Web lists-archives.com

Re: build* targets as root

(Changing subject line because I think this is a semi-separate issue)

On Wed, 18 Oct 2017 at 11:36:41 +0200, Guillem Jover wrote:
> Apparently this caused mass build failures, all due to packages (or
> their helpers) being very Debian policy non-compliant! These are all
> MUST requirements. Stuff like:
>   - build-targets [...] must be able to be
>     run with root privs (via binary-target invocation)

Which Policy requirements combine to imply this?

If this is really a Policy MUST then I'm not sure how realistic it
is to combine this with build-time testing, particularly if the root
privileges are allowed to be a lie and not actually imply any privilege
(fakeroot). fakeroot has some important limitations: in particular,
anything that has build-time tests involving a protocol that uses AF_UNIX
credentials passing (D-Bus, Wayland, PostgreSQL, some uses of X11)
is going to have a hard time.

I would be reluctant to introduce new incentives for maintainers to
disable build-time testing, given how frequent it seems to be for a
maintainer to turn on build-time tests and discover that their package
has always built but not actually worked on some architectures. If
there were enough autopkgtest workers for all architectures (perhaps
as a requirement for being in testing?) then we could perhaps rely on
autopkgtest to do that smoke-testing, but at the moment Debian only
tests amd64, and Ubuntu only tests 5 architectures (I'm not sure whether
that represents all the supported Ubuntu architectures or not, but it's
certainly fewer than the Debian release architectures).

I also don't think it's reasonable to expect test developers to support
running their tests under fakeroot, and I for one would not like to find
myself trying to explain to an upstream why we had considered this to
be a sensible thing to do. dbus fails tests under fakeroot, but with my
dbus upstream hat on I don't really consider that to be a bug - the tests
are successfully finding inconsistencies in the emulation of real root.

Requiring build-time tests to pass when run as real root (by inserting a
check and skip/early-return into tests that assume they are unprivileged,
if necessary) sounds a much more reasonable thing to expect to work,
although again, I could easily imagine reporting that failure to an
upstream and being told (quite justifiably) that I should never build as
root and the resulting failure was therefore not a bug.

So if Policy indirectly requires the build* targets to work under
fakeroot, I would prefer to see that part of Policy change.