Re: understanding 2025Q2 in aarch64

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

Ramiro Aceves <ea1abz%gmail.com@localhost> writes:

> Surfing ftp.netbsd.org I notice that in amd64:
>
> 10.1 -> 10.0 and
> 10.0 -> 10.0_2025Q2
>
> but in aarch64:
>
> 10.1 -> 10.0 and
> 10.0 -> 10.0_2025Q1

Just to let you know how it really is on the ftp server (which I realize
you can't inspect, and http doesn't really expose this):

   gdt 9 /ftp/pub/pkgsrc/packages/NetBSD > ls -ld amd64 aarch64 x86_64
   drwxrwxr-x  8           netbsd  512 Jul 22 06:10 aarch64
   lrwxrwxr-x  1           netbsd    6 May 28  2008 amd64 -> x86_64
   drwxrwxr-x  9           netbsd  512 Jul 18 18:04 x86_64

   gdt 10 /ftp/pub/pkgsrc/packages/NetBSD > ls -ld x86_64/10.*
   lrwxrwxr-x  1           netbsd   11 Jul 18 18:04 x86_64/10.0 -> 10.0_2025Q2
   drwxrwxr-x  3           netbsd  512 Mar 15 00:43 x86_64/10.0_2024Q4
   drwxrwxr-x  3           netbsd  512 Jun 10 12:43 x86_64/10.0_2025Q1
   drwxrwxr-x  3           netbsd  512 Aug  5 12:43 x86_64/10.0_2025Q2
   lrwxr-xr-x  1           netbsd    4 Dec 18  2024 x86_64/10.1 -> 10.0

   gdt 11 /ftp/pub/pkgsrc/packages/NetBSD > ls -ld aarch64/10*
   lrwxrwxr-x  1          netbsd   11 Apr 12 16:19 aarch64/10.0 -> 10.0_2025Q1
   drwxrwxr-x  3          netbsd  512 Dec 31  2024 aarch64/10.0_2024Q4
   drwxrwxr-x  3          netbsd  512 Apr 12 11:04 aarch64/10.0_2025Q1
   drwxrwxr-x  3          netbsd  512 Jul 18 09:59 aarch64/10.0_2025Q2
   lrwxr-xr-x  1          netbsd    4 Dec 18  2024 aarch64/10.1 -> 10.0

As you can see, the directories named for MACHINE_ARCH are the ones with
actual bits, and MACHINE aliases (e.g. amd64) are symlinks.

10.1 is just a symlink to 10.0, for people using their version instead
of N.0 in pkgin repo specs, if any such people exist, and is minor
clutter, if not.  It's tradition.  Similarly for 9.x -> 9.0, x > 0.

Directories 10.0_yyyyQn have contents (in All).

In theory, the 10.0 symlink points to the most recent quarterly build
that is believed to be more or less done and at least more or less ok
(which is usually really done and really ok!).

Further:

  gdt 13 /ftp/pub/pkgsrc/packages/NetBSD > du -s aarch64/10.0_202*
  72920264        aarch64/10.0_2024Q4
  76081388        aarch64/10.0_2025Q1
  76848848        aarch64/10.0_2025Q2

  gdt 15 /ftp/pub/pkgsrc/packages/NetBSD > for d in aarch64/10.0_202*; do echo -n "$d: "; ls $d/All|wc -l; done
  aarch64/10.0_2024Q4:    26459
  aarch64/10.0_2025Q1:    27475
  aarch64/10.0_2025Q2:    26514

but beware that size and package count is not as meaningful as you might
think, because when we stop building for an ancient python version, the
count drops a lot, but nothing that matters is lost.  However, these
values are a clue that 2025Q2 is at least pretty ok, and provide no
reason to suspect it is not ok.

> Does this mean that the contents of
>
> /pub/pkgsrc/packages/NetBSD/aarch64/10.0_2025Q2
>
> are not yet available for download?

No.  They should be downloadable.  You will need to change your pkgin
repository URL to have that path explicitly, rather than 10.0.  It's
always been possible for people to do that if they want to be early, or
if they want to stay on a package set after the ftp site has moved
forward.

I just used ftp(1) on NetBSD 10.0 amd64, and downloaded the aarch64
10.0_2025Q2 zip package, with no issues once I remembered the confusing
maze of twisty directory names (pub/pkgsrc/packages/NetBSD not
pub/NetBSD/packages!).

I build my own binary packages, just because I do, so I haven't been
paying attention to this lately.