[Rd] segfault on large number of open brackets (PR#7859)

[Prof Brian Ripley]

On Thu, 12 May 2005 ripley at stats.ox.ac.uk wrote:

> I suspect that is coincidence, as I can see no source change that would
> account for it.
>
> 2.0.1 and R-devel work for me and 2.1.0 segfaults with
>
> 999         KeepSource = *LOGICAL(GetOption(install("keep.source"), R_NilValue));
> (gdb) bt
> #0  0x080ca229 in ParseInit () at gram.y:999
> #1  0x080cb305 in token () at gram.y:1530
>
> I am not sure that is right, but I am pretty sure that
>
> static char	contextstack[50], *contextp;
>
> just above is an unchecked nesting limit, as in
>
> static void IfPush(void)
> {
>     if (*contextp==LBRACE ||
> 	*contextp=='['    ||
> 	*contextp=='('    ||
> 	*contextp == 'i')
> 	    *++contextp = 'i';
> }
>
> However, although I thought this was the problem, it seems only to apply
> to nested if()s.  So I do not currently have a solution.

I was almost right, just a different instance of *++contextp.

Fixed in R-patched, to give

> (((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
Error: contextstack overflow


-- 
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