Different variants of the same package on ftp.FreeBSD.org

[Lena at lena.kiev.ua]

On Thu, Jan 31, 2008 at 10:37:38AM +0100, Pav Lucistnik wrote:

> > > > ~ $ host ftp.freebsd.org
> > > > ftp.freebsd.org has address 204.152.184.73
> > > > ftp.freebsd.org has address 62.243.72.50
> > > > ftp.freebsd.org has IPv6 address 2001:6c8:6:4::7
> > > > ftp.freebsd.org has IPv6 address 2001:4f8:0:2::e
> > > > ~ $ echo -e "dir mathopd*\nquit" | ftp ftp://204.152.184.73/pub/FreeBSD/ports/i386/packages-6-stable/All/
> > > > Connected to 204.152.184.73.
> > > > 220 Welcome to freebsd.isc.org.
> > > > ...
> > > > -rw-r--r--    1 110      1002        56584 Jan 24 06:20 mathopd-1.5p6.tbz
> > > > ~ $ echo -e "dir mathopd*\nquit" | ftp ftp://62.243.72.50/pub/FreeBSD/ports/i386/packages-6-stable/All/
> > > > Connected to 62.243.72.50.
> > > > 220 ftp.FreeBSD.org NcFTPd Server (licensed copy) ready.
> > > > ...
> > > > -rw-r--r--   1 ftpuser  ftpusers     56520 Nov  4 14:48 mathopd-1.5p6.tbz
> > > > 
> > > > Different sizes, different dates. The November variant is broken
> > > > (installed package crashes at first connection), the January variant works.
> > > > I.e. half of users get the broken variant from ftp.FreeBSD.org.
> > > > Local mirrors contain the broken variant.
> > > > http://www.freshports.org/www/mathopd  lists last commit on August 13, 2007.
> > > > How, why different variants?
> > > 
> > > I'm not able to find out why the files are there on the 62.243 one,
> > > but they are from 6.2 while the one from January 24 is the FreeBSD
> > > 6.3 one.
> > 
> > The November variant crashes under 6.2-RELEASE-p1.
> > If I build mathopd from ports under the same 6.2-RELEASE-p1 then it works.
> 
> ftp-master have the January 24. Guess the mirrors are behind in mirroring the
> stuff. :)

I already found out how to get working variant. I want to attract attention
to the fact that for several months ftp.freebsd.org had the broken variant
which crashes under 6.2 though building from port under my 6.2 gives
working package. The mathopd port hasn't depencencies at all, one binary.
How the broken variant was compiled?
If someone wants to investigate, I uploaded the broken November 4 package at
http://lena.kiev.ua/broken-mathopd-1.5p6.tbz