Re. http://mail-index.netbsd.org/netbsd-help/2003/08/14/0001.html

Still having problems?

After seeing this message, I updated my system sources to current for
the first time in a while (BTW, assuming that you are talking about
*current*, shouldn't this be on current-users? for releases, you don't
need CVS...).

I found myself (eventually) facing similar errors, but fortunately, one
of the first errors was about the Makefile and missing an "operator".
"Did someone rewrite the Makefile for GNU make?  Surely not!"  I peeked
at the file, and saw it was munged, as if CVS didn't know how to merge
changes.

Of couse, one of the things that I did during the update was to start
a "cvs update -PAd", then I aborted with ^C.  I vaguely recall
that this kind of thing can cause CVS to do stupid things.

In my case, it was the top-level Makefile.  I erased the existing
munge that I could see (they were just comment lines anyway) and tried
again.

THEN I got errors that sounded very similar to what you described.


After a little thought, I just deleted the Makefile (fortunately the
first set of errors told me which file was munged; I simply assumed
more damage in the same file).  Then I asked for another udpate of
*just* the Makefile.

With a fixed Makefile, things seem okay.

You might do a "cvs diff" and see what that says, or (as someone
already suggested) get a fresh checkout of the entire tree.


(Side note: Kernel build times are now taking about 15 minutes.
On the same CPU, older *release* kernels take about 8 to 10 minutes.
Is gcc really that much slower, or is there really that much more work
for it to do with new code?  Or is it because I changed a *few*
options (NetBSD backwards-compat options, ext2fs, lfs, and perhaps
a couple more)?  I wouldn't think that those options would so
significantly affect build-times.)

1.5 to 2 times the compile-time seems like a big jump no matter how
you slice it.

...then again, this is more anecdotal, and based on my wonderful (ahem)
memory, than anything else; I haven't built a release kernel in a while.  (^&

-- 
  "I probably don't know what I'm talking about."  http://www.olib.org/~rkr/