[Rcpp-devel] Fwd: CRAN submission growfunctions 0.1

[Dirk Eddelbuettel]

Terrance,

Good news:  with a most valuable assist from Ryan who really knows the sparse
matrix data structure in Armadillo, this has been fixed in the now-current
version of RcppArmadillo on GitHub [1]    

To recap, it didn't trigger anything outside of ASAN tests, but when running
with the ASAN (address sanitizer) feature some out-of-area memory access was
noticed by the compiler.  This has now been rectified.

Thanks again for the report, and of course to Prof Ripley for running all
those tests on CRAN.

Dirk

[1] The current release is a pre-release of RcppArmadillo 0.4.400.0. Conrad
is preparing for Armadillo 4.400 and made a 4.399.2 available which I
wrapped, and tested against 80 CRAN dependents -- no issues. [2]  Expect a new
release, and CRAN upload, within a week or so. I won't upload this prerelease
as CRAN prefers about one release a month.

[2] Full details as always in the RcppCore/rcpp-logs/ repo. Out of 80, six
packages did not finish R CMD. All built: four lack a package not in Depends,
one requires rgl which fails when running headless, and one (still) has a
stop() condition on what seems to a numerical test that is too tight.

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