Hi Tony,
On Thu, Jul 23, 2020 at 11:41:40AM +0100, Tony Travis wrote:
> On 23/07/2020 07:56, Andreas Tille wrote:
> > > > Create sativas_GWAS.covar file...
> > > > smartpca -p sativas_GWAS.par > sativas_GWAS-smartpca.log
> > > > OOPS -i flag not specified at /usr/bin/smartpca line 21.
> > > > pique: error - Can't read sativas_GWAS.evec
> > > > make: *** [Makefile:29: input] Fehler 255
>
> The "pique-input" script should be running /usr/lib/eigensoft/smartpca, not
> /usr/bin/smartpca and my version does:
I'll check this. However, in principle I do not consider it the best
idea to use some program that is private to eigensoft and ignore the
available user interface in /usr/bin.
> > root@beluga:/opt/local/src/GWAS/PIQUE/test# make |& tee make.log
> > [...] > root@beluga:/opt/local/src/GWAS/PIQUE/test# fgrep smartpca *.log
> > make.log:/usr/lib/eigensoft/smartpca -p sativas_GWAS.par >
> > sativas_GWAS-smartpca.log
> > sativas_GWAS-pique-input.log:/usr/lib/eigensoft/smartpca -p sativas_GWAS.par
>
> Please will you try a "git pull" to make sure your copy of my PIQUE GitHub
> repo is up-to-date - Thanks.
For Debian we usually trust *released* version (if available) so
if you would please tag what should be packaged I could package
your tagged release.
> > libparallel-forkmanager-perl is in the list of Depends of the pique
> > package uploaded to new - so this should be settled.
>
> Yes, I understand that - I'm just letting you know that all the other
> dependencies of PIQUE are already satisfied by installing the "med-bio" and
> "med-bio-dev" meta-packages, which is a vast improvement over the way we
> used to install it in the bad old days :-)hadhad
I just want to make sure you will not create the false impression to
naive users that installing med-* is needed to just run pique (which
meanwhile was accepted in unstable).
> > OOPS -i flag not specified at /usr/bin/smartpca line 21.
> >
> > message? The version that was used in this test is 7.2.1+dfsg-1.
>
> See above, "pique-input" should run "/usr/lib/eigensoft/smartpca".
I can check this on your fresh release - but my recommendation is
not to do this.
> > also uses wget to download data. Could you provide a simple shell
> > script
> > that is using only the available data to test the functionality of pique
> > a) in a build time test
> > b) as autopkgtest
>
> It's not possible to test PIQUE without a data-set large enough to use for
> GWAS, which is why I use "wget" to download one and avoid storing large blob
> in the GitHub repo. The data-sets we normally use contain millions of SNPS,
> but this small data-set is what we used during the development of PIQUE and
> what we use when teaching how to use it.
>
> Not sure if that helps?
The information itself helps. As far as I know its permitted in autopkgtest
to download data - so if we know there is no better way than we can try
this.
Kind regards
Andreas.
--
http://fam-tille.de
