rbb 99/03/03 08:02:46
Modified:pthreads/src/main fdqueue.c http_main.c
Log:
Performance improvement, so we don't overfill our fdqueue array. I am just
moving
where we update how much blanks space we have in the array, so we don't get
the
blank space back until the worker t
rbb 99/02/23 11:11:23
Modified:pthreads/src/include fdqueue.h http_conf_globals.h
pthreads/src/main fdqueue.c http_core.c http_main.c
Log:
A hopefully final commit for the fdqueue logic. This removes the bug
where our queue head and tail where getting out of joi
rbb 99/02/22 12:05:47
Modified:pthreads/src/include fdqueue.h
pthreads/src/main fdqueue.c http_main.c
Log:
Removing Manoj's last fix for the accept thread. It puts the check for queue-
fullness in the wrong spot, and removes necessary logic to keep us from
ove
manoj 99/02/21 23:49:34
Modified:pthreads/src/include fdqueue.h
pthreads/src/main fdqueue.c http_main.c
Log:
Change the fdqueue's handling of blocking when there are no idle
threads. Advantages:
- We used to add to the queue before checking if it was full. Fix
rbb 99/02/20 19:20:05
Modified:pthreads/src/main fdqueue.c
Log:
We actually want to signal a thread everytime we add something to the
fdqueue, not just when the queue was empty before we added it. This
was resulting in a bug where only one thread was ever actually working on
rbb 99/02/19 12:33:25
Modified:pthreads/src/include fdqueue.h
pthreads/src/main fdqueue.c http_main.c
Log:
Fixes annoying fdqueue logic bug. It used to be that if long lived
connections
used up all of your threads, any new connections to the same process would
manoj 99/02/15 13:58:59
Modified:pthreads/src/main fdqueue.c
Log:
Fix lockup bug in fdqueue.c. Because of the recent change to make
queue_pop non-blocking, waits were happening without the corresponding
signal.
Revision ChangesPath
1.6 +3 -1 apache-apr/pt
rbb 99/02/12 12:37:00
Modified:pthreads/src/main fdqueue.c http_main.c
Log:
Bug fixes for graceful restart. Now, we return -1 for empty queue. This was
needed for the graceful restart which is also included in this patch. Most
of the changes from the last two patches were
rbb 99/02/11 14:15:56
Modified:pthreads/src/main fdqueue.c http_main.c
Log:
Another pass at logic for signal handling. SIGTERM works properly. SIGHUP
causes a zombie, and makes the socket unusable. Graceful restarts are not
close to working, because of signals issues with
manoj 99/02/09 14:04:16
Modified:pthreads/src/main fdqueue.c
Log:
Optimization: don't pthread_cond_signal every time something is added to
the queue, but only when there is a state change (empty -> not empty, full
-> not full).
Revision ChangesPath
1.2 +8 -3
10 matches
Mail list logo
