so illumos does `curses.ed:7:e !sort +2n ./tmp/keycaps`,

which it seems like the sort in /usr/gnu just stopped supporting, on bloody?

that's annoying, but an illumos bug for not being specific really, but I did a grep for other /sort \+[0-9/ and there's a large handful which I assume also don't work now

(I'd file a bug, but it seems so weird a thing to have broke I'm wondering if I've missed something

this seems to have been an issue introduced to switching building gnu-coreutils with gcc 14 on bloody (and setting the standard to XPG6 to get a clean build). a gnu-coreutils package build with gcc 13 has been published now so `pkg update` should get you back to a working gnu sort

we'll investigate and fix the package "properly" built with gcc 14

thanks!

How did this break the illumos build? I'd not expect GNU sort to be in the mix there.

Weird bug, I wonder what the root cause will turn out to be!

arm64-gate's bootstrap bit failed because libcurses used it without specifying a path or anything.

I pushed a one line workaround earlier

I am very suspicious of ' bool traditional_usage = ! (200112 <= posix_ver && posix_ver < 200809);'

there's a bunch of uses in illumos-gate, but I think that was the only build-time one

which is going to depend on the various whims by which people do or don't reset their path

Yes, it's because we're now building in an XPG6 environment, and sort determines posix_ver to be 200112 versus 199209 before

and chooses to disable traditional usage, including the +/-

It looks like the behaviour of tail, touch and uniq will be different too

_POSIX2_VERSION=199209 /usr/gnu/bin/sort +2n /etc/inet/hosts

^ this works

oof

See github.com/coreutils/gnulib/blob/master/m4/posixver.m4#L13-L18

What's that old quote? The nice thing about standards is there are so many to choose from?

which even calls out this particular example

doubly so when it's just different versions of the same standard.

yuck

standards mess is standards mess, it's the old way silently going away that isn't great

At least the fix is easy..

This is a great example of why if we're going to switch something big like the compiler version, we do it nice and early in the release cycle, to shake these things out. There have been a few surprises with gcc14 so far.

yeah, I mentioned to hadfl that that's why I kept on bloody even though I don't need to anymore

Appreciated :) and thanks for the report. We'll fix it so it won't break again next time we publish new bloody packages.

Night all.

good night