[Rd] Large number of linked environments makes save() crash R

[Prof Brian Ripley]

I think this is a C stack overflow problem.  Probably needs a test added 
internally to save().

I could not reproduce this on Linux: I got a C stack overflow that was 
caught.


On Wed, 26 Apr 2006, Henrik Bengtsson (max 7Mb) wrote:

> Hi,
>
> first, this not a problem originating from me, but it was observed
> trying to do similar things with my R.oo package, so I troubleshooted
> it and found the following.
>
> On R v2.3.0 (patched), the following will crash R on WinXP sooner or later:
>
> troubleLimit <- 20435;
>
> linkedList <- list(head=new.env());
>
> currEnv <- linkedList$head;
> for (kk in 1:(2*troubleLimit)) {
>  if (kk %% 100 == 0)
>    print(kk);
>  nextEnv <- new.env();
>  assign("nxt", nextEnv, envir=currEnv);
>  currEnv <- nextEnv;
>  if (kk > troubleLimit) {
>    cat("Saving linked list of length ", kk, "...", sep="");
>    save(linkedList, file="linkedList.RData");
>    cat("ok\n");
>  }
> }
>
> This gives:
> ...
> [1] 20300
> [1] 20400
> Saving linked list of length 20436...ok
> Saving linked list of length 20437...ok
> Saving linked list of length 20438...
>
> and then R crashes/terminates.  On my WinXP machine it always happens
> at the same number of environments.
>
> I tried the same on R v2.2.0 (sic!) on Unix and there it gave a
> segfault a bit after 10000 environments.  ...and, the purpose of the
> example is not to implement linked lists, it to show that you can
> torture R to crash.
>
> Cheers
>
> Henrik
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595