On Thu, 09 Nov 2000 12:15:46 +0100,
  Anders Franzen <uabfra@uab.ericsson.se> said:

uabfra> When I does 'make build' on the userland , the build fails on functions
uabfra> that initializes a ns_dtab (nsswitch.h). I think that it is the function

uabfra> prototype :
uabfra>        int             (*callback) __P((void *retval,
uabfra>                                                           void *cb_data,

uabfra>                                                        _BSD_VA_LIST_));

uabfra> that doesn't match ?? weird cause it look ok, the only difference
uabfra> between the proto type above and the actual functions that is
uabfra> used is that the _BSD_VA_LIST_ is a va_list in the functions.

I had an opportunity to install NetBSD/sparc64 disklessly to an Ultra
2 in my lab, and do make build with the binary of toolchain. Here are
the problems I found so far:


A. va_list vs _BSD_VA_LIST_

As noted by Anders :)

B. null format string in err(), warn(), etc
C. Y2K in national representation of date in strftime(3)

Since gcc 2.95.2 is kind enough to warn these problems, we are likely
to meet these problems in the other ports as well. I suppose that
toolchain hackers would get interested.

D. DBL_MAX, DBL_MIN, etc is defined more than once


After struggling about a week, I finally managed to do make in
/usr/src. I will take the diff of the source tree right after coming
back to home.

-- 
Seigo Tanimura <tanimura@r.dl.itc.u-tokyo.ac.jp> <tanimura@FreeBSD.org>