R devel - Jun 2016 - xxx not available for .C in package yyy

Hi R package developers. This is literally the strangest thing I've ever
seen.
Latest (as of a month ago) R under cygwin64.  I'm teaching an intern package
building and using the .C interface.  The package compiles, but when it gets
to "setting up lazy load" or some such it throws the error
"triang" not available for .C()
in package "randpkg".  Upon checking, after removing the current
version of the package
incrementing the package number just to make sure, commenting out the portion of
the
R code that calls .C(), then building, INSTALLING again, this time works. We are
able
to call .C("trian", ...) by hand inside of R, but then when we
uncomment out the call
in the .R file, it throws the error again. We've even retyped that portion
of code, we've
even dos2unix 'd everything.  Any thoughts?


Grant Izmirlian, Ph.D.
Mathematical Statistician
izmirlig at mail.nih.gov

Delivery Address:
9609 Medical Center Dr, RM 5E130
Rockville MD 20850

Postal Address:
BG 9609 RM 5E130 MSC 9789
9609 Medical Center Dr
Bethesda, MD 20892-9789

 ofc:  240-276-7025
 cell: 240-888-7367
  fax: 240-276-7845


	[[alternative HTML version deleted]]

On 20/06/2016 5:47 PM, Izmirlian, Grant (NIH/NCI) [E]
wrote:
> Hi R package developers. This is literally the strangest thing I've
ever seen.
> Latest (as of a month ago) R under cygwin64.  I'm teaching an intern
package
> building and using the .C interface.  The package compiles, but when it
gets
> to "setting up lazy load" or some such it throws the error
"triang" not available for .C()
> in package "randpkg".  Upon checking, after removing the current
version of the package
> incrementing the package number just to make sure, commenting out the
portion of the
> R code that calls .C(), then building, INSTALLING again, this time works.
We are able
> to call .C("trian", ...) by hand inside of R, but then when we
uncomment out the call
> in the .R file, it throws the error again. We've even retyped that
portion of code, we've
> even dos2unix 'd everything.  Any thoughts?
We don't support Cygwin builds.

If you get this using one of our builds, put together a simple 
reproducible instance of it and make it available.  If it's only in 
Cygwin, you should do the same, and send it to someone there.

Duncan Murdoch

Grant,

On 20 June 2016 at 21:47, Izmirlian, Grant (NIH/NCI) [E] wrote:
| Hi R package developers. This is literally the strangest thing I've ever
seen.
| Latest (as of a month ago) R under cygwin64.  I'm teaching an intern
package
| building and using the .C interface.  The package compiles, but when it gets
| to "setting up lazy load" or some such it throws the error
"triang" not available for .C()
| in package "randpkg".  Upon checking, after removing the current
version of the package
| incrementing the package number just to make sure, commenting out the portion
of the
| R code that calls .C(), then building, INSTALLING again, this time works. We
are able
| to call .C("trian", ...) by hand inside of R, but then when we
uncomment out the call
| in the .R file, it throws the error again. We've even retyped that portion
of code, we've
| even dos2unix 'd everything.  Any thoughts?

In the top-level file NAMESPACE, do you have

   useDynLib("randpkg")

to ensure the dynamic library is loaded?

Dirk

PS Please consider using .Call() instead of .C(). 

--
http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org