Re: Acer Aspire One NetBSD 11.0-RC1 slow man pages rendering in virtual consoles

[RVP <rvp%SDF.ORG@localhost>]

On Tue, 3 Mar 2026, Greg A. Woods wrote:

> At Tue, 3 Mar 2026 22:34:01 +0000 (UTC), RVP <rvp%SDF.ORG@localhost> wrote:
> > Edit /etc/ttys and set `constty' (vt100) to `off', and instead, set `ttyE0'
> > to `on'.
>
> That's arguably wrong, but the rest is very right.

Well, I do that because the other way around (constty on; ttyE0 off) leads
to

1. an inconsistency in the name of the 1st console which leads to issues of
   this sort:

https://mail-index.netbsd.org/pkgsrc-users/2024/03/23/msg039239.html

```
$ tty
/dev/constty

$ sysctl -n kern.consdev
ttyE0
```
and,

2. framebuffer graphics code not working because /dev/constty doesn't support
   mmap()--it does happily support all the ioctl()s leading up to it.

OTOH, I've found no issues with /dev/ttyE0 being the "console".

> Arguably the real bug is in the default /etc/ttys configuration.
>
> For default installs on the majority of "headfull" systems using wscons
> emulation the constty terminal type should be wsvt25m (or maybe wsvt25,
> see below).
> [...]
> BTW, I added the following comments to my copy of terminfo to make the
> difference between wsvt25 and wsvt25m more clear (and to elucidate the
> confusion caused by basing wsvt25 on vt220, which also has various
> delays encoded in its terminfo entry, though not so many):
> [..]

Ah, thanks for that. I didn't a `wsvt25m' type existed. I've been using the
`.metaesc' kbd variant in /etc/wscons.conf to compensate.

Two FYIs for the people reading this:

1. less(1) is _not_ slow on 10.x. It is only slow on 11.x because this change
   was made to less in the interim: less-563 -> less-581.2
   (own implementation of tputs(3))

https://github.com/gwsw/less/commit/0af04471e9f80b7d25b38b2d309b98316bb33703

2. And, less, even with that patch, won't be slow on FreeBSD or Linux because
   they're both tickless, so a 2ms delay will be a 2ms delay. On NetBSD, a
   2ms or 5ms delay will be rounded up to 1/HZ (10 ms, typically), and all
   those 10ms delays on each line will add up!

   Change HZ to 1000 and NetBSD will be on par with FreeBSD and Linux, I think.

-RVP