Michael Sperber writes:
> Emilio Lopes <eclig@gmx.net> writes:
>> Closing the request's input-port (in `get-bindings') results in a core
>> dump of the SUNet server. Ouch!
> Backtrace?
This was on Windows/Cygwin and it seems that I can't reproduce this
(and another "interesting" problem that I'm having) in a decent
environment.
It would be interesting to check if closing the request's input-port
(with the server running on a decent OS) does solve the problem with
Explorer. It would be a better/simple solution than twiddling the
buffering. But I don't have Windows at home and no Unix access at
work, so I can't test this. Anyone?
Also PLT's server has some code to cope with an Explorer bug. From
collects/web-server/private/request.ss:
; to work around a bug in MSIE for documents < 265 bytes when connecting
from the local
; machine. The server could pad the response as MSIIS does, but closing the
connection works, too.
; We do not check for version numbers since IE 6 under windows is 5.2 under
macosX
But this doesn't seem to be the same bug I encountered because I was
also having this problem when connection from another machine and it
was independent of the document size.
Oh, well...
--
Emílio C. Lopes
Munich, Germany
|