search

rcannood / princurve

Fits a Principal Curve in Arbitrary Dimension ⤵
41 stars 6 forks source link

Rchk calling allocating function #30

Closed rcannood closed 6 years ago

rcannood commented 6 years ago

Hi @kalibera!

I've been having a hard time figuring out why I'm getting these warnings.

I tried to replicate it by running rchk locally, but then I couldn't reproduce these warnings.

In my Rcpp code, I am calling no_init() 8 times. Since the warnings are being generated for 1 matrix and 2 vectors, I assume it must have something to do with these three lines.

Do you know where these warnings are coming from, and how I could fix them?

Robrecht

kalibera commented 6 years ago

Hi Robrecht,

these warnings are probably false alarms (I am not 100% sure though as I do not understand enough of what is Rcpp doing internally). I cannot reproduce these warnings locally, either, and I think it is because the checks were run with an older version of Rcpp. A newer version includes an extra protection specifically to make rchk happy, and hence these warnings related to no_init do not appear. As many rchk reports about Rcpp code seem to be false alarms, I've decided to filter out all Rcpp related reports in the CRAN checks (of course, this means true positives would be missed too, and I've seen at least one in the past). The filtering, however, apparently did not always work. I've updated the filtering and now I am re-running all checks - new results should appear in several days and reports for princurve should disappear.

Tomas

rcannood commented 6 years ago

Thanks a lot for looking into this! :bowing_man: It's good to hear that it was a false positive.