httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@znep.com>
Subject Re: updated status for 1.2b7
Date Mon, 17 Feb 1997 08:23:07 GMT
On Sun, 16 Feb 1997, Roy T. Fielding wrote:

> >  * server can hang on HUP
> >	Status: Marc posted patch, -1 Chuck because of sleep(2)... but I 
> >	don't like the alternatives.
> 
> I'd like to know what affect this has on a busy server that HUPs nightly.
> Perhaps using an exponential backoff with a select-based timeout in
> microseconds would be better.

On my busy server my patch means I _can_ HUP it without making it die all
the time. 

I think I will just use a blocking waitpid() with an alarm() acting as a
timeout if the child doesn't exit.  Shucks, guess I will need a longjmp in
there since waitpid() can be restartable.  I don't think I want to mess
with trying to set everything up so that for all systems the waitpid()
won't be restartable.

> >  * Solaris "accept: Too many levels of remote in path" [marc]
> 
> I've never seen this on my test machine, but I don't get external request
> on it either.

This server wasn't getting anything other than local test traffic either;
the odd thing is that I am sure it would be a very common report if people
did have trouble with it, yet I only have a faint recollection of hearing
about it once or twice in the past.  Unless it was introduced very
recently... guess we will find out.

> I think that the only two must-do fixes left are the byteranges fix
> and the tiny-change patch I posted.  All the rest are less important
> than getting a new beta out to the testers.

I still would like to look at the lingering_close changes Dean suggested.
I'm not convinced they can be implemented cleanly, but if they can they
could be a big performance win.  I'll know more tomorrow.


Mime
View raw message