Subject: Problems I had with a full userland build (1.3.1)
To: None <port-i386@NetBSD.ORG>
From: Kevin Cousins <kevinc@obelix.premium.com.au>
List: port-i386
Date: 04/23/1998 17:18:48
Feeling a little daring the other night, I applied the 1.3.1 source
patches to my 1.3 source tree, then did "make -k" in /usr/src.  Until
then, I was a userland-build virgin :)

[ Mind, running two hundred or so patches through the Emacs
  'epatch-file' utility was a bit of a joke towards the end---it gets
  very sllooowwwww! ]

Next morning, inspecting the build output, several slight problems
were manifest:

1. the compile of /usr/src/lib/libc/string/strftime.c warned about a
   few function declarations that it didn't consider to be full
   prototypes.  I worked this into shape by using ANSI prototypes
   throughout that file instead of the K&R ones that I found there,
   depending on the value of the __STDC__ macro (as is done throughout
   much of the rest of the source tree).  Prior to that change, libc
   wouldn't link in any of it's forms (.a, .po or .so);

2. The build of gdb died because it tries to link to a file called 'gdb'
   while there is already a source subdirectory called 'gdb' (build in
   /usr/src/gnu/usr.bin/gdb, and there there a ./gdb/ directory)!
   D'oh!

   Also, I am still uncertain as to why but at some point,
   /usr/bin/gdb seems to have been removed!  I still haven't found the
   time to poke about and solve this little teaser (building into
   another directory sounds like a good first cut), nor to get the 1.3
   /usr/bin/gdb out of my archived tarballs;

3. Compiling /usr/src/bin/csh/printf.c broke for some reason or other;
   I seem to recall that it was for a reason similar to that outlined
   above in 1. (/bin/csh was left intact), and

4. /usr/src/sbin/mount_nfs/getmntopts.c broke for some reason, I
   haven't yet looked into that, but shall later (though I still have
   a working /sbin/mount_nfs).

Now:

a. Did I do any stupid thing(s) which caused those four breakages, or
   is/are there some simple thing/s that I might have missed which
   would have avoided the problems?

b. Has anyone else seen problems similar to these with the NetBSD/i386 
   1.3.1 patches?

Otherwise, the whole userland build and install went impressively
smoothly---at least from a virgin's perspective! :)

--Kevin.