Re: [4.3] Current critical bugs

Thanks. I appreciate the response. Joseph Zeev Suraski wrote:

At 21:14 15/10/2002, Joseph Tate wrote:

Throw me a bone then. What is the suggested way to offer php developers the opportunity to run code after the connection has been closed?

I think I mentioned this as well; IMHO, if we want to create the ability for users under Apache/UNIX to run stuff after the connection is closed, we should add a dedicated function for it, it shouldn't be overloaded to register_shutdown_function(). It can be apache_register_shutdown_function() or something like that, so that it would be clear it takes advantage of platform dependent functionality.

Even if it only works under Apache on Linux? Also if the 4.1.0 behavior is the "correct" behavior, why is the function still documented as the 4.0.x behavior?

Because those who document the code and those who develop it are usually not the same people :)

I'll admit that the company I work for is probably the only one using this function, but we depend on it. We do heavy image processing work, using image magick, called from clients using wireless networks, where it hurts us to keep connections open for long periods of time. If we can't fork some processing to the background, our users think our app is slow, and we lose sales. It's that simple. "Correct" behavior or not, we need the functionality. My patch, I hope, will be ready by the end of the week. Adding a parameter to the register_shutdown_function will not be possible, since the function was changed a while back to allow users to specify arguments to the function being registered. Instead a second function will be created: register_offline_function offering the 4.0.x behavior.

I have no objection to adding the Apache specific equivalent to 4.3. As a matter of fact, if it solves a problem - I'm quite for it. I hope this bone is meaty enough :) Zeev