[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

No Subject



Re: server problem

Ok. about races:

The log_*() routines simply write to a file using a (hopefully) atomic
single write. Even if there would be a race it would cause some sort of
error in the calls. There are no errors in the calls and I check each and
every return (excepting fprintf(), but I do check errors on logfp from time
to time).

Remeber that even if there was an error in the log writes and the child
would run then the socket would be serviced. What happens is, there is
simply no return (0) from fork. 

Remeber that the code works perfectly on a SUN machine and that httpd (NCSA)
also displays the problem when run stand alone on the Linux machine.

There are virtually no mistakes in the code (paranoia code used to capture
almost any kind of error). It also works right, excepting for the @#@#$
sockets lingering in the table.

What I was asking was, does anyone have the same problem ? If someone is
still running a 1.2.13 kernel somewhere please try to start httpd (NCSA) on
an idle port and telnet to it 3-4 times (and complete a session). 
Then do netstat -a | less and tell me if you have any sockets in TIME_WAIT
sticking around (there should be one such socket after two trials according
to the symptoms of my problem).

Also I need to buy a 2.0 kernel cdrom. Who sells in the TA area, how much, I
need that like, say, yesterday. I want full sources, no need for
installation or precompiled kernels but must have working gcc and
development libs (guess why).

If someone has the Yggdrasil 6-er pro set with 2.0 kernels I want that. 

TIA, 
	Peter


Follow-Ups: