Marc

Thanks for the fix! I didn't actually replace (print) with (display). My hack was:

(begin ;; the (print) isn't -- don't know what the issue is, but it ;; sure doesn't work the way we want -- but then using (display) ;; does work. (display version port) (display " 200 OK" port) (display eol port) (display "Content-Length: " port) (display (u8vector-length message) port) (display eol port) (display "Content-Type: text/html; charset=ISO-8859-1" port) (display eol port) (display "Connection: close" port) (display eol port) (display eol port) )

I'll implement your fix, instead of my hack (it sure is cleaner). Thanks again!

Offtopic - you had mentioned future support of native threading. I have been using the Unix fork() primitive to use multi-core. This forces IPC using messages (or sockets). As part of this, I started playing around with Unix shm (shared memory). The problem is that I really couldn't get more performance, because data needs to be copied to the shared memory. It struck me that to get the performance I was after the copy would have to be removed. In other words, (shared) Scheme objects would have to reside *in* the shared memory region.

Just wondering what direction you are contemplating with Gambit-C (or am I completely off-base)?

FredW

On Tue, 2013-01-22 at 11:42 -0500, Marc Feeley wrote:

On 2013-01-22, at 11:14 AM, Fred Weigel fred_weigel@hotmail.com wrote:

...

A sample web server is supplied with Gambit-C. I am using gsi 4.6.6

[fred@dejah Gambit-C]$ gsi Gambit v4.6.6

...

and tried the web server (for inclusion into some of my own code). I do like the design, but there seems to be a bug. In http.scm line 827, there is a reference to function (print). The problem is that (print) doesn't produce ANY output on "port", and the contents are then interpreted as a flat text file. I replaced (print) with (display), and the thing works (as expected).

Running on Fedora 17 Linux, with the invocation

gsi base64 html http web-server 8000

(as recommended).

Strange that this incorrect call to print went unnoticed for so long... You can't simply replace the call to print by a call to display. The correct fix is:

            (print
             port: port
             (list version " 200 OK" eol
                   "Content-Length: " (u8vector-length message) eol
                   "Content-Type: text/html; charset=ISO-8859-1" eol
                   "Connection: close" eol
                   eol))

Thanks for reporting the problem!

Marc