R devel - May 2010 - dyn.load() strange behavior

Hello,

I am observing the following strange behavior when I try to
load a shared library using dyn.load() under Linux...

The library foo.so refers to a symbol in another package bar
(with shared library bar.so), I get:

dyn.load('foo.so') # error message because I forgot to load bar
library(bar) # so I fix it
dyn.load('foo.so') # works this time and everything is fine

But, if restart my R session and do it "right" the second
time there is a surprise:

library(bar)
dyn.load('foo.so') # fails
library(bar) # just to be sure?
dyn.load('foo.so) # fails again
dyn.load('/usr/local/lib64/library/bar/lib/libbar.so') # just to be
really
sure?
dyn.load('foo.so')

In other words, it is possible to load foo.so, but only
after a failed attempt. If I remember to load the required
package bar I cannot load the depenent library foo.so.

Any ideas?

Thanks,
Dominick

	[[alternative HTML version deleted]]

- Does bar really dyn.load its shared library on load?
- Have you told package foo (in the DESCRIPTION) it is LinkingTo: bar ?

Uwe Ligges



On 21.05.2010 22:35, Dominick Samperi wrote:
> Hello,
>
> I am observing the following strange behavior when I try to
> load a shared library using dyn.load() under Linux...
>
> The library foo.so refers to a symbol in another package bar
> (with shared library bar.so), I get:
>
> dyn.load('foo.so') # error message because I forgot to load bar
> library(bar) # so I fix it
> dyn.load('foo.so') # works this time and everything is fine
>
> But, if restart my R session and do it "right" the second
> time there is a surprise:
>
> library(bar)
> dyn.load('foo.so') # fails
> library(bar) # just to be sure?
> dyn.load('foo.so) # fails again
> dyn.load('/usr/local/lib64/library/bar/lib/libbar.so') # just to be
really
> sure?
> dyn.load('foo.so')
>
> In other words, it is possible to load foo.so, but only
> after a failed attempt. If I remember to load the required
> package bar I cannot load the depenent library foo.so.
>
> Any ideas?
>
> Thanks,
> Dominick
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel