SIZE different between MASTER_SITE(s) and ftp.FreeBSD.org for some ports

[Ion-Mihai Tetcu]

On Mon, 2 Feb 2004 17:11:27 -0500
Jason Harris <jharris at widomaker.com> wrote:

> In gmane.os.freebsd.devel.ports, you wrote:
> 
> >> =====> Fetching for /usr/ports/misc/floatator/
> > 
> > The distfile was re-rolled but "diff -burN" showed no changes.  I added
> > the new MD5 and size.
> 
> Unfortunately, this results in:
> 
>   fetch -S 52768 52770 http://floatator.cichon.com/download/floatator-0.2.1.tar.gz
> 
> So, I think removing the first size is the best (quick) fix, since those
> who have the original file won't be calling fetch and those who don't
> will likely grab it from a MASTER_SITE.
>
>  But if the MASTER_SITEs are
> down and the distfile caches report the file size (and have the original
> distfile), this will make the fetch fail unnecessarily.

I would prefer bumping port_revision or something; if that's
unacceptable change the SIZE and the md5 and let people fetch again. The
idea of having 2 "same" distfiles with different md5 and sizes just
doesn't seems right (and will likely produce more noise).

> However, unless bsd.port.mk (and fetch(1)) can be made to work properly
> with multiple SIZEs,

Not impossible, but unlikely as this will also require multiple MD5, if
I'm not missing something here; and my first thought would be something
very bad about security.

> I think removing the pre-reroll MD5 and SIZE and
> making the new distfile propagate to all distfile caches is the best
> long-term fix.  All MASTER_SITEs which are supposed to be mirrors but
> have the pre-reroll distfile(s) can be quickly skipped if they report
> file sizes or commented out if they result in unnecessary fetches.
> (Removing SIZEs and/or allowing outdated mirrors to go unnoticed until
> the next software release are not good alternatives, IMO.)



-- 
IOnut
Unregistered ;) FreeBSD user