Re: [cpp-netlib-devel] New http::client code errors

[Dean M. B. <mik...@gm...>]

Hi Divye

On Sun, Aug 10, 2008 at 7:41 PM, Divye Kapoor <div...@gm...> wrote:
> Hi Dean,
>    Response inline.
>
> On Sat, Aug 9, 2008 at 7:14 PM, Dean Michael Berris <mik...@gm...>
> wrote:
>>
>> Hi Guys,
>>
>> I've just checked in revision 48 which fixes the Python CGI script
>> that returns incomplete headers. I had to make it print the http
>> status line (which I'm not sure is required) and add the correct line
>> endings to the printed header: instead of just '\n', I made it print
>> "\r\n\r\n" in the end to signify that it was the last header responded
>> to.
>
> I tested the python server using  LiveHTTPHeaders addon to Firefox. The
> current output from the script (after your commit) is: (for
> http://localhost:8000/cgi-bin/requestinfo.py?q=1111)
>
> Content-type: text/plain; charset=us-ascii
>
>
> Query string: q=1111
> GET parameters: q - ['1111'] ;
> POST parameters:
>
> Clearly, the Content-type header is being shifted to the body. Please verify
> this if you can. I have changed the print statement to use sys.stdout.write
> to write unformatted strings, the print statement interprets the \n on
> windows to be \r\n - thus causing this issue. Hopefully it should be fixed.
> What was causing the CGI Test to fail initially? Was it the missing HTTP
> header?
>

The CGI test caused the localhost_test to fail because without the
'\r\n\r\n' combination from the HTTP response (that delimits the end
of the headers section), client.get(...) throws an 'end of file'
exception in boost::asio::read_until(..., "\r\n\r\n"). I added the
HTTP headers thinking it must be missing or the culprit, but
apparently just adding '\r\n\r\n' instead of '\n' to the Content-type
header should suffice.

I'm not sure if it's a bug in the Python CGI Server implementation, or
has something to do with Windows now handling the end-of-line
delimiters correctly.

I'm not sure also how to fix this, so I'm leaving it up to you. I just
did an update now and I see that the localhost_test fails in Windows
again, with the client throwing the same end-of-file exception.

>>
>> Divye, please make sure that this is the correct way to do CGI scripts
>> in Python -- I really have no idea how to do that, and I only hacked
>> the script to make it not fail in just the CGI case. I haven't tried
>> this in Windows yet, but I'm thinking it should be do-able.
>
> Hopefully this should be fixed.
>

I think it's fixed now too. Thanks. :)

>>
>> Another thing: maybe we should put the startup/shutdown of the server
>> in a global fixture in the test suite for localhost_test? Has anybody
>> done this with Boost.Test before? That way when we start the server,
>> we should be able to stop it as well at the end of the whole test.
>
> That's a good idea. I just learnt about them. I'll try my hand at it.
>

Hope to see your check-in soon. :)

BTW, I'm thinking of using fixtures in our unit tests so that we cut
down on repetitive code. I'll add more tests to the message framework
to show what I mean later.

For the meantime, I'm going to implement the head function call to
support HEAD queries from the client side.

HTH

-- 
Dean Michael C. Berris
Software Engineer, Friendster, Inc.