I dunno if I should bother reporting this, or to whom, so if I'm doing
this wrong, please flame.

I've been tracking the -current kernel for a few weeks now, doing a sup
and building from scratch overnight.  For the last few days, the build
has been crapping out with:

==========

cc  -O2 -Wall -Werror -Wstrict-prototypes -Wmissing-prototypes  -Wpointer-arith -Wno-main -G 0  -mno-abicalls -mno-half-pic -I. -I../../../../arch -I../../../.. -nostdinc -I../../../../../gnu/sys -DMIPS1 -DLKM -DNKMEMCLUSTERS=1024 -DHZ=256 -DMAXUSERS=64 -D_KERNEL -Dpmax  -c ../../../../arch/pmax/dev/rcons.c
cc1: warnings being treated as errors
../../../../arch/pmax/dev/rcons.c: In function `rcons_connect':
../../../../arch/pmax/dev/rcons.c:147: warning: passing arg 2 discards `const' from pointer target type
*** Error code 1

Stop.

==========

This is happening on the line:

        if (ri.ri_depth == 8 && info->fi_type.fb_boardtype != PMAX_FBTYPE_MFB)
                info->fi_driver->fbd_putcmap(info, rasops_cmap, 0, 256);

because "rasops_cmap" is "extern const u_char", and because the function
pointed to by fbd_putcmap is

        int     (*fbd_putcmap) __P ((struct fbinfo *, caddr_t, int, int));

in arch/pmax/include/fbvar.h.

(By the way, as one of the tiniest nits you'll ever see picked, fbvar.h does
not end with a newline ...).

I fixed this by casting rasops_cmap to caddr_t ...

The resultant kernel, built from sources supped at 0800 UTC on 12/08, fails
to load with an undefined reference to ibus_pmax_ca, so I'm going to go find
that next ...

FWIW, I'm running a stock 1.4.1 userland with an updated config(8) and new
configuration files based on GENERIC 1.56.

Terry R. Friedrichsen

terry@venus.sunquest.com