[Rcpp-devel] Compiled/installed version mismatch

Xavier Robin xavier at cbs.dtu.dk
Sat Apr 5 23:20:21 CEST 2014


Dear Dirk,

Thanks for your answer.

On 05. 04. 14 19:15, Dirk Eddelbuettel wrote:
> We try to be very clear and attempted to broadly communicate the need for
> rebuilds when Rcpp 0.11.0 was released.
I was more or less aware of if - although I didn't quite realize the
full extent and implications of this information.

But my question was rather aimed at the end user who doesn't know that
Rcpp is being used (or even what it is in the first place for that
matter, no offense meant) and has little understanding of what is going
on when he installs a package.
Is there a way to force a reinstallation on his side? From the rest of
your answer, I read I don't have one. This is ok - I thought I had
missed some piece of information (I usually do) but it is apparently not
the case this time.

I remember a previous update of R that broke some packages without a
namespace. It would print an error message stating that it was necessary
to re-install the package. I don't know if this kind of things could be
enforced with Rcpp as well - if so it would possibly give you more
freedom to make breaking changes when needed. But I'm probably
overstepping here.

> That said, we made several dozen Rcpp releases since 2008. And all but one
> (or two?) upgrades were non-breaking. 
>
> In general, we try our best to provide stable, consistent and maintained
> interfaces.  Sometimes a change is worth it the disruption. We feel this was
> one of these times.  I hope you still trust us that such events will be rare.
I am sorry it wasn't clear enough in my previous email. There is no
doubt I will keep using Rcpp in the future, and I cannot thank you all
enough for making it available. Orders of magnitude speedup is certainly
well worth a glitch from time to time.

Best wishes,
Xavier


-- who is about to submit to the CRAN now and getting ready for the
usual spanking.



More information about the Rcpp-devel mailing list