[Dovecot] Why p_strdup and other string functions uses loops instead strlen? (dovecot 2.0.rc4)

[Len7hir]

I did Dovecot profiling on huge e-mail system, and p_strdup was very 
high on list.

I do minor change:
p_strdup:
/*
        for (len = 0; (str)[len] != '\0'; )
                len++;
        len++;
*/
        len = strlen(str) + 1;
p_strndup:
/*
        len = 0;
        while (len < max_chars && ((const char *) str)[len] != '\0')
                len++;
*/
        len = strnlen(str, max_chars);

And after changes strdup drop down on the profile list.


This is output from oprofile:

samples  %        symbol name
28000    10.9764  p_strdup
8507      3.3349  safe_memset
7857      3.0801  .plt
7627      2.9899  buffer_write
6871      2.6935  parse_body_add_block

And after (about four times smaller samples, but shows everything):
samples  %        symbol name
9595     26.2625  parse_next_body_to_boundary
8247     22.5729  parse_body_add_block
772       2.1130  .plt
672       1.8393  buffer_write
658       1.8010  __i686.get_pc_thunk.bx
614       1.6806  safe_memset
586       1.6039  pool_alloconly_malloc
559       1.5300  p_strdup
533       1.4589  hash_table_insert_node


I wonder why You use loops instead strlen, which is optimalised on every 
platforms.

Redgards,
Len7hir