Re: RFP for bioperl's Bio-EUtilities
Looks like there are two unrelated errors.
The first is from a missing dependency (Bio::ASN1::EntrezGene), which is really optional (the comp test should skip that directory). The other is from XML::Simple, which is unusual; I’m wondering whether the underlying XML parser is checking the XML schema for the test reports. Any idea what the specific XML::SAX backend parser module used was?
chris
On 1/23/17, 1:52 PM, "Andreas Tille" <andreas@an3as.eu> wrote:
[CCed debian-perl@l.d.o - discussion is about test failures of new RFPed
package https://anonscm.debian.org/git/debian-med/libbio-eutilities-perl.git]
Hi Christopher,
On Mon, Jan 23, 2017 at 07:25:38PM +0000, Fields, Christopher J wrote:
> Just ran tests on a non-networked laptop (both latest from Github and the 1.75 release from CPAN) and they were fine, all passed and the relevant network tests are skipped. Can you be more specific as to the build failure?
In the download tarball from CPAN the 1.75 release has only a t directory
(no xt). I've attached the relevant part of the build log.
Hope this helps
Andreas.
> On 1/23/17, 12:46 PM, "Fields, Christopher J" <cjfields@illinois.edu> wrote:
>
> Actually, when looking at this, the network tests are already in ‘xt’ on the repo, and should only tested when making a release:
>
> https://github.com/bioperl/Bio-EUtilities/tree/master/xt/release
>
> I’ll pull down the latest CPAN and run it from a non-networked computer to see if I can pinpoint what the problem is.
>
> chris
>
> On 1/23/17, 9:12 AM, "Fields, Christopher J" <cjfields@illinois.edu> wrote:
>
> My suggestion is to take the avenue of easiest solution first, then most stable solution long-term if needed.
>
> The easiest may be to move the network tests to ‘xt’, which as Carnė mentioned would be run upon release only. This should be easy enough w/ a Dist::Zilla release. Longer term a mock setup may be better.
>
> chris
>
> On 1/23/17, 9:03 AM, "carandraug@gmail.com on behalf of Carnë Draug" <carandraug@gmail.com on behalf of carandraug+dev@gmail.com> wrote:
>
> On 23 January 2017 at 12:40, Andreas Tille <andreas@an3as.eu> wrote:
> > Hi Carnė,
> >
> > On Fri, Jan 20, 2017 at 05:44:00PM +0000, Carnė Draug wrote:
> >> I have filled a RFP (bug # 852004) for bioperl's Bio-EUtilities
> >> package [1]. Unlike Bio-Coordinate, which was split from bioperl and
> >> was recently packaged in Debian, Bio-EUtilities development started
> >> already after bioperl commenced its splitting.
> >>
> >> I was wondering if it was possible for the debian-med team to package
> >> it. While I am not a debian maintainer, I am one of the
> >> Bio-EUtilities developers, have an interest on seeing it packaged in
> >> Debian, and I'm willing to support it upstream.
> >
> > I have commited some initial packaging to
> >
> > https://anonscm.debian.org/git/debian-med/libbio-eutilities-perl.git
> >
> > This build fails due to the failure of several tests - as far as I can
> > see due to the attempt to access the internet. It would help if you
> > could provide an option: "Just do all tests than can be done offline"
> > since the Debian packaging process needs to run fully offline.
> >
> > Kind regards
> >
> > Andreas.
>
> I have asked on #debian-perl if there was any standard method or
> debian preferred method to skip those tests and apparently there is
> none. But there are some suggestions that seem to be common and
> acceptable to the Debian.
>
> 1. check for an environment variable that defines whether network
> tests should be skipped. Some variables used in Debian are
> NO_NETWORK, NOINTERNET, TEST_INTERNET, and NETWORK_TESTING. See for
> example discussion on debian bug #764868 [1]
>
> 2. mock EUtilities using Test::LWP::UserAgent. This is probably the
> most involved but probably the technically most correct way. It has
> the problem of not failing if upstream service ever changes (but maybe
> if that happens and only the testsuite notices, then this whole module
> is not needed).
>
> 3. move all tests from t/ to xt/ so that they are only ran at release
> time by the package author "since obviously if the author isn't
> running their own tests before releasing, you already have a bigger
> problem"
>
> Note that the test must not even attempt network access [2] so
> skipping the test because a ping to the entrez servers fails is also
> not allowed.
>
> I'm CC'ing Chris Fields (also developer on Bio-EUtilities) to discuss
> what would Debian and upstream prefer.
>
> Carnë
>
> [1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=764868
> [2] https://www.debian.org/doc/debian-policy/
>
>
>
>
>
>
--
http://fam-tille.de
Reply to: