Re: 2024Q2; uemacs-4.0nb2 drops core on NetBSD-10

[Greg Troxel <gdt%lexort.com@localhost>]

Van Ly <van.ly%sdf.org@localhost> writes:

> First, I installed the binary package from:
>
>  --> https://cdn.netbsd.org/pub/pkgsrc/packages/NetBSD/$arch/10.0_2024Q1/All
>  --> https://cdn.netbsd.org/pub/pkgsrc/packages/NetBSD/$arch/10.0_2023Q4/All
>  --> https://cdn.netbsd.org/pub/pkgsrc/packages/NetBSD/$arch/10.0/All
>
> and uemacs would fail as described.

Those are all old by now.   That's ok to try them - but notably
different than pkgsrc HEAD.

> Second, I built the package locally with non-default settings after
> calling, cvs up -A, when the pkgsrc freeze was posted.  Removed the
> downloaded pkgsrc binary, installed the local build, and experienced the
> uemacs failure.

That's  exactly what you should do to debug.  Technically, all packages
should be  from the same source tree, but in this case I doubt it matters.

> sysctl.conf has the lines:
>
>    ddb.onpanic?=0
>    kern.defcorename=/tmp/%n.core

Those seem extremely unlikely to be relevant.  I asked because of what
was in mk.conf
>
> The non-default settings you see, I assembled from posts on bsd related
> forums.  Is there a page documenting recommended defaults?  The
> impression I get is there's bit entropy in the documentation and it is a
> diy ordeal.

The recommendation is not to set anything.  Basically pkgsrc by default
does what we have decided is best in the general case for normal users.

In general my view is that if you don't understand a config line or
can't explain what it does and why you have chosen it, you probably
shouldn't have it.

> I have three compilers installed:
>
>  - gcc (nb3 20231008) 10.5.0 from netbsd-10 base
>  - gcc (GCC) 12.3.0 from pkgsrc
>  - gcc (GCC) 13.2.0 from pkgsrc

Generally when debugging pkgsrc you should use defaults.  Or at least
start there.  Also, people say gcc is not ABI compatible across
versions, at least for C++.  So mixing 2024Q1 official binary packages
(which are by doctrine built with an essentially empty mk.conf, as far
as things that change the packages go) using the base compiler and
pkgsrc HEAD with gcc12 is quite a departure from the standard approach.

(As for mixing, the license lets you do it, and I'm not telling you not
to.  Just that I don't want to debug the mixed case when things are
vague.  If you get specific info about a bug, that's very useful.)

> Since a working uemacs is observed on netbsd-current, I suppose in time
> the build infrastructure will sync and I should be able to fetch a
> working binary or build locally sticking to the defaults.

I wouldn't assume that at all.  There is a bug someplace, and I think it
is probably in uemacs.   The evidence is consistent with it sometimes
being expressed and sometimes being latent.

pkgsrc-current and netbsd-current are separate concepts.

netbsd-current will eventually become NetBSD 11.  I have not heard an
announcement of timing.  Based on the last two releases I expect it
between 2026 and 2028.

If you care about uemacs working my recommendation is to use a debugger
and find out what is actually wrong with the code.  There is some chance
you'll find a bug not in uemacs, which is also useful of course.