Paolo, > Bruno, ... Can you shed light on what __collseq_table_lookup is supposed > to mean?
It is a runtime lookup function into a table that maps Unicode characters to uint32_t values. For a 'char' value, the most efficient way to implement a mapping from 'char' to uint32_t is through an array: uint32_t[UCHAR_MAX+1]. For a 'wchar_t' value whose width is up to 21 bits, the data structure we use in glibc (and also in gnulib / libunistring) is a 3-level lookup table. See the file locale/programs/3level.h for details. In regcomp.c and regexec.c the _NL_COLLATE_COLLSEQWC field of the LC_COLLATE part of the locale is encoded in this way. In glibc/locale/programs/ld-collate.c this field is being constructed from a table called 'collate->wcseqorder'. The role of this table is to be used in regular expression matching and wildcard matching. The table is derived from (but does not represent the entire information from) the LC_COLLATE portion of the locale input file. Bruno
