On Sun, 27 Jun 2004, Dinesh Nair wrote:

> been looking over the code of mini_httpd, specifically my changes to allow
> concurrent connection, especially with respect to the problem many of you
> have been facing with the apparent "hanging" of the webGUI.
> 
> there was a bug which ommitted to turn off an alarm, and the parent
> mini_httpd died when the alarm was triggered, thus killing the webGUI in
> the process. this only happens when you have hit 8 simultanoues
> connections to the webGUI, and you'd be blissfully unaware of it
> otherwise.

Did the "died" part include using a bad pointer?  If not, it wouldn't
explain the SIGSEGV.

> the problem manifested itself more in IE because of the way IE does not
> close a connection when it's done with it. thus IE users would have been
> more likely to hit the bug, though other browsers would have eventually
> hit it as well. testing on IE made it more apparent.

Safari keeps HTTP/1.1 sessions open as well.  I'd expect that to be
increasingly common.

					Fred Wright