Re: Please track you new packages in Blends tasks (Was: pullseq_1.0.2-1_amd64.changes is NEW)
Am 31.03.21 um 06:11 schrieb Andreas Tille:
> Hi Nilesh,
>
> On Wed, Mar 31, 2021 at 01:02:01AM +0530, Nilesh Patra wrote:
>>> I'm very happy to see your performance to add new packages to the pool.
>>> That's really cool. It would be nice if you would add packages you
>>> are working on to the relevent Debian Med task even if it is only in
>>> Git (not even uploaded to new). The information is gathered with about
>>> 24 hours delay. It helps to keep our metapackages up to date and is
>>> also helpful to get our users informed what we are doing.
>> If I'm not mistaken, I need to add it here[1], right?
> Yes.
>
>> Since these NEW packages are not yet accepted, is
>> adding in "Recommends:" still OK?
> Definitely. The toolset will deal with this. The code to generate the
> metapackages will decrease any Recommends that can not be resolved in
> *testing* to Suggests automatically. So even if we might loose some
> packages due to RC bugs (I guess dicompyler is some candidate for this
> due to [2]) the metapackages will be re-rendered short before the real
> release to reflect this loss. There is no need to edit the tasks file
> for this.
>
>> (Or is this done with the ulterior motive that these will be accepted
>> someday?)
> The tasks files are containing packages in Git only also with
> recommends. Yes, once they migrated to *testing* the metapackages will
> include the packages. The web site will render the packages in the
> according section. Please use "Suggests" *only* if the package is of
> some lower relevance for the task.
We never got to it, but maybe now with the workflows this would
something to consider preparing for our Debian Med policy. In my mind,
you recommend a package if that problem maps to a biological problem.
So, for RNA-seq I would recommend any of the RNA-seq pipelines that we
have already packaged. The filtering of SNPs with a particular property
I would consider to be a sub-problem of that analysis and not recommend
it but only suggest it. But SNP-filtering you also need for discussing
genome-wide association studies in large cohorts that can also come from
microarray data? Fine, then the workflow for that is recommended and it
would be another pipeline that then drags in the SNP-filtering tool.
The libraries Debian Med maintains I would also put as suggested only.
We currently have no means to express that we need to have an eye on a
dependency even though that library with the Java team or the Python
team. I mean, we have all our Debian QA pages, but this is not reflected
by the tasks - and should not IMHO.
So, I suggest to follow Andreas' instructions and at some point we just
need to talk about all the packages.
Best,
Steffen
Reply to: