[Rcpp-devel] Large size of libs directory on some systems

[Dean Bodenham]

Dear Dirk,

Thank you very much for your detailed replies!

The Ubuntu system does indeed seem to be using the -g flag. I did try a
local compile/installation with the ~/.R/Makevars file you suggested, but
couldn't quite get it to work - I get warnings "linking not done" for the
.o files, and then an error that the "ffstream.so" file is not found.

But, even if it did work locally, then, as you say, relying on a Makevars
file could lead to issues later when trying to get the package onto CRAN (I
have encountered difficulties with this approach previously).

Perhaps I will leave the package as it is for the moment, but it is very
good to know the reason why the libs directory is (sometimes) so large.

Thank you again for your help, and I will watch out for any future
announcements about this issue.

Best regards,
Dean

On Fri, Nov 25, 2016 at 2:38 PM, Dirk Eddelbuettel <edd at debian.org> wrote:

>
> On 25 November 2016 at 07:23, Dirk Eddelbuettel wrote:
> | Dan Dillon has some experimental code to provide a stripping plugin.
> Until
> | then, and just for local tests, do something like this in src/Makevars:
> |
> |    PKG_LIBS= -Wl,-S
> |
> | This asks the linker to use -S to strip debugging symbols (but not
> general
> | symbols, which R needs).  On my Ubuntu 16.04 box I get a change from
> |
> |    9.0 mb    (default)
> |
> | to
> |
> |    734 kb    (with stripping)
> |
> | for twelve-fold reduction in size _on the same system_.
> |
> | If/when we have something more general I am sure it will be announced
> here.
> | If you want to play with the plugin hit me and/or Dan (CC'ed).
>
> PS The reason we have not talked about this / used this more is that we
> think
> it raise the "not portable" flags at CRAN so we want to (ideally) do it
> with
> some code that behaves correctly.
>
> For local use in ~/.R/Makevars it is perfectly fine.
>
> Dirk
>
> --
> http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rcpp-devel/attachments/20161126/30fb0e4d/attachment.html>