[Rsiena-help] Possible error with multi-core estimation in rev. 119

[Guido Conaldi]

Dear Ruth,

RSienaTest keeps crashing when I try estimating my models:

*** glibc detected *** /usr/lib64/R/bin/exec/R: free(): invalid next
size (fast): 0x000000000506b9f0 ***

I can save the backtrace for later inspection if that may help.

However, I patched and build the tar ball as you suggested and
multi-core is working fine again.

Thank you very much and enjoy your holiday,
Regards,
Guido

On Tue, 2010-08-31 at 11:58 +0100, Ruth M. Ripley wrote: 
> Dear Guido,
> 
> As Tom says, I am on holiday. But I know what the bug is and it should not 
> exist in RSienaTest. Please try using RSienaTest from R-forge. It should 
> give the same answers for one core and not fail for more than one.
> 
> Alternatively, if you are able to unpack the tar ball and build the 
> package yourself, try making a one line change to R/simstatsc.r. After 
> line 1762, which is
> 
> nGroup <- ff$nGroup
> 
> add
> 
> types <- ff$types
> 
> and rebuild.
> 
> I will sort RSiena out when I get back to work.
> 
> Regards,
> 
> Ruth
> 
> On Tue, 31 Aug 2010, Guido Conaldi wrote:
> 
> > Dear Ruth,
> >
> > I updated to rev.119 and since then I experience an error if I try to
> > use more than one core (3 in this case, as you can see):
> >
> > Error in checkForRemoteErrors(lapply(cl, recvResult)) :
> > 3 nodes produced errors; first error: object 'types' not found
> >
> > With one core, estimation seems to be running fine.
> >
> > I hope you can help,
> > Regards
> > Guido
> >
> > _______________________________________________
> > Rsiena-help mailing list
> > Rsiena-help at lists.r-forge.r-project.org
> > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rsiena-help
> >