On Sun, 7 Dec 2003, Gregg C Levine wrote:

> I imagine people will correct me, on this issue, but on the i386 list,
> which I monitor from a different e-mail address, which is also
> subscribed to this list, they did discuss the merits behind the
> release you've mentioned. That is, using version 3.x versus 2.95.3 for
> building stuff for NetBSD. It seems 3.x is still broken for the VAX
> families, and several others. Not surprisingly, on the Linux-VAX list,
> they are having the devil's own time with building a tool-chain to
> build even a kernel for the VAX.

I'm not too suprised about there being difficulties with the move to gcc
3.x, but I'm out of the loop on the details of it.  I really hope we can
get gcc 3.x working for the VAX before the NetBSD 2.0 release.

> I imagine that each release of the i386 branch is built native. The
> same goes for nearly every single working port in the lists, but what
> about the VAX? Johnny, you're more familiar with that area, then I
> know I am, can you confirm that side of things? The same for you,
> Brian.

I think most of the VAX releases prior to 1.6 were compiled natively on
VAX systems.  Or if not, then certainly this was the case before the 1.5
release.

Tracking -current natively on a VAX can be quite challenging.  Even on a
fairly respectable 6 VUP VS4K/VLC, I'd estimate that it takes around a
week to build a full release from a clean tree.  By the time I hit a
problem in my build, several days may have passed since I ran the cvs
update for the build in progress.  By then, the problem I'm seeing may
already have been fixed-- and of course new and different problems might
also be introduced.

Getting from the 1.6.1 release to -current on the VAX proved to be a
rather big challenge.  Cross-compiling from the 1.6U snapshot on i386
was relatively painless, though the build did have its problems.  Now
that I've got a 1.6ZG running on my VAX, I'm running another native
build to see if it proceeds more smoothly this time around.

-brian.