Patch source: https://vcs.pcre.org/pcre/code/trunk/pcre_internal.h?r1=1649&r2=1688&sortby=date https://vcs.pcre.org/pcre/code/trunk/pcre_ucd.c?r1=1490&r2=1688&sortby=date --- a/pcre_internal.h 2016/05/21 13:34:44 1649 +++ b/pcre_internal.h 2017/02/24 17:30:30 1688 @@ -2772,6 +2772,9 @@ extern const pcre_uint16 PRIV(ucd_stage2)[]; extern const pcre_uint32 PRIV(ucp_gentype)[]; extern const pcre_uint32 PRIV(ucp_gbtable)[]; +#ifdef COMPILE_PCRE32 +extern const ucd_record PRIV(dummy_ucd_record)[]; +#endif #ifdef SUPPORT_JIT extern const int PRIV(ucp_typerange)[]; #endif @@ -2780,9 +2783,15 @@ /* UCD access macros */ #define UCD_BLOCK_SIZE 128 -#define GET_UCD(ch) (PRIV(ucd_records) + \ +#define REAL_GET_UCD(ch) (PRIV(ucd_records) + \ PRIV(ucd_stage2)[PRIV(ucd_stage1)[(int)(ch) / UCD_BLOCK_SIZE] * \ UCD_BLOCK_SIZE + (int)(ch) % UCD_BLOCK_SIZE]) + +#ifdef COMPILE_PCRE32 +#define GET_UCD(ch) ((ch > 0x10ffff)? PRIV(dummy_ucd_record) : REAL_GET_UCD(ch)) +#else +#define GET_UCD(ch) REAL_GET_UCD(ch) +#endif #define UCD_CHARTYPE(ch) GET_UCD(ch)->chartype #define UCD_SCRIPT(ch) GET_UCD(ch)->script --- a/pcre_ucd.c 2014/06/19 07:51:39 1490 +++ a/pcre_ucd.c 2017/02/24 17:30:30 1688 @@ -38,6 +38,20 @@ const pcre_uint32 PRIV(ucd_caseless_sets)[] = {0}; #else +/* If the 32-bit library is run in non-32-bit mode, character values +greater than 0x10ffff may be encountered. For these we set up a +special record. */ + +#ifdef COMPILE_PCRE32 +const ucd_record PRIV(dummy_ucd_record)[] = {{ + ucp_Common, /* script */ + ucp_Cn, /* type unassigned */ + ucp_gbOther, /* grapheme break property */ + 0, /* case set */ + 0, /* other case */ + }}; +#endif + /* When recompiling tables with a new Unicode version, please check the types in this structure definition from pcre_internal.h (the actual field names will be different):