On Fri, Jul 8, 2022 at 7:08 PM Ivan Zhakov <[email protected]> wrote:
>
> On Tue, 25 Jan 2022 at 20:34, <[email protected]> wrote:
> >
> > - child_handles[i] = CreateThread(NULL, ap_thread_stacksize,
> > - worker_main, (void *) i,
> > - stack_res_flag, &tid);
> > - if (child_handles[i] == 0) {
> > - ap_log_error(APLOG_MARK, APLOG_CRIT, apr_get_os_error(),
> > - ap_server_conf, APLOGNO(00355)
> > - "Child: CreateThread failed. Unable to "
> > + workers[i].num = i;
> > + workers[i].handle = CreateEvent(NULL, TRUE, FALSE, NULL);
> > + if (!workers[i].handle) {
> > + rv = apr_get_os_error();
> > + }
> > + else {
> > + apr_threadattr_t *thread_attr = NULL;
> > + apr_threadattr_create(&thread_attr, pchild);
> > + if (ap_thread_stacksize != 0) {
> > + apr_threadattr_stacksize_set(thread_attr,
> > + ap_thread_stacksize);
> > + }
> > + rv = ap_thread_create(&workers[i].thd, thread_attr,
> > + worker_thread, &workers[i], pchild);
>
> This is performance regression: before this change stack memory was
> 'reserved'. See stack_res_flag in CreateThread call. Now stack memory
> is committed.
So STACK_SIZE_PARAM_IS_A_RESERVATION needs to go to apr_thread_create()?
> >
> > while (threads_created)
> > {
> > - HANDLE handle = child_handles[threads_created - 1];
> > + struct worker_info *info = workers[threads_created - 1];
>
> This code doesn't compile:
> [[[
> server\mpm\winnt\child.c(1210,1): error C2440: 'initializing': cannot
> convert from 'worker_info' to 'worker_info *'
> ]]]
r1902636, hopefully.
Regards;
Yann.
