[Rd] uninformative error message when building package: ‘x’ must be an atomic vector

Dan Tenenbaum dtenenba at fhcrc.org
Tue Oct 8 00:40:45 CEST 2013


Hi,

When building this Bioconductor package with R-devel:

http://www.bioconductor.org/packages/devel/bioc/src/contrib/ROntoTools_1.1.2.tar.gz

I get the following:

[...]
Error: processing vignette ‘rontotools.Rnw’ failed with diagnostics:
‘x’ must be an atomic vector
Execution halted

If I install the package, Stangle and then source the vignette, I see that the "real" error is:

[...]
> ###################################################
> ### code chunk number 14: rontotools.Rnw:177-178
> ###################################################
> peRes <- pe(x = fc, graphs = kpg, ref = ref,  nboot = 200, verbose = FALSE)
Error in La.svd(x, nu, nv) : error code 1 from Lapack routine 'dgesdd'

It would be great if R CMD build could produce this error instead of the uninformative 
‘x’ must be an atomic vector. 

I see this with many different packages and I always have to do the same dance in order to find out what is really going on.

BTW, I don't think it's referring to the x in the line of code right before the error, because that x is an atomic vector:

> class(fc)
[1] "numeric"
> length(fc)
[1] 2864

Thanks,
Dan



More information about the R-devel mailing list