and ispell Makefiles. The same problem
applies to 12), incidentally.
> 10) The usual krb5 DB lossage, which causes at least /bin/athena/login
> to lose, because it can't read the passwd database.
krb5-current appears to have renamed all the symbols from things like
__hash_open to __kdb2_hash_open. This is not particularly wise, but
should solve the symptom if not the problem. I'll see if I can get
the krb5 team to fix this problem the right way before krb5 1.1.
In the short run, I'm not sure what to do, since I don't particularly
want to introduce several hundred merge conflicts into the Athena
source tree or source-sipb.
