Don't have heavy load yet myself, but afraid I will have same problem on one machine. Here is my thinking from experience with other code:
Heavy load and Win32 indicates the problem has to do with concurrency by threading. Reasoning: Easy load doesn't cause much concurrency. And under Linux, as far as I am reading in documentation, most installations run with a separate process for each connection. So what is unique to heavy load under Win32 is concurrency by threading instead of by processes. Hope that reminder saves someone time in figuring this out. Maybe I'll get to it later. Also would be interested in hints about how to debug modules under Win32. - Leo At 09:30 AM 1/17/02 -0600, you wrote: >I could have written your email word for word! I've been trying to keep >Apache/mod_ssl running on WIN32 since early December, and still haven't figured out >what's wrong. Josef's patch seemed to work for him, but even after applying it, my >SSL-enabled Apache only runs a few minutes under heavy load before crashing. > >As far as getting a debug version built, I've had lots of trouble with that as well. >It seems to build just fine in debug mode in Visual C++, but when I attempt to debug >the EXE after a crash, I get no source code to trace. > >Is there anyone out there successfully running an SSL-enabled version of Apache on >WIN32? If so, did you have to do anything special? And either way, does someone >have a write-up explaining how to put together a debug build? > >Thanks. > >Jay > >-----Original Message----- >From: Enrico Demarin [mailto:[EMAIL PROTECTED]] >Sent: Thursday, January 17, 2002 12:19 PM >To: [EMAIL PROTECTED] >Subject: mod-ssl on win32 (mod_ssl bug 569, apache crash) > >I know it's not supported but... seems like something is still not good in mod_ssl , >as reported by Josef Goebel, >who posted a patch, there is something odd in mod_ssl for win32. I applied his patch >to ap_ctx.c , that cures >the crash in most of the cases >in ap_ctx_get() adds: >if ((ctx==NULL) || (key==NULL)) >{ >ap_log_assert("ap_ctx_get_bug: ctx or key are NULL!", >__FILE__, __LINE__); >return NULL; >} >and i also added the same check in ap_ctx_set. The problem is that sometimes mod_ssl >calls such functions with ctx set to null and key set to null, and crashes apache. I >have had other crashes which i haven't been able to debug ( how do i properly build a >debug version of apache-ssl on win32 ? when the visual c debug window pops up after a >crash all i can see is assembly code ). As reported by Josef the "null" calls seem to >be generated by ssl_io_suck_read() in ssl_engine_io.c, >but why does it happen ? is there some memory corruption going inside mod_ssl that >causes the structures to be set to null ? Apache with aforementioned patch is >"almost" stable but... i have the sensation that there is a problem somewhere else. >- Enrico > >______________________________________________________________________ >Apache Interface to OpenSSL (mod_ssl) www.modssl.org >User Support Mailing List [EMAIL PROTECTED] >Automated List Manager [EMAIL PROTECTED] ______________________________________________________________________ Apache Interface to OpenSSL (mod_ssl) www.modssl.org User Support Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
