misc/41255: gcc 4.3.3 won't compile some of the source tree

[lidl%pix.net@localhost]

>Number:         41255
>Category:       misc
>Synopsis:       gcc 4.3.3 won't compile some of the source tree
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    misc-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Apr 21 02:40:00 +0000 2009
>Originator:     Kurt Lidl
>Release:        netbsd-5
>Organization:
>Environment:
SunOS lsl.pix.net 5.11 snv_110 i86pc i386 i86pc
(Cross compiling for netbsd-5 on a new target architecture)

>Description:
Gcc 4.3.3 is more picky that the in-tree gcc 4.1.2 that netbsd-5
has.  However, it supports the processor (Freescale Coldfire) that
is being used for this project.

crypto/dist/heimdal/lib/hx509/softp11.c
-- missing va_start before varargs processing

dist/ntp/ntpd/ntp_loopfilter.c
-- bad code in #ifdef/#else clause that will not parse

sys/arch/sparc/dev/cgfourteen.c
-- single quote in comment that gcc will not accept

sys/fs/hfs/libhfs.c
-- Initialize a data structure that gcc cannot determine if it
   used or not during this function.  This patch makes it
   compile, but I'm not 100% sure it is correct.  Feedback
   welcome.

sys/kern/vfs_wapbl.c
-- Comment in #ifdef/#endif block that gcc will not parse,
   just change it to be a proper "C" multi-line comment

sys/nfs/nfs_vnops.c
-- I think this should be a pid_t and not a short, but am not
   100% sure that it matters.  Feedback welcome.

sys/rump/librump/rumpuser/rumpuser_pth.c
-- Mark the function as __dead so it matches its prototype,
   and then gcc 4.3.3 will compile it.

tests/Makefile
-- OK, this isn't actually a "gcc" problem, but tests/net relies
   on ATF, so if MKATF==no, don't try the "net" directory either

usr.bin/mail/lex.c
-- Get rid of "might be clobbered by `longjmp' or `vfork'" error.
   This might be an inappropriate way to fix the problem.
   Feedback welcome.

usr.bin/mixerctl/mixerctl.c
-- I don't see how this could have worked before, properly.

usr.bin/rdist/expand.c
-- #ifdef out a unused static function

usr.bin/systat/main.c
-- Use strcmp to figure out argment match.  I'm not sure this is
   correct or not.  The way that argument parsing is done in this
   program seems kinda fragile to me.

>How-To-Repeat:
Build a gcc 4.3.3 cross-compiler, cross-build netbsd-5 with that
compiler, watch the problems roll in.

>Fix:
I will attach a diff to this PR once I get the number back from the
server.