Re: Line 751 in http_mime_db.c

[Brian Behlendorf <br...@organic.com>]

Context - Rob and I just tracked down the last fd leak, meaning the 
problems we were having (at least on BSDI) with too many open fd's is fixed.
Other issues remain until 0.7.x is to be made public:

On Wed, 21 Jun 1995, Rob Hartill wrote:
> > Yup!  That did it.  yay!  I'll go install this on www.links.net now, to 
> > do some serious load testing.... I'd say this is finally ready for public 
> > consumption.
> 
> Almost.   inetd operation has so far been ignored. I think it
> can be handled by the standalone code, by changing the 
> MaxRequestsPerChild to 1, and have the parent exit on a SIGCHLD.. or
> maybe straight after it spawns the child.
> 
> If so, all the "if standalone" checks and extra routines in httpd.c
> for inetd can be dropped into the bit-bucket where they belong.

Sounds good to me.

> I'd also like to see a system to handle .html -> .shtml mapping
> to bring an end to XBITHACK.

Okay - is the mechanism we discussed (if file.html isn't found, file is 
looked for, and if multiviews is turned on then file.shtml will be 
returned) workable/acceptible?

> There were some earlier problem reports with Linux and Solaris. Do the
> problems still exist ?

Dunno - when 0.7.3g is uploaded, maybe other people can download and try 
it out...

> I'm sure there are also some fixes and patches in the bugs mailbox
> that need to be dealt with. I haven't had time to check them for
> a couple of days.

Okay.  I can't think of any other feature enhancements we should do 
before a public release of 0.7 other than the ones discussed above.

	Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com  brian@hyperreal.com  http://www.[hyperreal,organic].com/

Re: Line 751 in http_mime_db.c

[Ryan Watkins <va...@dimensionx.com>]

On Wed, 21 Jun 1995, Brian Behlendorf wrote:

> > There were some earlier problem reports with Linux and Solaris. Do the
> > problems still exist ?
> 
> Dunno - when 0.7.3g is uploaded, maybe other people can download and try 
> it out...

I just pulled down 0.7.3g a few minutes ago, and the same problems I ran
into before persist.  It seems that some children just hang and dont
answer requests at all.  I talked the server by hand and noticed it was
answering, but never responded to the GET.  When the connection is
aborted with Netscape they produce a 'socket error: accept failed' in the
error log and then the child dies and is replaced.  It seems to happen in
about a quarter of the time, and I had it running with only 4 httpds.

----

[Wed Jun 21 17:28:05 1995] socket error: accept failed
[Wed Jun 21 17:28:05 1995] could not get local address
[Wed Jun 21 17:28:05 1995] httpd: caught SIGSEGV, dumping core
[Wed Jun 21 17:28:05 1995] httpd: parent replaced dead child #2724 with #2725
[Wed Jun 21 17:29:09 1995] socket error: accept failed
[Wed Jun 21 17:29:09 1995] could not get local address
[Wed Jun 21 17:29:09 1995] httpd: caught SIGSEGV, dumping core
[Wed Jun 21 17:29:10 1995] httpd: parent replaced dead child #2723 with #2730

---- 

There is also a warning during compilation of 

gcc -c -O2 -DVIRTUAL_HOST -DSOLARIS2 httpd.c
httpd.c: In function `child_main':
httpd.c:530: warning: passing arg 2 of `accept' from incompatible pointer type

but other than thats its a clean compile.  This is all currently under
Solaris 2.4 on a Sparc2 with multiple interfaces.  Has anyone else had a
chance to try 0.7 under Solaris 2.4 yet? 


Ryan L. Watkins                email: vamp@dnx.com
HTML Hacker / System Admin       url: http://www.dnx.com/vamp/vamp.html
Dimension X, Inc.              phone: 415.296.0100