On Thu, Feb 27, 2025 at 12:01:07PM +0100, Paul Gevers wrote: > Hi, > > On 27-02-2025 11:04, Francesco Poli (wintermute) wrote: > > It seems that the tests that need Internet access (and are marked as > > "needs-internet" in 'debian/tests/control') fail with the following error: > > > > 302s /usr/bin/auto-apt-proxy: 176: avahi-browse: not found > > 303s E: Failed to open TCP connection to :80 (Connection refused - > > connect(2) for "" port 80) > > 303s E: Exiting with error
This failure mode ("avahi-browse: not found") is a regression in
auto-apt-proxy 16.2. It should check whether avahi-browser is installed
before using it (as documented). I will upload a fixed version soon.
> > It should be noted that apt-listbugs will attempt to run an auto proxy
> > detect command, if one is found in APT configuration
> > ( acquire::http::proxy-auto-detect ).
> > I guess the testbed has 'auto-apt-proxy' installed, but is it working
> > correctly?
>
>
> Unfortunately, not all our riscv64 workers are configured in the same way,
> as in one location we have a dedicated local proxy, and in the other we
> don't. All the workers nodes in Shanghai have a proxy configured and I think
> autopkgtest picks that up and installs auto-apt-proxy in the testbed and
> that seems to work for the purpose of downloading packages (I think the
> warning about avahi-browser is only a warning). The nodes in Beijing don't
> have a dedicated proxy and instead have apt-cacher-ng installed on the node
> itself. I didn't install auto-apt-proxy on those nodes because that didn't
> seem to work (sorry for not filling a bug report), but autopkgtest on those
> nodes seemed to work nicely without it. I'm not sure if auto-apt-proxy gets
> installed on those testbeds. I'll test that shortly.
>
> > Is there anything wrong in the tests defined in package 'apt-listbugs'?
> >
> > Is there anything that fails to work in the riscv64 Debian CI testbed?
> >
> > Is there anything that needs to be fixed anywhere else?
>
>
> Maybe.
>
> I have re-triggered your test several times and it succeeded once on a
> Beijing host, so that should get it out of your way. Thanks for letting us
> know.
I think this one is on me.
signature.asc
Description: PGP signature
