/* NB: breaking out of this loop jumps to log_and_exit() */
out_cnt = 0;
+ pfd[FROM_CGI].fd = fromCgi_rd;
+ pfd[FROM_CGI].events = POLLIN;
+ pfd[TO_CGI].fd = toCgi_wr;
while (1) {
/* Note: even pfd[0].events == 0 won't prevent
* revents == POLLHUP|POLLERR reports from closed stdin.
- * This works: */
+ * Setting fd to -1 works: */
pfd[0].fd = -1;
+ pfd[0].events = POLLIN;
+ pfd[0].revents = 0; /* probably not needed, paranoia */
- pfd[FROM_CGI].fd = fromCgi_rd;
- pfd[FROM_CGI].events = POLLIN;
+ /* We always poll this fd, thus kernel always sets revents: */
+ /*pfd[FROM_CGI].events = POLLIN; - moved out of loop */
+ /*pfd[FROM_CGI].revents = 0; - not needed */
- pfd[TO_CGI].fd = toCgi_wr;
+ /* gcc-4.8.0 still doesnt fill two shorts with one insn :( */
+ /* http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47059 */
+ /* hopefully one day it will... */
pfd[TO_CGI].events = POLLOUT;
+ pfd[TO_CGI].revents = 0; /* needed! */
if (toCgi_wr && hdr_cnt <= 0) {
if (post_len > 0) {
/* Expect more POST data from network */
pfd[0].fd = 0;
- pfd[0].events = POLLIN;
} else {
/* post_len <= 0 && hdr_cnt <= 0:
* no more POST data to CGI,