Am Montag, den 05.03.2007, 13:42 -0600 schrieb William A. Rowe, Jr.:
> Just register one of two free functions as cleanups of the pool, based
> on the origin of the data.
How could that help me? I think I don't get the idea.
I think I should restate my problem. I can do something like
-------------------------------------------------------------
static apr_pool_t *volatile p_cur = NULL;
static void *xml2_alloc(size_t sz)
{
void *mem = apr_palloc(p_cur, sz + sizeof(size_t));
size_t *psz = mem;
*psz = sz;
return (char *) mem + sizeof(size_t);
}
-------------------------------------------------------------
There is the obvious ugliness of the current pool, which is easily
possible with prefork, but difficult with others.
There is also the problem that usage of pool memory in a filter is a de
facto leak when large documents are encountered.
I would prefer to do something like
const unsigned int mgc = 0x88DEFEA7;
static void *xml2_alloc(size_t sz)
{
void *mem = malloc(sz + sizeof(int));
const unsigned int *mark = mem;
*mark = mgc;
return (char *) mem + sizeof(int);
}
static void xml2_free(void *emem)
{
void *mem = (char *)emem - sizeof(int);
const unsigned int *mark = mem;
if (*mark == mgc) {
free(mem);
}
}
However I don't know if I can safely do that with apache memory (both,
bucket alloc and pool) and if there is value for mgc I can use.
Sincerely,
Joachim
