[Bioc-devel] Reproducing erros on build and check

[Martin Morgan]

On 10/17/2011 11:59 PM, Stefan McKinnon Høj-Edwards wrote:
> Hi,
>
> I have just noticed that my package AnnotationFuncs does pass build and check for the next release on the Windows hosts. On the Linux and Max OS X hosts there are no problems.
> The error logs are these:
> - http://bioconductor.org/checkResults/devel/bioc-LATEST/AnnotationFuncs/liverpool-buildsrc.html
> - http://bioconductor.org/checkResults/devel/bioc-LATEST/AnnotationFuncs/moscato1-checksrc.html
>
> My personal workstation is also running Windows (7), and I have checked and built the package using the latest R release (R version 2.14.0 alpha (2011-10-04 r57169)), but I am not able to reproduce the errors displayed on the log at BioConductor.
>
> I have also been looking through my code, and I cannot trace the error giving problems in the build and check above:
>
> Error in function (x)  : no function to return from, jumping to top level
> Execution halted
> Error in function (x)  : attempt to apply non-function
> Execution halted

These errors seem to be an issue with the build system  / R. They show 
up intermitently on all platforms, but much more commonly on our windows 
build machines. We don't know what the precise issue is (maybe 
contention of multiple installation threads for some R-wide resource? a 
memory allocation bug in R?) but these seem to be on our side of the 
fence and not yours. Actually, though, we are quite stumped about the 
root cause so any insights from the broader community would be great...

While on the topic, it can also be disconcerting to see your package 
fail, but tracing to some dependency -- all authors should check 
carefully, especially at this point in the release cycle, that their 
changes work. The recommended strategy for PkgA is to make a local svn 
export

   cd /tmp
   svn export /home/mtmorgan/mypkgs/PkgA

then in the /tmp directory build

   R CMD build PkgA

then check

   R CMD check PkgA_0.99.0.tar.gz

The local svn export makes sure that everything you think is in svn 
actually is. R CMD build creates the vignette; the --no-vignette flag 
can be useful to speed things up for other-than-final checks.

Martin

>
>
> Any suggestions?
>
> Kind regards,
>
> Stefan McKinnon Høj-Edwards              Dept. of Molecular Biology and Genetics
> Ph.D. Fellow                             Aarhus University
>                                           Blichers Allé 20, Postboks 50
>                                           DK-8830 Tjele
> Tel.: +45 8715 7969                      Tel.: +45 8715 6000
> Email: Stefan.Hoj-Edwards at agrsci.dk      Web: www.agrsci.dk
>
> _______________________________________________
> Bioc-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/bioc-devel


-- 
Computational Biology
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N. PO Box 19024 Seattle, WA 98109

Location: M1-B861
Telephone: 206 667-2793