Re: portsnap snapshot corruption?

On Fri, Sep 21, 2007 at 02:56:57AM +0400, Boris Samorodov wrote:
> On Thu, 20 Sep 2007 15:43:36 -0700 Marcus Reid wrote:
> > On Thu, Sep 20, 2007 at 03:30:15PM -0700, Marcus Reid wrote:
> > > On Thu, Sep 20, 2007 at 09:46:06PM +0300, Giorgos Keramidas wrote:
> > > > On 2007-09-20 21:42, Giorgos Keramidas <keramida_at_FreeBSD.org> wrote:
> > > > >
> > > > > Is something wrong with the portsnap servers, or should I try to see if
> > > > > there's something odd with my latest CURRENT upgrade?
> > > > 
> > > > Damn, right after having spent an hour on this *and* posting a message,
> > > > our local admin notified us that our network has started using a
> > > > transparent proxy -- which is apparently broken.
> > > > 
> > > > Sorry for the noise.  I'll try to resolve this with our IT guys :)
> > > 
> > > I've been having the same problem and I'm not behind any sort of proxy.
> > > Also, what sort of proxy would corrupt that 49MB gzipped tar file in a
> > > way that it passes gunzip -t?  I think this is a real problem with
> > > the snapshots being served up by the portsnap servers.  I posted this
> > > issue to freebsd-ports_at_ yesterday.
> 
> > I should provide a couple more details of what I've seen:
> 
> >   - The portsnap fetch downloads the 49MB .tgz, and it extracts
> >     properly.  This file is not corrupted (passes gzip checksum).
> 
> >   - In the portsnap/snap directory, many of the <checksum>.gz
> >     files are correct, but many (looks like 1/4?) of them are
> >     truncated somehow.  They are gzip files, but they are corrupted.
> 
> Isn't it related to a recent libarchive backout?:
> . src/lib/libarchive/archive_write_disk.c rev.1.16
> . src/lib/libarchive/test/test_write_disk.c rev.1.5

Yes, that would be it.  Both of the machines that I was testing on
have the version with the regressions.  Thanks!

Marcus