Re: [libvirt] [PATCH] Don't ship autogen.sh in release tarballs
On 04/13/2017 10:19 AM, Daniel P. Berrange wrote:
On Thu, Apr 13, 2017 at 10:11:50AM -0500, Eric Blake wrote:
> On 04/13/2017 10:06 AM, Andrea Bolognani wrote:
>> autogen.sh is only useful for developers, not users, and we
>> expect developers to have a git checkout handy, so there's
>> no point in shipping the script in release tarballs.
>
> I'm worried this breaks the GPL. autogen.sh is our preferred way for
> rebuilding autotools in preparation for a release, and thus I think the
> script belongs in a tarball even if it is not expected to be used by the
> end user.
Aside from the licensing question, IMHO, the tar.xz should always contain
all the files we have in GIT, plus whatever auto-generated files we decide
are needed. If nothing else that gives users clear visibility into what
generated the auto-generated files, even if they don't need to re-run that
auto-generation process themselves.
As another data point, recent coreutils has moved towards the tarball
only having recent changes, plus documentation about how to find the
source repository for older changelogs, storing the older changes only
in the source repository. But again, licensing wise, the changelogs are
NOT the preferred source form for creating a tarball, while the
bootstrap and autogen.sh scripts ARE, so there's a big difference
between trimming tarball size by dropping build scripts vs. trimming
tarball size by dropping data that is not a build impact.
You have to have a really strong reason for making a tarball that is not
a superset of a git checkout, as it gets very hard to prove that the
tarball is then sufficient to create a fork (and the fact remains that
the GPL requires anyone getting a libvirt binary be afforded the chance
to fork from the same sources used to build that binary).
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
Attachments:
- signature.asc (application/pgp-signature — 604 bytes)