httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Roy T. Fielding" <>
Subject Re: solaris HUP blues
Date Fri, 25 Jul 1997 04:10:56 GMT
>Ok so what if SIGHUP didn't close the sockets just on SOLARIS2 < 260.  We
>can even emit a log message "retaining sockets from pre-HUP to work around
>OS problems".  If the server gets into that accept problem then the user
>is going to have to do a full restart. 

Nope, SIGHUPs are routinely done when nobody is around, and the non-routine
ones are often to fix problems that require the socket close.  Like I said,
a normal server doesn't need any of this hackery, so why screw them?
I'd rather emit a message during SOLARIS2 < 260 Configure that says
"OS is limited to XX file descriptors and has too many bugs to
automatically work around.  Since this may impact sites with many virtual
hosts and separate logfiles, please see http://blah/blah for manual
workarounds if you encounter descriptor limitations."

And the latter would say "You can define -DNO_SLACK, but ...

In any case, I'd personally prefer it if the slack code was disabled
if the number of listeners is less than 5.


View raw message