Re: PR 7492
I converted a working Apache1.3 server to Apache2.0 on a SuSE Linux system with IPV6 support. All httpd.conf address references are in IPV4 format, all log/status/error messages show all addresses in IPV4 format. However, I had to use IPV6 address in the VHosts sections or the incoming requests would not match. VirtualHost :::10.168.xxx.yyy:8080 This certainly was not intuitive, and took several days of fussing to discover. ...don [EMAIL PROTECTED]
configtest
Trying to use configtest with a httpd.conf file that has a line like: VirtualHost 82 gives: ../apachectl: line 187: 2066 Sementation fault $HTTPD - t ...don [EMAIL PROTECTED]
setsockopt error
Apache2.0 If all of my vhosts are of the form: VirtualHost *:nn and if there is a Listen for a port (81) that does NOT appear in a VirtualHost directive, then I get: ... [crit] (88)Socket operation on non-socket: make_sock: for address [::]:81, setsockopt: (SO_REUSEADDR) ...don [EMAIL PROTECTED]
Re: configtest
Don Hughes [EMAIL PROTECTED] writes: Trying to use configtest with a httpd.conf file that has a line like: VirtualHost 82 gives: ../apachectl: line 187: 2066 Sementation fault $HTTPD - t cool! easy to reproduce, probably easy to fix... keep those cards and letters coming -- Jeff Trawick | [EMAIL PROTECTED] | PGP public key at web site: http://www.geocities.com/SiliconValley/Park/9289/ Born in Roswell... married an alien...
Re: related config directives
On Mon, Dec 24, 2001 at 09:40:03AM -0800, Ryan Bloom wrote: I thought of another reason to do this in the tree itself. The point of the config tree is to keep the current config in memory. If you are going to modify the config, you need to modify it in the tree. By doing to work on the tree, you get that for free. I'm confused (it may just be the eggnog :), but aren't we talking about two different things? There are fatal and non-fatal errors in the config. We want to give a meaningful message when we discover a fatal error, but we just want to issue a warning and try to adjust something sane for non-fatal errors. I can see how fidgeting with absurd settings by rearranging the order in the tree will make it easier to deal with some of the non-fatal errors, but doesn't it seem easiest to deal with the fatal errors all at once after we've got a stable config tree with all the defaults in place in the tree? -aaron
Re: cvs commit: httpd-2.0/server/mpm/worker worker.c
Ugh! I've been meaning to patch the prototype for apr_thread_exit for a long time (so that it takes an apr_status_t instead of a pointer), but it got lost in the shuffle. I'll see if I can fix this sometime this week so we don't have to have ackward rv stuff like this. -aaron On Sun, Dec 23, 2001 at 01:56:49PM -, [EMAIL PROTECTED] wrote: dreid 01/12/23 05:56:49 Modified:server/mpm/worker worker.c Log: This fixes a segfault that showed up on BeOS and may catch other systems. Revision ChangesPath 1.51 +4 -2 httpd-2.0/server/mpm/worker/worker.c Index: worker.c === RCS file: /home/cvs/httpd-2.0/server/mpm/worker/worker.c,v retrieving revision 1.50 retrieving revision 1.51 diff -u -r1.50 -r1.51 --- worker.c2001/12/19 14:49:22 1.50 +++ worker.c2001/12/23 13:56:49 1.51 @@ -790,7 +790,8 @@ worker_thread_count--; apr_thread_mutex_unlock(worker_thread_count_mutex); -apr_thread_exit(thd, APR_SUCCESS); +rv = APR_SUCCESS; +apr_thread_exit(thd, rv); return NULL; } @@ -879,7 +880,8 @@ * life_status is almost right, but it's in the worker's structure, and * the name could be clearer. gla */ -apr_thread_exit(thd, APR_SUCCESS); +rv = APR_SUCCESS; +apr_thread_exit(thd, rv); return NULL; }
Re: setsockopt error
On Tue, Dec 25, 2001 at 02:46:08PM -0500, Don Hughes wrote: Apache2.0 If all of my vhosts are of the form: VirtualHost *:nn and if there is a Listen for a port (81) that does NOT appear in a VirtualHost directive, then I get: ... [crit] (88)Socket operation on non-socket: make_sock: for address [::]:81, setsockopt: (SO_REUSEADDR) How are you starting Apache? bin/apachectl? -aaron
[PATCH] fix prep_walk_cache performance
This change may be controversial, so I'm soliciting
comments before committing...
The prep_walk_cache() function in server/request.c
is one of the slower parts of the core server. About
60% of its run time is due to the apr_pool_userdata_*
functions.
The attached patch eliminates the use of the apr_pool_userdata_*
functions by adding hooks for the various walk caches to the
core_request_config structure (stored in r-request_config).
--Brian
Index: include/http_core.h
===
RCS file: /home/cvs/httpd-2.0/include/http_core.h,v
retrieving revision 1.55
diff -u -r1.55 http_core.h
--- include/http_core.h 2001/12/13 14:50:36 1.55
+++ include/http_core.h 2001/12/26 06:58:26
@@ -330,10 +330,21 @@
/* Per-request configuration */
+typedef enum {
+AP_WALK_DIRECTORY,
+AP_WALK_LOCATION,
+AP_WALK_FILE,
+AP_NUM_WALK_CACHES
+} ap_walk_cache_type;
+
typedef struct {
/* bucket brigade used by getline for look-ahead and
* ap_get_client_block for holding left-over request body */
struct apr_bucket_brigade *bb;
+
+/* a place to hold per-request working data for
+ * ap_directory_walk, ap_location_walk, and ap_file_walk */
+void *walk_cache[AP_NUM_WALK_CACHES];
} core_request_config;
/* Per-directory configuration */
Index: server/core.c
===
RCS file: /home/cvs/httpd-2.0/server/core.c,v
retrieving revision 1.121
diff -u -r1.121 core.c
--- server/core.c 2001/12/18 13:48:52 1.121
+++ server/core.c 2001/12/26 06:58:28
@@ -3399,17 +3399,18 @@
static int core_create_req(request_rec *r)
{
+core_request_config *req_cfg;
+req_cfg = apr_palloc(r-pool, sizeof(core_request_config));
+memset(req_cfg, 0, sizeof(*req_cfg));
if (r-main) {
-ap_set_module_config(r-request_config, core_module,
- ap_get_module_config(r-main-request_config, core_module));
+core_request_config *main_req_cfg = (core_request_config *)
+ap_get_module_config(r-main-request_config, core_module);
+req_cfg-bb = main_req_cfg-bb;
}
else {
-core_request_config *req_cfg;
-
-req_cfg = apr_pcalloc(r-pool, sizeof(core_request_config));
req_cfg-bb = apr_brigade_create(r-pool);
-ap_set_module_config(r-request_config, core_module, req_cfg);
}
+ap_set_module_config(r-request_config, core_module, req_cfg);
ap_add_input_filter(NET_TIME, NULL, r, r-connection);
Index: server/request.c
===
RCS file: /home/cvs/httpd-2.0/server/request.c,v
retrieving revision 1.86
diff -u -r1.86 request.c
--- server/request.c2001/12/14 03:30:23 1.86
+++ server/request.c2001/12/26 06:58:29
@@ -296,9 +296,10 @@
apr_array_header_t *walked; /* The list of walk_walked_t results */
} walk_cache_t;
-static walk_cache_t *prep_walk_cache(const char *cache_name, request_rec *r)
+static walk_cache_t *prep_walk_cache(ap_walk_cache_type t, request_rec *r)
{
walk_cache_t *cache;
+core_request_config *my_req_cfg;
/* Find the most relevant, recent entry to work from. That would be
* this request (on the second call), or the parent request of a
@@ -306,25 +307,32 @@
* this _walk()er with a copy it is allowed to munge. If there is no
* parent or prior cached request, then create a new walk cache.
*/
-if ((apr_pool_userdata_get((void **)cache, cache_name, r-pool)
-!= APR_SUCCESS)
-|| !cache) {
-if ((r-main (apr_pool_userdata_get((void **)cache,
- cache_name,
- r-main-pool)
- == APR_SUCCESS) cache)
- || (r-prev (apr_pool_userdata_get((void **)cache,
- cache_name,
- r-prev-pool)
- == APR_SUCCESS) cache)) {
-cache = apr_pmemdup(r-pool, cache, sizeof(*cache));
+my_req_cfg = (core_request_config *)
+ap_get_module_config(r-request_config, core_module);
+
+if (!my_req_cfg || !(cache = my_req_cfg-walk_cache[t])) {
+core_request_config *req_cfg;
+if ((r-main
+ (req_cfg = (core_request_config *)
+ ap_get_module_config(r-main-request_config, core_module))
+ req_cfg-walk_cache[t]) ||
+(r-prev
+ (req_cfg = (core_request_config *)
+ ap_get_module_config(r-prev-request_config, core_module))
+ req_cfg-walk_cache[t])) {
+cache = apr_pmemdup(r-pool, req_cfg-walk_cache[t],
+sizeof(*cache));
cache-walked = apr_array_copy(r-pool, cache-walked);
}
else {
cache = apr_pcalloc(r-pool,
