Subject: | |
From: | |
Reply To: | |
Date: | Tue, 14 Oct 2014 22:06:20 -0400 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
On Tue, Oct 14, 2014 at 1:10 PM, Kenneth MacDonald
<[log in to unmask]> wrote:
> While experimenting with a local mirror over rsync, I notice that the
> Scientific Linux doesn't seem to be maintaining hard links between
> source RPMs anymore.
Seems to work for me at mirrors.kernel.org. Note that the *mirror* may
not have hard links configured correctly with the "rsync -H" option.
> For example ...
>
> 6.4/SRPMS/vendor/at-3.1.10-44.el6_5.2.src.rpm
> 6.5/SRPMS/vendor/at-3.1.10-44.el6_5.2.src.rpm
>
> However, the older files ...
>
> 6.4/SRPMS/vendor/at-3.1.10-42.el6.src.rpm
> 6.5/SRPMS/vendor/at-3.1.10-42.el6.src.rpm
>
> ... are hard linked.
If you're making a local mirror you might try installing the
'hardlink' program, and running 'hardlink -n' on your local mirror.
> In 6.[45]/SRPMS/vendor, the oldest non-hardlinked files are both ...
>
> -rw-r--r-- 1 root root 2254830 Oct 15 2012 tar-1.23-9.el6.src.rpm
>
> But they have the same md5sum, so I would expect them to be hard linked.
Checksum != timestamp and owner and permissions, which are normally
set before checked before making hardlinks. Also, if they're created
by doing some more comlex process than running 'hardlink', it's easy
to make small mistakes.
> Is this something that is already known? Perhaps it's the upstream
> vendor's change in behaviour?
>
> Cheers,
>
> Kenny.
|
|
|