Re: new new beta (Re: new beta released)

At 29.10.2003 17:28 +0100, you wrote:
>On Tue, Oct 28, 2003 at 12:21:20PM +0100, [email protected] wrote:
> > > On Mon, Oct 27, 2003 at 03:24:46PM +0100, [email protected] wrote:
> > > > > I'm sorry to say, but b6 breaks something in the CGI stuff.
> > > >
> > > > Cool! A bug. :) I'll try to reproduce. What operating system are you
> > > > using?
> > >
> > > CRUX Linux, glibc 2.3.2, gcc 3.2.3, binutils 2.14, kernel 2.4.22
> >
> > Ok I think I spotted it. What happens is that linux appears to set then
> > POLLHUP flag whenever an end-of-file is received on a descriptor. This is
> > a problem for mathopd because it does not expect this.
> >
> > I will release b7 within a few minutes, which should solve the problem.
>
>Most of all, thanks ! Unforunately one of my favourite application
>(cvstrac, http://www.cvstrac.org/) won't work, with b5 it does.

the bug occurs on line 288 in stub.c
the message is printed when the script outputs the content-length itself.

quick and dirty hack:
    if (/* *cp != '\n' || */ul >= UINT_MAX) {


i had no time to check this closer, but the hack works for me


regards,
    thomas


>I got the following error messages (timestamp removed):
>
>sh-2.05b# mathopd -n -d -f /etc/mathopd.conf
>[18564] *** Mathopd/1.5b8 starting
>[18564] fork_request: child process 18839 created
>[18564] convert_cgi_headers: illegal content-length header 
>"Content-Length: 1665"
>[18564] pipe_run: problem in convert_cgi_headers
>[18564] child process 18839 exited with status 0
>[18564] timeout to 192.168.0.1[34607]
>
>Similar messages also with my eruby scripts. Anyway, many thanks for
>your work, Michiel
>
>Greetings
>Jürgen
>
>
>--
>[email protected]