In article <> you wrote:
> 15-Jan-99 18:35 you wrote:
>> We've already recognized that on some not-such-smart platforms Apache+mod_ssl
>> can dump core under the DSO situation. I've already traced it down and the
>> problem has to do with the fact that Apache _reloads_ the DSOs at startup-time
>> and that some dangling references to EAPI hooks exists this way. On platforms
>> like FreeBSD and Linux (where I develop mod_ssl) the modules always seem to be
>> reloaded to the same memory address, so I never recognized this problem. But
>> now Khimenko Victor's EAPI replacement patches gave me the essential hint that
>> the hooks are not unregistered correctly. This doesn't harm when the module
>> was reloaded to the same address, but causes core dumps when this isn't the
>> case.
> BTW my replacement patches are designed so that Apache will coredump without
> properly unregistered hooks even on "smart" platforms so there are ALL hooks
> (in mod_log_config, mod_proxy and mod_ssl itself) are correctly unregistered
> on module unload event. So mod_ssl with my patches should be safe even on
> non-such-smart platforms...
Be careful, I'm still not convinced this is really the case. Because inside
mod_log_config the "hooks" are not used for calling. The are used for
registering function calls. And the problem are not dangling function pointers
inside EAPI, the problem are those dangling function pointers inside
mod_log_config's own private structures as far as I see the problem.
Ralf S. Engelschall
[EMAIL PROTECTED]
www.engelschall.com
______________________________________________________________________
Apache Interface to SSLeay (mod_ssl) www.engelschall.com/sw/mod_ssl/
Official Support Mailing List [EMAIL PROTECTED]
Automated List Manager [EMAIL PROTECTED]
