Autor: Philip Hazel Data: A: Ralf Junker CC: pcre-dev Assumpte: Re: [pcre-dev] Internal errors and crashes with quantified
subroutines
On Tue, 22 Nov 2011, Ralf Junker wrote:
> To accommodate changes in LINK_SIZE, would it be reasonable to redefine
> COMPILE_WORK_SIZE like this:
>
> #define COMPILE_WORK_SIZE (4096 / 2 * LINK_SIZE)
That would only raise the maximum number of forward references from
around 1500 in the LINK_SIZE=3 case to around 2000, the same as the
LINK_SIZE=2 case. I do not think that really helps (ND has quoted
examples with repeat counts of 35000.)
