ID: 27412
Updated by: [EMAIL PROTECTED]
Reported By: chris at seismo dot usbr dot gov dot spamfree
-Status: Feedback
+Status: No Feedback
Bug Type: iPlanet related
Operating System: IRIX 6.5
PHP Version: 4CVS, 5CVS (2004-02-27)
New Comment:
No feedback was provided for this bug for over a week, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".
Previous Comments:
------------------------------------------------------------------------
[2004-07-15 17:12:20] [EMAIL PROTECTED]
Is it a solution to use a newer version of the SunONE servers? Because
since version 6 they use on all platforms pthreads.
------------------------------------------------------------------------
[2004-03-28 13:40:31] chris at seismo dot usbr dot gov dot spamfree
Further testing indicates that Netscape/iPlanet Enterprise Server uses
the sproc/sprocsp model of threading on IRIX, at least for versions 3
and 4. According to the sproc man page "The sproc model of threading is
incompatible with POSIX threads."
A problem with sproc threads is that they can quickly use up all
available process memory unless the per-process stack size is kept
*small* enough. With PHP NSAPI, this will result in a segment violation
(SIGSEGV), with a "not enough memory to lock stack" syslog error. So
with PHP, one may need to limit the per-process stacksize when starting
ns-httpd. For example:
(limit stacksize 8m && $NSES_PATH/start)
------------------------------------------------------------------------
[2004-03-09 05:25:58] [EMAIL PROTECTED]
Can you send me the patch as Unified Diff to my email?
------------------------------------------------------------------------
[2004-03-08 20:11:44] chris at seismo dot usbr dot gov dot spamfree
TSRM.c needed a few more minor changes to have complete NSAPI support.
Without these, you eventually get SIGSEGV.
The point of these changes is to permit choice of using NSAPI thread
API, thus avoiding dependency on native threads (and resulting problems
on some platforms).
Here are the complete context diffs for php-4.3.5RC3:
*** 91,96 ****
--- 91,98 ----
#if defined(PTHREADS)
/* Thread local storage */
static pthread_key_t tls_key;
+ #elif defined(NSAPI)
+ static int tls_key;
#elif defined(TSRM_ST)
static int tls_key;
#elif defined(TSRM_WIN32)
***************
*** 106,111 ****
--- 108,115 ----
pth_init();
#elif defined(PTHREADS)
pthread_key_create( &tls_key, 0 );
+ #elif defined(NSAPI)
+ tls_key = systhread_newkey();
#elif defined(TSRM_ST)
st_init();
st_key_create(&tls_key, 0);
***************
*** 186,191 ****
--- 190,197 ----
#elif defined(PTHREADS)
pthread_setspecific(tls_key, 0);
pthread_key_delete(tls_key);
+ #elif defined(NSAPI)
+ /* systhread_setdata(tls_key, 0); /* as bogus as
pthread_setspecific(key,0) */
#elif defined(TSRM_WIN32)
TlsFree(tls_key);
#endif
***************
*** 261,266 ****
--- 267,274 ----
#if defined(PTHREADS)
/* Set thread local storage to this new thread resources
structure */
pthread_setspecific(tls_key, (void *) *thread_resources_ptr);
+ #elif defined(NSAPI)
+ systhread_setdata(tls_key, (void *) *thread_resources_ptr);
#elif defined(TSRM_ST)
st_thread_setspecific(tls_key, (void *)
*thread_resources_ptr);
#elif defined(TSRM_WIN32)
***************
*** 302,307 ****
--- 310,317 ----
* and our hashtable lookup.
*/
thread_resources = pthread_getspecific(tls_key);
+ #elif defined(NSAPI)
+ thread_resources = systhread_getdata(tls_key);
#elif defined(TSRM_ST)
thread_resources = st_thread_getspecific(tls_key);
#elif defined(TSRM_WIN32)
***************
*** 390,395 ****
--- 400,407 ----
}
#if defined(PTHREADS)
pthread_setspecific(tls_key, 0);
+ #elif defined(NSAPI)
+ /* systhread_setdata(tls_key, 0); /* as bogus
as pthread_setspecific(key,0) */
#elif defined(TSRM_WIN32)
TlsSetValue(tls_key, 0);
#endif
***************
*** 524,530 ****
#elif defined(PTHREADS)
return pthread_mutex_lock(mutexp);
#elif defined(NSAPI)
! return crit_enter(mutexp);
#elif defined(PI3WEB)
return PISync_lock(mutexp);
#elif defined(TSRM_ST)
--- 536,543 ----
#elif defined(PTHREADS)
return pthread_mutex_lock(mutexp);
#elif defined(NSAPI)
! crit_enter(mutexp); /* returns void */
! return 0;
#elif defined(PI3WEB)
return PISync_lock(mutexp);
#elif defined(TSRM_ST)
***************
*** 551,557 ****
#elif defined(PTHREADS)
return pthread_mutex_unlock(mutexp);
#elif defined(NSAPI)
! return crit_exit(mutexp);
#elif defined(PI3WEB)
return PISync_unlock(mutexp);
#elif defined(TSRM_ST)
--- 564,571 ----
#elif defined(PTHREADS)
return pthread_mutex_unlock(mutexp);
#elif defined(NSAPI)
! crit_exit(mutexp); /* returns void */
! return 0;
#elif defined(PI3WEB)
return PISync_unlock(mutexp);
#elif defined(TSRM_ST)
For completeness, here are the context diffs for TSRM.h:
*** 45,50 ****
--- 45,52 ----
# include <pth.h>
#elif defined(PTHREADS)
# include <pthread.h>
+ #elif defined(NSAPI)
+ # include <nsapi.h>
#elif defined(TSRM_ST)
# include <st.h>
#elif defined(BETHREADS)
As discribed above, configure still needs hacking to turn off pthreads
and get the proper defines. If I get a chance, I'll make the m4 changes
to do that cleanly, and post those as well.
------------------------------------------------------------------------
[2004-02-27 19:09:07] chris at seismo dot usbr dot gov dot spamfree
Thanks for taking a look. Given Netscape's definitive statement that
NSAPI does not support pthreads, it may be more luck than anything else
if it appears to be working for some. More likely it's implementation
dependent, as indicated by the bug reports where people just gave up on
PHP - even with Solaris. The biggest risk could be not knowing what
other problems pthreads are causing.
A more robust approach might be to always use NES/iPlanet's thread API
(NSAPI) - no matter what version or platform. Since NSAPI seems to
already include a common thread API, does PHP really need to worry with
those details?
Both NSAPI and XP_UNIX should be defined in CFLAGS because TSRM.h
includes nsapi.h. Just defining XP_UNIX or NSAPI in nsapi.c won't work.
------------------------------------------------------------------------
The remainder of the comments for this report are too long. To view
the rest of the comments, please view the bug report online at
http://bugs.php.net/27412
--
Edit this bug report at http://bugs.php.net/?id=27412&edit=1
