On 12 October 2001, Johann Visagie said:
> Expert, no.  But the DOS/Windows null device is indeed NUL.  I've just
> checked that this still exists on Win2K, at least.

Great.  Can someone intrepid Quixote-on-Windows user try a few things
for us:

  * set DEBUG_LOG = None in the config file and start a Quixote
    application (eg. access "/q/" if you've installed the demo
    according to the instructions in doc/demo.txt)

  * if Windows does not recognize "/dev/null" (as we suspect), it should
    crash

(If it does *not* crash, please try this in a Python interactive session:

  f = open("/dev/null", "a")
  print >>f, "to the bit bucket"

I'm curious what happens.)

Then try applying this patch:

--- publish.py  2001/10/09 21:38:15     1.99
+++ publish.py  2001/10/12 14:25:09
@@ -44,2 +44,11 @@

+if os.name == "posix":
+    NULL_FILE = "/dev/null"
+elif os.name == "nt":
+    NULL_FILE = "NUL"
+else:
+    # Will bomb if-and-only-if we enter setup_logs() with
+    # config.debug_log unset.
+    NULL_FILE = None
+
 class Publisher:
@@ -136,3 +145,3 @@
             # to the bit-bucket.
-            sys.stdout = open('/dev/null', 'w')
+            sys.stdout = open(NULL_FILE, 'w')
         else:

...and see if it works (ie. doesn't crash) this time.

Thanks --

        Greg
--
Greg Ward - software developer                gward@mems-exchange.org
MEMS Exchange                            http://www.mems-exchange.org