xorg-server 1.13.1, - checksum post-mortem
Peter Hutterer
peter.hutterer at who-t.net
Thu Jan 3 15:38:02 PST 2013
On Sat, Dec 15, 2012 at 08:33:29AM +0100, Matthieu Herrb wrote:
> On Fri, Dec 14, 2012 at 03:04:30PM -0700, Matt Dew wrote:
> >
> > What I didn't realize is that the new tarballs hadn't been uploaded to
> > the server, rather the old ones were there. So the checksums didn't match.
> >
> > I untarred both the tarballs on my local machine and the ones on the fdo
> > server. The ones on my local machine matched the checksums in the
> > announcement I sent out, but the checksums on the fdo server were
> > different, but did infact match the ones a couple of you reported.
> >
> > I diffed the contents of those tarballs. No content differences.
>
> Yes this is one of the things that always annoys me with autotools's
> make distcheck.
>
> The tarballs created by 'make distcheck' contain directory entries
> which are created every time you run make distcheck and thus have
> different creation times every time. So you end up with different tar
> balls everytime.
>
> Moreover gzip also adds the mtime of the compressed file to its
> header, so even if you generate the same archive twice (whith the same
> checksums before compression) you end up with 2 different files once
> gzip'd. bzip2 doesn't include the timestamp so isn't affected by this
> 2nd issue.
is there any particular reason we're still generating gz and bzip2? one
should be enough, isn't it?
Cheers,
Peter
More information about the xorg
mailing list