You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by "William A. Rowe, Jr." <wr...@lnd.com> on 2000/07/27 02:11:19 UTC

server_conf and the Win32 MPM

Two questions,

Win32 had a function to read the httpd.conf file for the PidFile
directive.  If we pass server_conf through ap_run_pre_config, is
there a function to read one root level config directive from
the server_conf structure?  Would anyone object to server_conf
becoming an argument of ap_run_pre_config?

The reason is simple, we don't care that there are errors, when
the admin types apache -k stop, we want to die, no questions asked.


Secondly... I'm cleaning up my mess, and the Apache 2.0 code now
lives and breathes the default service config ("Apache").  If we
provide no argument, it assumes we are asking to invoke the service
"Apache" and all it's ConfigOpts.  Without an installed, default
service we get errors.  I'm thinking of eliminating this.

The question to the group, is 'everything' a service under Win32,
and if not, when is the Apache command a service, and when isn't it.
Apache -k [install|uninstall(remove)|start|stop(shutdown)|restart]
should be obvious, but what about no -k argument?  Do we approve
of this behavior or not?  A related issue, should Apache -k stop 
stop either the "Apache" service if it's started and a command-line
apache invocation?  Or is it time to stop worring about stoping and
restarting console apache (which is already visible, and shouldn't
need new hacks to hide it for Win9x?)

The risk is that a user typing Apache expects the same tests as
the user typing Apache -k start ... if the system doesn't invoke
the installed service's options (-D, -c, -C etc...) then they
can't tell what is working and what is broken.

Off to find food.

Bill



RE: server_conf and the Win32 MPM

Posted by "William A. Rowe, Jr." <wr...@lnd.com>.
> From: William A. Rowe, Jr. [mailto:wrowe@lnd.com]
> Sent: Wednesday, July 26, 2000 7:11 PM
> 
> Secondly... I'm cleaning up my mess, and the Apache 2.0 code now
> lives and breathes the default service config ("Apache").  If we
> provide no argument, it assumes we are asking to invoke the service
> "Apache" and all it's ConfigOpts.  Without an installed, default
> service we get errors.  I'm thinking of eliminating this.
> 
> [snip]  Or is it time to stop worring about stoping and
> restarting console apache (which is already visible, and shouldn't
> need new hacks to hide it for Win9x?)

Or... Ctrl+break and Ctrl+C are different signals... we could always
document (and splat to the console) for the user to Ctrl+Break to
halt the server, or Ctrl+C to gracefully restart :?  Opinions?